id
stringlengths 27
33
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 13
1.81M
|
|---|---|---|---|
warning/0001/astro-ph0001192.html | ar5iv | text | # An Instability-driven Dynamo for Gamma-ray Burst Black Holes
## I Introduction
The combined redshift and fluence measurements of at least five $`\gamma `$-ray burst sources plus very large photon energy detections in certain bursts and tight size constraints derived from the rapid risetimes of burst triggers strongly suggest that the release of energy is highly focused by the central engine that propels a $`\gamma `$-ray burst. In spite of the very large $`\gamma `$-ray energy requirements, the efficiency of energy deposition into electromagnetic channels is likely to be very poor if the burst is driven by a neutrino burst in analogy to the processes thought to give birth to supernovæ (MacFadyen & Woosley 1998, 1999) or if the burst involves major energy losses to gravitational radiation such as might be the case in compact object merger scenarios (Rasio and Shapiro 1994, Davies et al. 1994, Ruffert & Janka 1998). Indeed, these measurements pose a serious energy budget problem for arguably all gravitational collapse powered models of $`\gamma `$-ray bursts if the energy release is not moderately collimated.
An attractive solution to this problem starts with a (directional) Poynting-flux dominated outflow (Thompson 1994, Mészáros & Rees 1997) under the premise that such a flow may carry very little baryonic contamination if deposited along a centrifugally (or gravitationally) evacuated funnel such as the angular momentum axis of a black hole-accretion disk system. Yet, formal motivation for an external field of the desired strength and topology has yet to be investigated in this setting.
We motivate a reasonable set of heuristic two dimensional dynamo equations for magnetic field components in the comoving frame of a mildly advective disk under the premise of negligible generation of meridional field (invoking Parker’s undulate instability to promote the growth (loss) of vertical (horizontal) field is questionable in turbulent disks where the turbulence is fed by MRI’s). A self-consistent turbulent steady state is achieved when the non-linear damping rate of the turbulent cascade equals the inverse of the linear growth timescale (Zhang, Diamond & Vishniac 1994). Accretion disk models are far from being self-consistent in this respect.
The heuristic dynamo equations account for field generation by the shear flow and by the non-axisymmetric MRI in Paczyński’s pseudo-potential well $`\mathrm{\Phi }(rr_g)^1`$ where $`r_g=2GM/c^2`$; with flux buoyancy and turbulent reconnection providing for field loss terms. The model predicts azimuthally averaged field components and depends explicitly on the magnetic Mach number of the turbulence (assumed Alfvenic), on local pressure ratios, and on the relativistic generalization of the shear parameter (A$`{}_{\mathrm{Oort}}{}^{}{}_{}{}^{\mathrm{Rel}}`$ = Oort’s A constant).
## II Dynamo Equations and Scalings
In Lagrangian coordinates
$$_tB_\phi =\frac{B_r}{\tau __𝒮}\frac{B_\phi }{\tau _{_{}}}\frac{B_\phi }{\tau _r}\mathrm{and}_tB_r=\frac{B_\phi }{\tau _{_{}}}\frac{B_r}{\tau _{_{}}}\frac{B_r}{\tau _r}.$$
(1)
The shear flow is parameterized linearly by a relativistic generalization of Oort’s first constant (Novikov & Thorne 1974), $`\tau __𝒮^1=`$2A$`{}_{\mathrm{Oort}}{}^{}=\gamma ^2d_{\mathrm{ln}r}\mathrm{\Omega }`$, where $`\gamma `$ is the bulk Lorentz factor of the flow. Shear forces the radial wavenumber of perturbations to evolve according to $`k_r(t)=k_r^02\mathrm{A}k_\phi t`$.
We use exact analytical MRI scalings from Foglizzo & Tagger (1995). With $`\alpha _\phi B_\phi ^2/(8\pi p)`$ and $`\widehat{\mathrm{A}}\mathrm{A}/\mathrm{\Omega }<0`$, the maximum growth of non-axisymmetric MRI modes occurs for wavenumber
$$k_\phi \eta \frac{\mathrm{\Omega }}{v_\phi ^{\mathrm{Alf}}}\stackrel{\alpha _\phi 1}{}\{2\widehat{\mathrm{A}}+(1+\alpha _\phi )\widehat{\mathrm{A}}^2\}\frac{\mathrm{\Omega }}{v_\phi ^{\mathrm{Alf}}}$$
(2)
at a rate, $`\tau _{}^1`$, that obeys
$$|\widehat{\mathrm{A}}|^2\tau _{_{}}^2\stackrel{\alpha _\phi 1}{}1+\alpha _\phi (2+\widehat{\mathrm{A}})$$
(3)
as long as $`k_\phi \text{ }>\text{ }1/r`$. On the other hand, in a very strongly sheared flow
$$\widehat{\mathrm{A}}>1+(2\alpha _\phi +1)^1,$$
(4)
the slow branch of MHD propagation (to which the toroidal MRI belongs) is destabilized into a radial interchange mode (T. Foglizzo, Priv. Comm.) in accordance with the Rayleigh criterium.
A plausible mechanism that promotes baryon unloading from field lines is turbulent pumping (Vishniac 1995a) by the MRI which must favorably compete with turbulent diffusion of matter back onto flux ropes. Under this assumption, the stretch, twist and fold of field lines by (enthalpy-weighted) sub-Alfvenic turbulence augments the field energy density and releases matter from field lines that otherwise would be “frozen-in”. For the marginal case of Alfvenic turbulence, nearly empty B flux ropes (i.e. flux residing on surfaces perpendicular to the local meridian) in a gas pressure dominated disk acquire a drag limited buoyant velocity $`v_b(v_\theta ^{\mathrm{Alf}})^2/c_s`$. Moreover, assuming efficient diffusion of radiation and $`e^\pm `$ pairs into the flux tubes, in this picture the buoyancy loss rate, $`\tau _{}^1=v_b/_\theta `$, is enhanced by a factor $`\text{ }<\text{ }p/p_{\mathrm{gas}}\xi `$ (Vishniac 1995b).
Reconnection of the field at sub-MRI optimal lengthscales (where the field lines are only weakly stochastic) occurs at the Alfvén speed for Alfvenic turbulence (Lazarian & Vishniac 1999). The rates may be written as inverse Alfvén transit times calculated from the component of the field that undergoes reconnection, $`\tau _{\mathrm{rec}}l_{\mathrm{rec}}^i/v_i^{\mathrm{Alf}}=\sqrt{2/\mathrm{\Gamma }}(l_{\mathrm{rec}}^i/_\mathrm{\Theta })(c_s/v_i^{\mathrm{Alf}})\mathrm{\Omega }^1`$, where $`\mathrm{\Gamma }`$ is the adiabatic index of the fluid.
The perpendicular lengthscales, $`l_{\mathrm{rec}}^i`$, associated with the mean distance for field reversal are derived from the fundamental linear lengthscales for coherent field pumping. These are supplied by half of the toroidal MRI’s (wave)length scale. Azimuthal, radial field reversal, $`Y_r`$, directly involves the optimal wavenumber of the non-axisymmetric MRI. In addition, following Tout & Pringle (1992), an estimate of the radial, azimuthal field reversal lengthscale, $`X_\phi `$ follows by noting that the time evolution of wavenumbers implied by shear during one MRI timescale couples the azimuthal lengthscale to the radial lengthscale, i.e. $`l_y^{}=(\tau _{_{}}/\tau __𝒮)\times l_x^{}`$. Thus $`Y_r=\pi /k_{_{}}\mathrm{and}X_\phi =(\tau __𝒮/\tau _{_{}})\times (\pi /k_{_{}})`$.
## III Equilibrium Solutions
Scaling wavenumbers to the inverse pressure scale height $`k\widehat{k}/_\mathrm{\Theta }`$, and redefining $`\widehat{\mathrm{A}}|\mathrm{A}|/\mathrm{\Omega }>0`$; a set of normalized dynamo equations follows by representing fields in (Alfvén ) velocity units and normalized to the soundspeed ($`B^{}=B\times \sqrt{4\pi \varrho }c_s`$), and the time normalized to the inverse of the Keplerian frequency $`t^{}=\mathrm{\Omega }t`$.
In a steady state, these equations must satisfy
$$B_r^{}=\frac{B_\phi ^{}}{2\widehat{\mathrm{A}}}\left\{\frac{1}{\xi }B_{}^{}{}_{}{}^{2}+\frac{\sqrt{2}}{\pi }\frac{\tau _{_{}}^{}}{\tau __𝒮^{}}\eta \right\},\mathrm{and}B_\phi ^{}=\frac{B_r^{}}{\widehat{\mathrm{A}}}\left\{\frac{1}{\xi }B_{}^{}{}_{}{}^{2}+\frac{\sqrt{2}}{\pi }\frac{B_r^{}}{B_\phi ^{}}\eta \right\}$$
(5)
which we solve numerically.
## IV The energy deposition rate
The accretion disk setting is envisioned to follow the standard hyper-accreting black hole model of Popham, Woosley & Fryer (1999, hereafter PWF) where $`M_{\mathrm{bh}}=3\mathrm{M}_{};\alpha _{\mathrm{SS}}=0.1,\mathrm{and}\dot{M}=0.1\mathrm{M}_{}\mathrm{sec}^1`$. By adopting their published pressure ratios and assuming Keplerian rotation for $`r[2.25,20]r_g`$, we “piggyback” the hydromagnetic energy conversion process on this model.
We find that the magnetic output rate from buoyancy, $`\dot{E}1.77_{+51}`$ erg sec<sup>-1</sup>, is comparable to the neutrino luminosity $`L_\nu 3.3_{+51}`$ erg sec<sup>-1</sup> (PWF). The ‘half-luminosity’ radius is located at $`r_{}5.75r_g`$ and the IDD becomes operational (against the radial interchange instability, c.f. Eq and Araya-Gochez 1999) at $`r_{\mathrm{min}}2.55r_g`$. Curiously, the derived value of the magnetic viscosity $`\alpha _{\mathrm{SS}}=B_r^{}B_\phi ^{}`$ hovers on 0.1 at the innermost radii and decreases to about .087 at $`r_{}`$ in good agreement with the adopted value. Lastly, we note that the MRI pumps the field preferably at the lowest wavenumbers for a near-equipartition field in a strongly sheared flow. Thus, most of this energy could in principle go into a Poynting jet with interesting consequences for $`\gamma `$-ray bursts. |
warning/0001/cond-mat0001260.html | ar5iv | text | # Pseudogap effects on the c-axis charge dynamics in copper oxide materials
## I Introduction
It has become clear in the past several years that copper oxide materials are among the most complex systems studied in condensed matter physics, and show many unusual normal-state properties. The complications arise mainly from (1) strong anisotropy in the properties parallel and perpendicular to the CuO<sub>2</sub> planes which are the key structural element in the whole copper oxide superconducting materials, and (2) extreme sensitivity of the properties to the compositions (stoichiometry) which control the carrier density in the CuO<sub>2</sub> plane , while the unusual normal-state feature is then closely related to the fact that these copper oxide materials are doped Mott insulators, obtained by chemically adding charge carriers to a strongly correlated antiferromagnetic (AF) insulating state, therefore the physical properties of these systems mainly depend on the extent of dopings, and the regimes have been classified into the underdoped, optimally doped, and overdoped, respectively . The normal-state properties of copper oxide materials in the underdoped and optimally doped regimes exhibit a number of anomalous properties in the sense that they do not fit in the conventional Fermi-liquid theory , and the mechanism for the superconductivity in copper oxide materials has been widely recognized to be closely associated with the anisotropic normal-state properties . Among the striking features of the normal-state properties in the underdoped and optimally doped regimes, the physical quantity which most evidently displays the anisotropic property in copper oxide materials is the charge dynamics , which is manifested by the optical conductivity and resistivity. It has been show from the experiments that the in-plane charge dynamics is rather universal within the whole copper oxide materials . The in-plane optical conductivity for the same doping is nearly materials independent both in the magnitude and energy dependence, and shows the non-Drude behavior at low energies and anomalous midinfrared band in the charge-transfer gap, while the in-plane resistivity $`\rho _{ab}(T)`$ exhibits a linear behavior in the temperature in the optimally doped regime and a nearly temperature linear dependence with deviations at low temperatures in the underdoped regime . By contrast, the magnitude of the c-axis charge dynamics in the underdoped and optimally doped regimes is strongly materials dependent, i.e., it is dependent on the species of the building blocks in between the CuO<sub>2</sub> planes . In the underdoped and optimally doped regimes, the experimental results show that the ratio $`R=\rho _c(T)/\rho _{ab}(T)`$ ranges from $`R100`$ to $`R>10^5`$, this large magnitude of the resistivity anisotropy reflects that the c-axis mean free path is shorter than the interlayer distance, and the carriers are tightly confined to the CuO<sub>2</sub> planes, and also is the evidence of the incoherent charge dynamics in the c-axis direction. For the copper oxide materials without the Cu-O chains in between the CuO<sub>2</sub> planes , such as La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub> systems, the transferred weight in the c-axis conductivity forms a band peaked at high energy $`\omega 2ev`$, and the low-energy spectral weight is quite small and spread over a wide energy range instead of forming a peak at low energies, in this case the behavior of the c-axis temperature dependent resistivity $`\rho _c(T)`$ is characterized by a crossover from the high temperature metallic-like to the low temperature semiconducting-like . However, for these copper oxide materials with the Cu-O chains in between the CuO<sub>2</sub> planes , such as YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> systems, the c-axis conductivity exhibits the non-Drude behavior at low energies and weak midinfrared band, moreover, this weak midinfrared band rapidly decrease with reducing dopings or increasing temperatures, while the c-axis resistivity $`\rho _c(T)`$ is linear in temperatures in the optimally doped regime, and shows a crossover from the high temperature metallic-like behavior to the low temperature semiconducting-like behavior in the underdoped regime . Therefore there are some subtle differences between the chain and no-chain copper oxide materials.
The c-axis charge dynamics of copper oxide materials has been addressed from several theoretical viewpoints . Based on the concept of dynamical dephasing, Leggett thus proposed that the c-axis conduction has to do with scatterings from in-plane thermal fluctuations, and depends on the ratio of the interlayer hopping rate of CuO<sub>2</sub> sheets to the thermal energy. While the theory of tunneling c-axis conductivity in the incoherent regime has been given by many researchers . Based on a highly anisotropic Fermi-liquid, some effect from the interlayer static disorder or dynamical one has been discussed . The similar incoherent conductivity in the coupled fermion chains has been in more detail studied by many authors within the framework of the non-Fermi-liquid theory . Moreover, the most reliable result for the c-axis charge dynamics from the model relevant to copper oxide materials has been obtained by the numerical simulation . It has been argued that the in-plane resistivity deviates from the temperature linear behavior and temperature coefficient of the c-axis resistivity change sign, showing semiconducting-like behavior at low temperatures are associated with the effect of the pseudogap . To shed light on this issue, we, in this paper, apply the fermion-spin approach to study the c-axis charge dynamics by considering the interlayer coupling.
The paper is organized as follows. The theoretical framework is presented in Sec. II. In the case of the incoherent interlayer hopping, the c-axis current-current correlation function (then the c-axis optical conductivity) is calculated in terms of the in-plane single-particle spectral function by using standard formalisms for the tunneling in metal-insulator-metal junctions . Within this theoretical framework, we discuss the c-axis charge dynamics of the chain copper oxide materials in Sec. III. It is shown that the c-axis charge dynamics of the chain copper oxide materials is mainly governed by the scattering from in-plane charged holons due to spinon fluctuations, and the behavior of the c-axis resistivity is the metallic-like in the optimally doped regime and the semiconducting-like in the underdoped regime at low temperatures. In Sec. IV, the c-axis charge dynamics of the no-chain copper oxide materials is discussed. Our result shows that the scattering from the in-plane fluctuation incorporating with the interlayer disorder dominates the c-axis charge dynamics for the no-chain copper oxide materials. In this case, the c-axis resistivity exhibits the semiconducting-like behavior in the underdoped and optimally doped regimes at low temperatures. Sec. V is devoted to a summary and discussions. Our results also show that the crossover to the semiconducting-like range in $`\rho _c(T)`$ is obviously linked with the crossover from the temperature linear to the nonlinear range in $`\rho _{ab}(T)`$, and the common origin for these crossovers is due to the existence of the holon pseudogap at low temperatures and lower doping levels.
## II Theoretical framework
Among the microscopic models the most simplest for the discussion of doped Mott insulators is the $`t`$-$`J`$ model , which is originally introduced as an effective Hamiltonian of the Hubbard model in the strong coupling regime, where the electron become strongly correlated to avoid the double occupancy. The interest in the $`t`$-$`J`$ model is stimulated by many researchers’ suggestions that it may contain the essential physics of copper oxide materials . On the other hand, there is a lot of evidence from the experiments and numerical simulations in favour of the $`t`$-$`J`$ model as the basic underlying microscopic model . Within each CuO<sub>2</sub> plane, the physics property may be described by the two-dimensional (2D) $`t`$-$`J`$ model,
$`H_l=t{\displaystyle \underset{i\widehat{\eta }\sigma }{}}C_{li\sigma }^{}C_{li+\widehat{\eta }\sigma }+h.c.\mu {\displaystyle \underset{i\sigma }{}}C_{li\sigma }^{}C_{li\sigma }+J{\displaystyle \underset{i\widehat{\eta }}{}}𝐒_{li}𝐒_{li+\widehat{\eta }},`$ (1)
supplemented by the on-site local constraint $`_\sigma C_{li\sigma }^{}C_{li\sigma }1`$ to avoid the double occupancy, where $`\widehat{\eta }=\pm a_0\widehat{x},\pm a_0\widehat{y}`$, $`a_0`$ is the lattice constant of the square planar lattice, which is set as the unit hereafter, $`i`$ refers to planar sites within the l-th CuO<sub>2</sub> plane, $`C_{li\sigma }^{}`$ ($`C_{li\sigma }`$) are the electron creation (annihilation) operators, $`𝐒_{li}=C_{li}^{}\sigma C_{li}/2`$ are the spin operators with $`\sigma =(\sigma _x,\sigma _y,\sigma _z)`$ as the Pauli matrices, and $`\mu `$ is the chemical potential. Then the hopping between CuO<sub>2</sub> planes is considered as
$`H=t_c{\displaystyle \underset{l\widehat{\eta }_ci\sigma }{}}C_{li\sigma }^{}C_{l+\widehat{\eta }_ci\sigma }+h.c.+{\displaystyle \underset{l}{}}H_l,`$ (2)
where $`\widehat{\eta }_c=\pm c_0\widehat{z}`$, $`c_0`$ is the interlayer distance, and has been determined from the experiments as $`c_0>2a_0`$. As mentioned above, the experimental results show that the c-axis charge dynamics in the underdoped and optimally doped regimes is incoherent, therefore the c-axis momentum can not be defined . Moreover, the absence of the coherent c-axis charge dynamics is a consequence of the weak interlayer hopping matrix element $`t_c`$, but also of a strong intralayer scattering, i.e., $`t_ct`$, and therefore the common CuO<sub>2</sub> planes in copper oxide materials clearly dominate the most normal-state properties. In this case, the most relevant for the study of the c-axis charge dynamics is the results on the in-plane conductivity $`\sigma _{ab}(\omega )`$ and related single-particle spectral function $`A(k,\omega )`$.
Since the strong electron correlation in the $`t`$-$`J`$ model manifests itself by the electron single occupancy on-site local constraint, then the crucial requirement is to impose this electron on-site local constraint for a proper understanding of the physics of copper oxide materials. To incorporate this local constraint, the fermion-spin theory based on the charge-spin separation has been proposed . According to the fermion-spin theory, the constrained electron operators in the $`t`$-$`J`$ model is decomposed as ,
$`C_{li}=h_{li}^{}S_{li}^{},C_{li}=h_{li}^{}S_{li}^+,`$ (3)
with the spinless fermion operator $`h_i`$ keeps track of the charge (holon), while the pseudospin operator $`S_i`$ keeps track of the spin (spinon). The main advantage of this approach is that the electron on-site local constraint can be treated exactly in analytical calculations. In this case, the low-energy behavior of the $`t`$-$`J`$ model (2) in the fermion-spin representation can be written as ,
$`H`$ $`=`$ $`t_c{\displaystyle \underset{l\widehat{\eta }_ci}{}}h_{l+\widehat{\eta }_ci}^{}h_{li}(S_{li}^+S_{l+\widehat{\eta }_ci}^{}+S_{li}^{}S_{l+\widehat{\eta }_ci}^+)+{\displaystyle \underset{l}{}}H_l,`$ (5)
$`H_l`$ $`=`$ $`t{\displaystyle \underset{i\widehat{\eta }}{}}h_{li+\widehat{\eta }}^{}h_{li}(S_{li}^+S_{li+\widehat{\eta }}^{}+S_{li}^{}S_{li+\widehat{\eta }}^+)+\mu {\displaystyle \underset{i}{}}h_{li}^{}h_{li}+J_{eff}{\displaystyle \underset{i\widehat{\eta }}{}}(𝐒_{li}𝐒_{li+\widehat{\eta }}),`$ (6)
where $`J_{eff}=J[(1\delta )^2\varphi ^2]`$, the in-plane holon particle-hole parameter $`\varphi =h_{li}^{}h_{li+\widehat{\eta }}`$, and $`S_{li}^+`$ and $`S_{li}^{}`$ are the pseudospin raising and lowering operators, respectively. As a consequence, the kinetic part in the $`t`$-$`J`$ model has been expressed as the holon-spinon interaction in the fermion-spin representation, which dominates the charge and spin dynamics in copper oxide materials in the underdoped and optimally doped regimes. The spinon and holon may be separated at the mean-field level, but they are strongly coupled beyond mean-field approximation (MFA) due to fluctuations.
The mean-field theory within the fermion-spin formalism in the underdoped and optimally doped regimes without AF long-range-order (AFLRO) has been developed , and the in-plane mean-field spinon and holon Green’s functions $`D_{ab}^{(0)}(ij,\tau \tau ^{})=T_\tau S_{li}^+(\tau )S_{lj}^{}(\tau ^{})_0`$ and $`g_{ab}^{(0)}(ij,\tau \tau ^{})=T_\tau h_{li}(\tau )h_{lj}^{}(\tau ^{})_0`$ have been evaluated as,
$`D_{ab}^{(0)}(𝐤,\omega )={\displaystyle \frac{B_k}{2\omega (k)}}\left({\displaystyle \frac{1}{\omega \omega (k)}}{\displaystyle \frac{1}{\omega +\omega (k)}}\right),`$ (7)
$`g_{ab}^{(0)}(𝐤,\omega )={\displaystyle \frac{1}{\omega \xi _k}},`$ (8)
respectively, where $`B_k=\lambda [(2ϵ\chi _z+\chi )\gamma _k(ϵ\chi +2\chi _z)]`$, $`\gamma _𝐤=(1/Z)_\eta e^{i𝐤\widehat{\eta }}`$, $`\lambda =2ZJ_{eff}`$, $`ϵ=1+2t\varphi /J_{eff}`$, $`Z`$ is the number of the nearest neighbor sites at the plane, while the in-plane mean-field spinon spectrum
$`\omega ^2(k)=\lambda ^2\left(\alpha ϵ[\chi _z\gamma _k+{\displaystyle \frac{1}{2Z}}\chi ][\alpha C_z+{\displaystyle \frac{1}{4Z}}(1\alpha )]\right)(ϵ\gamma _k1)`$ (9)
$`+\lambda ^2\left(\alpha ϵ[{\displaystyle \frac{1}{2}}\chi \gamma _k+{\displaystyle \frac{1}{Z}}\chi _z]{\displaystyle \frac{1}{2}}ϵ[\alpha C+{\displaystyle \frac{1}{2Z}}(1\alpha )]\right)(\gamma _kϵ),`$ (10)
and the in-plane mean-field holon spectrum $`\xi _k=2Z\chi t\gamma _k+\mu `$, with the in-plane spinon correlation functions $`\chi =S_{li}^+S_{li+\widehat{\eta }}^{}`$, $`\chi _z=S_{li}^zS_{li+\widehat{\eta }}^z`$, $`C=(1/Z^2)_{\widehat{\eta },\widehat{\eta ^{}}}S_{li+\widehat{\eta }}^+S_{li+\widehat{\eta ^{}}}^{}`$, and $`C_z=(1/Z^2)_{\widehat{\eta },\widehat{\eta ^{}}}S_{li+\widehat{\eta }}^zS_{li+\widehat{\eta ^{}}}^z`$. In order not to violate the sum rule of the correlation function $`S_{li}^+S_{li}^{}=1/2`$ in the case without AFLRO, the important decoupling parameter $`\alpha `$ has been introduced in the mean-field calculation , which can be regarded as the vertex correction. The mean-field order parameter $`\chi `$, $`C`$, $`\chi _z`$, $`C_z`$, $`\varphi `$ and chemical potential $`\mu `$ have been determined by the self-consistent equations.
Within the 2D $`t`$-$`J`$ model, the in-plane charge dynamics in copper oxide materials has been discussed by considering fluctuations around this mean-field solution, and the result exhibits a behavior similar to that seen in the experiments and numerical simulations . In the framework of the charge-spin separation, an electron is represented by the product of a holon and a spinon, then the external field can only be coupled to one of them. According to the Ioffe-Larkin combination rule , the physical c-axis conductivity $`\sigma _c(\omega )`$ is given by,
$`\sigma _c^1(\omega )=\sigma _c^{(h)1}(\omega )+\sigma _c^{(s)1}(\omega ),`$ (11)
where $`\sigma _c^{(h)}(\omega )`$ and $`\sigma _c^{(s)}(\omega )`$ are the contributions to the c-axis conductivity from holons and spinons, respectively, and can be expressed as,
$`\sigma _c^{(h)}(\omega )={\displaystyle \frac{\mathrm{Im}\mathrm{\Pi }_c^{(h)}(\omega )}{\omega }},\sigma _c^{(s)}(\omega )={\displaystyle \frac{\mathrm{Im}\mathrm{\Pi }_c^{(s)}(\omega )}{\omega }},`$ (12)
with $`\mathrm{\Pi }_c^{(h)}(\omega )`$ and $`\mathrm{\Pi }_c^{(s)}(\omega )`$ are the holon and spinon c-axis current-current correlation function, respectively, which are defined as,
$`\mathrm{\Pi }_c^{(s)}(\tau \tau ^{})=T_\tau j_c^{(s)}(\tau )j_c^{(s)}(\tau ^{}),\mathrm{\Pi }_c^{(h)}(\tau \tau ^{})=T_\tau j_c^{(h)}(\tau )j_c^{(h)}(\tau ^{}).`$ (13)
Within the Hamiltonian (4), the c-axis current densities of spinons and holons are obtained by the time derivation of the polarization operator using Heisenberg’s equation of motion as, $`j_c^{(h)}=2\stackrel{~}{t}_ce\chi _{l\widehat{\eta }_ci}\widehat{\eta }_ch_{l+\widehat{\eta }_ci}^{}h_{li}`$ and $`j_c^{(s)}=t_ce\varphi _c_{l\widehat{\eta }_ci}\widehat{\eta }_c(S_{li}^+S_{l+\widehat{\eta }_ci}^{}+S_{li}^{}S_{l+\widehat{\eta }_ci}^+)`$, respectively, with $`\stackrel{~}{t}_c=t_c\chi _c/\chi `$ is the effective interlayer holon hopping matrix element, and the order parameters $`\chi _c`$ and $`\varphi _c`$ are defined as $`\chi _c=S_{li}^+S_{l+\widehat{\eta }_ci}^{}`$, and $`\varphi _c=h_{li}^{}h_{l+\widehat{\eta }_ci}`$, respectively. As in the previous discussions , a formal calculation for the spinon part shows that there is no the direct contribution to the current-current correlation from spinons, but the strongly correlation between holons and spinons is considered through the spinon’s order parameters entering in the holon part of the contribution to the current-current correlation, therefore the charge dynamics in copper oxide materials is mainly caused by charged holons within the CuO<sub>2</sub> planes, which are strongly renormalized because of the strong interaction with fluctuations of surrounding spinon excitations.
In the case of the incoherent charge dynamics in the c-axis direction, i.e., the independent electron propagation in each layer, the c-axis holon current-current correlation function is then proportional to the tunneling rate between just two adjacent planes, and can be calculated in terms of the in-plane holon Green’s function $`g_{ab}(k,\omega )`$ by using standard formalisms for the tunneling in metal-insulator-metal junctions as,
$`\mathrm{\Pi }_c^{(h)}(i\omega _n)=(4\stackrel{~}{t}_ce\chi c_0)^2{\displaystyle \frac{1}{N}}{\displaystyle \underset{k}{}}{\displaystyle \frac{1}{\beta }}{\displaystyle \underset{i\omega _m^{}}{}}g_{ab}(k,i\omega _m^{}+i\omega _n)g_{ab}(k,i\omega _m^{}),`$ (14)
where $`i\omega _n`$ is the Matsubara frequency. Therefore the c-axis current-current correlation function is essentially determined by the property of the in-plane full holon propagator $`g_{ab}(k,i\omega _n)`$, which can be expressed as,
$`g_{ab}(k,i\omega _n)={\displaystyle \frac{1}{g_{ab}^{(0)1}(k,i\omega _n)\mathrm{\Sigma }_h^{(2)}(k,i\omega _n)}}={\displaystyle \frac{1}{i\omega _n\xi _k\mathrm{\Sigma }_h^{(2)}(k,i\omega _n)}},`$ (15)
where the holon self-energy has been obtained by considering the second-order correction due to the spinon pair bubble as ,
$`\mathrm{\Sigma }_h^{(2)}(k,i\omega _n)=(Zt)^2{\displaystyle \frac{1}{N^2}}{\displaystyle \underset{pp^{}}{}}(\gamma _{p^{}k}+\gamma _{p^{}+p+k})^2{\displaystyle \frac{B_p^{}B_{p+p^{}}}{4\omega _p^{}\omega _{p+p^{}}}}\times `$ (16)
$`(2{\displaystyle \frac{n_F(\xi _{p+k})[n_B(\omega _p^{})n_B(\omega _{p+p^{}})]n_B(\omega _{p+p^{}})n_B(\omega _p^{})}{i\omega _n+\omega _{p+p^{}}\omega _p^{}\xi _{p+k}}}`$ (17)
$`+{\displaystyle \frac{n_F(\xi _{p+k})[n_B(\omega _{p+p^{}})n_B(\omega _p^{})]+n_B(\omega _p^{})n_B(\omega _{p+p^{}})}{i\omega _n+\omega _p^{}+\omega _{p+p^{}}\xi _{p+k}}}`$ (18)
$`{\displaystyle \frac{n_F(\xi _{p+k)}[n_B(\omega _{p+p^{}})n_B(\omega _p^{})]n_B(\omega _p^{})n_B(\omega _{p+p^{}})}{i\omega _n\omega _{p+p^{}}\omega _p^{}\xi _{p+k}}}),`$ (19)
with $`n_F(\xi _k)`$ and $`n_B(\omega _k)`$ are the Fermi and Bose distribution functions, respectively. For the convenience in the following discussions, the above full holon in-plane Green’s function $`g_{ab}(k,i\omega _n)`$ also can be expressed as frequency integrals in terms of the spectral representation as,
$`g_{ab}(k,i\omega _n)={\displaystyle _{\mathrm{}}^{\mathrm{}}}{\displaystyle \frac{d\omega }{2\pi }}{\displaystyle \frac{A_h(k,\omega )}{i\omega _n\omega }},`$ (20)
with the in-plane holon spectral function $`A_h(k,\omega )=2\mathrm{I}\mathrm{m}g_{ab}(k,\omega )`$. Then the c-axis optical conductivity in the present theoretical framework is expressed as $`\sigma _c(\omega )=\mathrm{Im}\mathrm{\Pi }_c^{(h)}(\omega )/\omega `$.
## III Charge dynamics for the chain copper oxide materials
We firstly consider the chain copper oxide materials. From the experiments testing the c-axis charge dynamics , it has been shown that the presence of the rather conductive Cu-O chains in the underdoped and optimally doped regimes can reduce the blocking effect, and therefore the c-axis charge dynamics in this system is effected by the same electron interaction as that in the in-plane. In this case, we substitute Eq. (14) into Eq. (11), and evaluate the frequency summations, then the c-axis optical conductivity for the chain copper oxide materials can be obtained as ,
$`\sigma _c(\omega )={\displaystyle \frac{1}{2}}(4\stackrel{~}{t}_ce\chi c_0)^2{\displaystyle \frac{1}{N}}{\displaystyle \underset{k}{}}{\displaystyle _{\mathrm{}}^{\mathrm{}}}{\displaystyle \frac{d\omega ^{}}{2\pi }}A_h(k,\omega ^{}+\omega )A_h(k,\omega ^{}){\displaystyle \frac{n_F(\omega ^{}+\omega )n_F(\omega ^{})}{\omega }},`$ (21)
where the in-plane momentum is conserved. This c-axis conductivity $`\sigma _c(\omega )`$ has been calculated numerically, and the result at the doping $`\delta =0.12`$ (solid line), $`\delta =0.09`$ (dashed line), and $`\delta =0.06`$ (dot-dashed line) for the parameters $`t/J=2.5`$, $`\stackrel{~}{t}_c/t=0.04`$, and $`c_0/a_0=2.5`$ at the temperature $`T`$=0 is shown in Fig. 1, where the charge $`e`$ has been set as the unit. From Fig. 1, we find that there is two bands in $`\sigma _c(\omega )`$ separated at $`\omega 0.4t`$, the higher-energy band, corresponding to the ”midinfrared band” in the in-plane optical conductivity $`\sigma _{ab}(\omega )`$ , shows a broad peak at $`\omega 0.7t`$, in particular, the weight of this band is strongly doping dependent, and decreasing rapidly with dopings, but the peak position does not appreciably shift to higher energies. On the other hand, the transferred weight of the lower-energy band forms a sharp peak at $`\omega <0.4t`$, which can be described formally by the non-Drude formula, and our analysis indicates that this peak decay is $`1/\omega `$ at low energies as in the case of $`\sigma _{ab}(\omega )`$ . In comparison with $`\sigma _{ab}(\omega )`$ , the present result also shows that the values of $`\sigma _c(\omega )`$ are by $`23`$ orders of magnitude smaller than those of $`\sigma _{ab}(\omega )`$ in the corresponding energy range. The finite temperature behavior of $`\sigma _c(\omega )`$ also has been discussed, and the result shows that $`\sigma _c(\omega )`$ is temperature dependent, the higher-energy band is severely suppressed with increasing temperatures, and vanishes at higher temperatures. These results are qualitatively consistent with the experimental results of the chain copper oxide materials and numerical simulations .
With the help of the c-axis conductivity, the c-axis resistivity can be obtained as $`\rho _c=1/lim_{\omega 0}\sigma _c(\omega )`$, and the numerical result at the doping $`\delta =0.12`$ and $`\delta =0.06`$ for the parameters $`t/J=2.5`$, $`\stackrel{~}{t}_c/t=0.04`$, and $`c_0/a_0=2.5`$ is shown in Fig. 2(a) and Fig. 2(b), respectively. In the underdoped regime, the behavior of the temperature dependence of $`\rho _c(T)`$ shows a crossover from the high temperature metallic-like ($`d\rho _c(T)/dT>0`$) to the low temperature semiconducting-like ($`d\rho _c(T)/dT<0`$), but the metallic-like temperature dependence dominates over a wide temperature range. Therefore in this case, there is a general trend that the chain copper oxide materials show nonmetallic $`\rho _c(T)`$ in the underdoped regime at low temperatures. While in the optimally doped regime, $`\rho _c(T)`$ is linear in temperatures, and shows the metallic-like behavior for all temperatures. These results are also qualitatively consistent with the experimental results of the chain copper oxide materials and numerical simulation .
## IV Charge dynamics for the no-chain copper oxide materials
Now we turn to discuss the c-axis charge dynamics of the no-chain copper oxide materials. It has been indicated from the experiments that for the no-chain copper oxide materials the doped holes may introduce a disorder in between the CuO<sub>2</sub> planes, contrary to the case of the chain copper oxide materials , where the increasing doping reduces the disorder in between the CuO<sub>2</sub> planes due to the effect of the Cu-O chains. Therefore for the no-chain copper oxide materials, the disorder introduced by doped holes residing between the CuO<sub>2</sub> planes modifies the interlayer hopping elements as the random matrix elements. In this case, only the in-plane holon density of states (DOS) $`\mathrm{\Omega }_h(\omega )=1/N_kA_h(k,\omega )`$ enters the holon current-current correlation function as in disordered systems , and after a similar discussion as in Sec. II, we find that the corresponding momentum-nonconserving expression of the c-axis conductivity $`\sigma _c^{(n)}(\omega )`$ for the no-chain copper oxide materials is obtained by the replacement of the in-plane holon spectral function $`A_h(k,\omega )`$ in Eq. (15) with the in-plane holon DOS $`\mathrm{\Omega }_h(\omega )`$ as ,
$`\sigma _c^{(n)}(\omega )={\displaystyle \frac{1}{2}}(4\overline{t}_ce\chi c_0)^2{\displaystyle _{\mathrm{}}^{\mathrm{}}}{\displaystyle \frac{d\omega ^{}}{2\pi }}\mathrm{\Omega }_h(\omega ^{}+\omega )\mathrm{\Omega }_h(\omega ^{}){\displaystyle \frac{n_F(\omega ^{}+\omega )n_F(\omega ^{})}{\omega }},`$ (22)
where the $`\overline{t}_c`$ is some average of the random interlayer hopping matrix elements $`(\stackrel{~}{t}_c)_{li}`$. We have performed a numerical calculation for this c-axis conductivity $`\sigma _c^{(n)}(\omega )`$, and the result at the doping $`\delta =0.10`$ (solid line), $`\delta =0.08`$ (dashed line), and $`\delta =0.06`$ (dash-dotted line) for the parameters $`t/J=2.5`$, $`\overline{t}_c/t=0.04`$ with the temperature $`T=0`$ is plotted in Fig. 3. This result shows that the c-axis conductivity $`\sigma _c^{(n)}(\omega )`$ contains two bands, the higher-energy band, corresponding to the midinfrared band in the in-plane conductivity $`\sigma _{ab}(\omega )`$ , shows a broad peak at $`0.3t`$. The weight of this band is increased with dopings, but the peak position does not appreciably shift to lower energies. As a consequence of this pinning of the transferred spectral weight, the weight of the lower-energy band, corresponding to the non-Drude peak in $`\sigma _{ab}(\omega )`$ , is quite small and does not form a well-defined peak at low energies in the underdoped and optimally doped regimes. In this case, the conductivity $`\sigma _c^{(n)}(\omega )`$ at low energies cannot be described by the over-damped Drude-like formula even an $`\omega `$-dependence of the in-plane holon scattering rate has been taken into consideration. In comparison with the momentum-conserving $`\sigma _c(\omega )`$, our result also shows that at low energies the suppression of the momentum-nonconserving $`\sigma _c^{(n)}(\omega )`$ is due to the disordered effect introduced by doped holes residing between the CuO<sub>2</sub> planes. These results are qualitatively consistent with the experimental results of the no-chain copper oxide materials .
For the further understanding the transport property of the no-chain copper oxide materials, we have also performed the numerical calculation for the c-axis resistivity $`\rho _c^{(n)}=1/lim_{\omega 0}\sigma _c^{(n)}(\omega )`$, and the results at the doping $`\delta =0.10`$ and $`\delta =0.06`$ for the parameters $`t/J=2.5`$, $`\overline{t}_c/t=0.04`$, and $`c_0=2.5a_0`$ are shown in Fig. 4(a) and Fig. 4(b), respectively. In accordance with the c-axis conductivity $`\sigma _c^{(n)}(\omega )`$, the behavior of the c-axis resistivity $`\rho _c^{(n)}(T)`$ in the underdoped and optimally doped regimes is the semiconducting-like at low temperatures, and metallic-like at higher temperatures. In comparison with the in-plane resistivity $`\rho _{ab}(T)`$ , it is shown that the values of the c-axis resistivity $`\rho _c^{(n)}(T)`$ for the no-chain copper oxide materials are by $`34`$ orders of magnitude larger than these of the in-plane resistivity $`\rho _{ab}(T)`$ in the corresponding energy range, which are also qualitatively consistent with the experimental results of the no-chain copper oxide materials .
## V Summary and discussions
In the above discussions, the central concerns of the c-axis charge dynamics in copper oxide materials are the two dimensionality of the electron state and incoherent hopping between the CuO<sub>2</sub> planes, and therefore the c-axis charge dynamics in the present fermion-spin picture based on the charge-spin separation is mainly determined by the in-plane charged holon fluctuation for the chain copper oxide materials and the in-plane charged holon fluctuation incorporating with the interlayer disorder for the no-chain copper oxide materials. In comparison with the in-plane resistivity $`\rho _{ab}(T)`$ , it is shown that the crossover to the semiconducting-like range in $`\rho _c(T)`$ is obviously linked with the crossover from the temperature linear to the nonlinear range in $`\rho _{ab}(T)`$, i.e., they should have a common origin.
In the fermion-spin theory , the charge and spin degrees of freedom of the physical electron are separated as the holon and spinon, respectively. Although both holons and spinons contributed to the charge and spin dynamics, it has been shown that the scattering of spinons dominates the spin dynamics , while the results of the in-plane charge dynamics and present c-axis charge dynamics show that scattering of holons dominates the charge dynamics, the two rates observed in the experiments are attributed to the scattering of two distinct excitations, spinons and holons. It has been shown that an remarkable point of the pseudogap is that it appears in both of spinon and holon excitations . The present study indicates that the observed crossovers of $`\rho _{ab}`$ and $`\rho _c`$ for copper oxide materials seem to be connected with the pseudogap in the in-plane charge holon excitations, which can be understood from the physical property of the in-plane holon DOS. The numerical result of the in-plane holon DOS $`\mathrm{\Omega }_h(\omega )`$ at the doping $`\delta =0.06`$, $`\delta =0.12`$, and $`\delta =0.15`$ for the parameter $`t/J=2.5`$ at the temperature $`T`$=0 is shown in Fig. 5(a), Fig. 5(b), and Fig. 5(c), respectively. For comparison, the corresponding mean-field result (dashed line) is also shown in Fig. 5. While the in-plane holon density of states $`\mathrm{\Omega }_h(\omega )`$ in the underdoping $`\delta =0.06`$ as a function of energy for the temperature (a) $`T=0`$, (b) $`T=0.01J`$, and (c) $`T=0.2J`$ is plotted in Fig. 6. From Fig. 5 and Fig. 6, we therefore find that the in-plane holon DOS in MFA consists of the central part only, which comes from the noninteracting particles as pointed in Ref. . After including fluctuations the central part is renormalized and two side bands and a V-shape holon pseudogap near the chemical potential $`\mu `$ in the underdoped regime appear. But these two side bands are almost doping and temperature independent, while the V-shape holon pseudogap is doping and temperature dependent, and grows monotonously as the doping $`\delta `$ decreases, and disappear in the overdoped regime. Moreover, this holon pseudogap also decreases with increasing temperatures, and vanishes at higher temperatures. Since the full holon Green’s function (then the holon spectral function and DOS) is obtained by considering the second-order correction due to the spinon pair bubble, then the holon pseudogap is closely related to the spinon fluctuation. For small dopings and lower temperatures, the holon kinetic energy is much smaller than the magnetic energy, i.e., $`\delta tJ`$, in this case the magnetic fluctuation is strong enough to lead to the holon pseudogap. This holon pseudogap would reduce the in-plane holon scattering and thus is responsible for the metallic to semiconducting crossover in the c-axis resistivity $`\rho _c`$ and the deviation from the temperature linear behavior in the in-plane resistivity $`\rho _{ab}`$ . This holon pseudogap will also lead to form the normal-state gap in the system, and the similar result has been obtained from the doped kagomé and triangular antiferromagnets , where the strong quantum fluctuation of spinons due to the geometric frustration leads to the normal-state gap. With increasing temperatures or dopings, the holon kinetic energy is increased, while the spinon magnetic energy is decreased. In the region where the holon pseudogap closes, at high temperatures or at higher doping levels, the charged holon scattering would give rise to the temperature linear in-plane resistivity as well as the metallic temperature dependence of the c-axis resistivity. Our results also show that $`\rho _{ab}(T)`$ is only slightly affected by this holon pseudogap , while $`\rho _c(T)`$ is more sensitive to the underlying mechanism.
In summary, we have studied the c-axis charge dynamics of copper oxide materials in the underdoped and optimally doped regimes within the $`t`$-$`J`$ model by considering the incoherent interlayer hopping. Our result shows the c-axis charge dynamics for the chain copper oxide materials is mainly governed by the scattering from the in-plane fluctuation, and the c-axis charge dynamics for the no-chain copper oxide materials is dominated by the scattering from the in-plane fluctuation incorporating with the interlayer disorder, which would be suppressed when the holon pseudogap opens at low temperatures and lower doping levels, leading to the crossovers to the semiconducting-like range in the c-axis resistivity $`\rho _c(T)`$ and the temperature linear to the nonlinear range in the in-plane resistivity $`\rho _{ab}(T)`$. Because copper oxide materials are very complex systems, it is also possible that the actual c-axis conductivity may be a linear combination of the momentum-conserving $`\sigma _c(\omega )`$ and momentum-nonconserving $`\sigma _c^{(n)}(\omega )`$, but we believe that $`\sigma _c(\omega )`$ should be the major part of the c-axis conductivity in the chain copper oxide materials, while $`\sigma _c^{(n)}`$ should be the major part of the c-axis conductivity in the no-chain copper oxide materials.
###### Acknowledgements.
The authors would like to thank Professor Ru-Shan Han, Professor H. Q. Lin, and Professor T. Xiang for helpful discussions. This work was supported by the National Natural Science Foundation under Grant No. 19774014 and the State Education Department of China through the Foundation of Doctoral Training. |
warning/0001/hep-lat0001029.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In a recent paper Lüscher has investigated generic structures of chiral anomaly for abelian gauge theory on the lattice. His work is extended to arbitrary higher dimensions in our previous papers , where the topological part of the axial anomaly is shown to be interpretable as a lattice generalization of Chern character within the framework of noncommutative differential calculus . In the continuum theory the Chern character gives the integer topological winding number when integrated over the base manifold and it coincides with the index of the Dirac operator . The chiral anomaly on the lattice can also be related with the index of the Ginsparg-Wilson Dirac operator . So it is very natural to expect some extension from the continuum to the lattice of the index theorem relating the analytical index of the Dirac operator with the topological invariant of the manifold on which the Dirac operator is defined . In this respect, it is, however, not clear in the constructions of ref. whether the lattice analogue of Chern character can be related to some topology of the gauge theory on the lattice.
One might think that it would make no sense to argue the topological configurations on the lattice since any lattice fields could be continuously deformed into the trivial configuration and no nontrivial topological invariants could be constructed. But this is not the case. As argued in refs. , it is indeed possible to define a smooth fiber bundle and hence a topological winding number for a given lattice field configuration if it contains no exceptional link variables . In the case of abelian theories Lüscher has shown in ref. that the configuration space of the link variables satisfying the admissibility condition has a rich topological structure. It is considered as a kind of smoothness condition for the gauge field configuration, ensuring the existence of gauge potentials continuously parameterizing the link variables. The essential point here is that the configuration space of the admissible link variables is topologically disconnected.
In this paper we investigate the topological charge of abelian gauge theory on a periodic lattice in arbitrary even dimensions and argue that the lattice analogue of Chern character obtained in refs. indeed gives an integer-valued topological invariant by relating it to the topological winding number of a $`U(1)`$ bundle constructed from the lattice gauge fields by the interpolation method of refs. .
We should add a brief argument concerning the theorem given in refs. , where it is supposed for an infinite hypercubic regular lattice. We can extend it to topologically nontrivial lattices $`\mathrm{\Lambda }`$ without boundaries by restricting to ultralocal functions. We assume that $`\mathrm{\Lambda }`$ is locally hypercubic and regular. This means that for any point $`n\mathrm{\Lambda }`$ one can find a set $`U_n`$ of lattice points and links with a hypercubic regular lattice structure of the same dimensions. Hypercubic regular lattices with periodic boundary conditions, which we shall consider, are examples for $`\mathrm{\Lambda }`$. We call functions $`f`$ on $`\mathrm{\Lambda }`$ ultralocal if $`f(n)`$ for any $`n\mathrm{\Lambda }`$ depends only on the gauge potentials associated to links within the subset $`U_n`$ of $`\mathrm{\Lambda }`$. The abelian gauge potentials on the lattice will be treated in Sect. 3 in detail. See also refs. . Throughout this paper we assume the lattice spacing $`a=1`$. The forward and backward difference operators $`\Delta _\mu `$ and $`\Delta _\mu ^{}`$ are then defined by
$`\Delta _\mu f(n)=f(n+\widehat{\mu })f(n),\Delta _\mu ^{}f(n)=f(n)f(n\widehat{\mu }).`$ (1.1)
Then the theorem is extended to the lattice $`\mathrm{\Lambda }`$ as
Theorem : Let $`q`$ be gauge invariant and smooth ultralocal function of the abelian gauge potentials $`A_\mu `$ on a locally hypercubic regular lattice $`\mathrm{\Lambda }`$ of dimensions $`D`$ without boundaries that satisfies the topological invariance
$`{\displaystyle \underset{n\mathrm{\Lambda }}{}}\delta q(n)`$ $`=`$ $`0`$ (1.2)
for arbitrary local variations of the gauge potentials $`A_\mu A_\mu +\delta A_\mu `$, then $`q(n)`$ for arbitrary $`n\mathrm{\Lambda }`$ takes the form
$`q(n)`$ $`=`$ $`{\displaystyle \underset{l=0}{\overset{[D/2]}{}}}\beta _{\mu _1\nu _1\mathrm{}\mu _l\nu _l}F_{\mu _1\nu _1}(n)F_{\mu _2\nu _2}(n+\mu _1+\nu _1)`$ (1.3)
$`\times \mathrm{}\times F_{\mu _l\nu _l}(n+\mu _1+\nu _1+\mathrm{}+\mu _{l1}+\nu _{l1})+\Delta _\mu ^{}k_\mu (n),`$
where $`F_{\mu \nu }(n)=\Delta _\mu A_\nu (n)\Delta _\nu A_\mu (n)`$ is the field strength, the coefficient $`\beta _{\mu _1\nu _1\mathrm{}\mu _n\nu _n}`$ is antisymmetric in its indices and the current $`k_\mu `$ can be chosen to be gauge invariant and ultralocal in the gauge potential. For functions $`q`$ on the infinite lattice $`𝐙^D`$ the theorem holds true. Since $`\mathrm{\Lambda }`$ is assumed to be locally hypercubic and regular, the same identity should also follow for ultralocal functions $`q`$. The point here is that the gauge invariant current $`k_\mu `$ can be chosen to be ultralocal since both $`q`$ and the topological terms are ultralocal and (1.3) must be an identity for any configuration of the gauge potentials. As we will see, the topological charge density in general has a complicated form. The theorem will be used to rewrite it to the standard form (1.3).
This paper is organized as follows. In the next section we describe the topological winding number of abelian gauge theory on a $`D`$ dimensional torus, giving an explicit formula for the topological charge in terms of the transition functions. Sect. 3 deals with abelian gauge fields on a $`D`$ dimensional periodic lattice. We argue the precise relation among the link variables, the field strengths and the gauge potentials on the lattice. In Sect. 4 we give a formulation of interpolation of the parallel transport functions from the discrete lattice to the continuum. The interpolated transition functions are obtained in closed form. In Sect. 5 we define the topological charge of the lattice abelian gauge theory by the Chern number of the fiber bundle with the interpolated transition functions given in Sect. 4. The connection between the topological charge and the chiral anomaly obtained by the method of noncommutative differential calculus in our previous works is established. We also show that the topological charge can be expressed solely by the magnetic fluxes introduced in ref. . A mathematical lemma concerning the existence of periodic potential functions for flux-free field configurations is given in Sect. 6. In the proof of the lemma we show the systematic way to isolate the flux contributions from the field strengths. For illustration we present a simple but nontrivial field configuration of constant field strengths with unit topological charges in two dimensions. Sect. 7 is devoted to summary and discussion.
## 2 Topological charge of abelian gauge theory on $`T^D`$
Fiber bundles over a manifold are topologically classified by the equivalence class of transition functions. Our first main concern is to give a formula for the topological winding number of the fiber bundle in terms of transition functions. In this paper we consider $`U(1)`$ bundles over a torus $`T^D`$ of dimensions $`D=2N`$ defined by the identification
$`xx+L\widehat{\mu }\text{for}x𝐑^D,\mu =1,\mathrm{},D`$ (2.1)
where $`\widehat{\mu }`$ denotes the unit vector in the $`\mu `$-th direction and the period $`L`$ of the torus is assumed to be a positive integer. A hypercubic periodic lattice $`\mathrm{\Lambda }`$ of dimensions $`D`$ is defined as the set of integral lattice points in $`T^D`$.
We divide $`T^D`$ into $`L^D`$ hypercubes $`c(n)`$ ($`n\mathrm{\Lambda }`$) defined by
$`c(n)=\{xT^D|x=n+{\displaystyle \underset{\mu =1}{\overset{D}{}}}y_\mu \widehat{\mu },0y_\mu 1\}.`$ (2.2)
We assume that $`L`$ is large enough so that any restricted bundle over $`c(n)`$ is trivial. Mathematically this can be achieved for $`L2`$. For later convenience, let us denote the intersection of $`c(n)`$ and $`c(n\widehat{\mu })`$ by $`p(n,\mu )`$ and the common boundary of $`p(n,\mu )`$, $`p(n,\nu )`$, $`\mathrm{}`$, $`p(n,\sigma )`$ by $`p(n,\mu ,\nu ,\mathrm{},\sigma )`$.
Let $`A^{(n)}(x)`$ be the gauge potential 1-formThe gauge field $`A_\mu ^{(n)}(x)`$ is assumed to be real. on the hypercube $`c(n)`$, then the gauge potentials $`A^{(n\widehat{\mu })}`$ and $`A^{(n)}`$ are related by a gauge transformation on $`p(n,\mu )`$
$`A^{(n\widehat{\mu })}(x)`$ $`=`$ $`A^{(n)}(x)+d\mathrm{\Lambda }_{n,\mu }(x),`$ (2.3)
where $`d`$ is the ordinary exterior differential and $`\mathrm{\Lambda }_{n,\mu }`$ is defined by the transition functions $`v_{n,\mu }`$ as
$`v_{n,\mu }=\mathrm{e}^{i\mathrm{\Lambda }_{n,\mu }}.`$ (2.4)
The transition functions must satisfy cocycle conditions for $`xp(n,\mu ,\nu )`$
$`v_{n\widehat{\mu },\nu }(x)v_{n,\mu }(x)=v_{n\widehat{\nu },\mu }(x)v_{n,\nu }(x).`$ (2.5)
In terms of $`\mathrm{\Lambda }_{n,\mu }`$ they can be written as
$`\Delta _\mu ^{}\mathrm{\Lambda }_{n,\nu }=\Delta _\nu ^{}\mathrm{\Lambda }_{n,\mu }(\text{mod}2\pi ),`$ (2.6)
where $`\Delta _\mu ^{}`$ is the backward difference operator defined by (1.1).
The topological charge of the abelian gauge theory is given by
$`Q=C_N{\displaystyle _{T^D}}F^N,C_N={\displaystyle \frac{1}{(2\pi )^NN!}},`$ (2.7)
where $`F(x)=dA^{(n)}(x)`$ is the field strength 2-form for $`xc(n)`$. It is gauge invariant and globally defined on $`T^D`$. By the Bianchi identity $`dF=0`$, we have $`F^N=d(A^{(n)}F^{N1})`$ and hence
$`Q`$ $`=`$ $`C_N{\displaystyle \underset{n\mathrm{\Lambda }}{}}{\displaystyle _{c(n)}}d(A^{(n)}F^{N1})`$ (2.8)
$`=`$ $`C_N{\displaystyle \underset{n\mathrm{\Lambda }}{}}{\displaystyle \underset{\mu }{}}\left({\displaystyle _{p(n+\mu ,\mu )}}A^{(n)}F^{N1}{\displaystyle _{p(n,\mu )}}A^{(n)}F^{N1}\right)`$
$`=`$ $`C_N{\displaystyle \underset{n\mathrm{\Lambda }}{}}{\displaystyle \underset{\mu }{}}{\displaystyle _{p(n,\mu )}}(A^{(n\widehat{\mu })}A^{(n)})F^{N1}`$
$`=`$ $`C_N{\displaystyle \underset{n\mathrm{\Lambda }}{}}{\displaystyle \underset{\mu }{}}{\displaystyle _{p(n,\mu )}}𝑑\mathrm{\Lambda }_{n,\mu }F^{N1}.`$
In the last step we have used the relation (2.3). Since $`d\mathrm{\Lambda }_{n,\mu }`$ is a closed form, we can write $`d\mathrm{\Lambda }_{n,\mu }F^{N1}=d(d\mathrm{\Lambda }_{n,\mu }A^{(n)}F^{N2})`$ and $`Q`$ is further reduced as
$`Q`$ $`=`$ $`C_N{\displaystyle \underset{n}{}}{\displaystyle \underset{\mu ,\nu }{\overset{}{}}}{\displaystyle _{p(n,\mu ,\nu )}}\left\{\Delta _\nu ^{}d\mathrm{\Lambda }_{n,\mu }A^{(n)}F^{N2}+d\mathrm{\Lambda }_{n\widehat{\nu },\mu }d\mathrm{\Lambda }_{n,\nu }F^{N2}\right\},`$ (2.9)
where $`{\displaystyle \underset{\mu ,\nu ,\mathrm{}}{\overset{}{}}}`$ implies antisymmetrized summation on $`\mu ,\nu ,\mathrm{}`$ satisfying
$`{\displaystyle \underset{\mu ,\nu ,\mathrm{}}{\overset{}{}}}f_{\mu \nu \mathrm{}}=f_{12\mathrm{}}+\mathrm{},{\displaystyle \underset{\mathrm{},\mu _i,\mathrm{},\mu _j\mathrm{}}{\overset{}{}}}={\displaystyle \underset{\mathrm{},\mu _j,\mathrm{},\mu _i\mathrm{}}{\overset{}{}}}.`$ (2.10)
Since $`d`$ commutes with $`\Delta _\mu ^{}`$ and $`d\Delta _\nu ^{}\mathrm{\Lambda }_{n,\mu }`$ is symmetric in $`\mu `$ and $`\nu `$ by the cocycle condition (2.6), the first term in the integrand of (2.9) does not contribute to $`Q`$. We thus obtain
$`Q`$ $`=`$ $`C_N{\displaystyle \underset{n}{}}{\displaystyle \underset{\mu ,\nu }{\overset{}{}}}{\displaystyle _{p(n,\mu ,\nu )}}𝑑\mathrm{\Lambda }_{n\widehat{\nu },\mu }𝑑\mathrm{\Lambda }_{n,\nu }F^{N2}.`$ (2.11)
Similar procedure can be repeated until all the field strengths disappear. The final expression for $`Q`$ is given by
$`Q`$ $`=`$ $`(1)^{\frac{N(N1)}{2}}C_N{\displaystyle \underset{n}{}}{\displaystyle \underset{\mu _1,\mathrm{},\mu _N}{\overset{}{}}}{\displaystyle _{p(n,\mu _1,\mu _2,\mathrm{},\mu _N)}}𝑑\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_N,\mu _1}𝑑\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_N,\mu _2}\mathrm{}𝑑\mathrm{\Lambda }_{n,\mu _N}`$ (2.12)
$`=`$ $`(1)^{\frac{N(N1)}{2}}C_N{\displaystyle \underset{n}{}}ϵ_{\mu _1\mu _2\mathrm{}\mu _D}{\displaystyle _{p(n,\mu _1,\mu _2,\mathrm{},\mu _N)}}d^Nx_{\mu _{N+1}}\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_N,\mu _1}_{\mu _{N+2}}\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_N,\mu _2}`$
$`\times \mathrm{}\times _{\mu _D}\mathrm{\Lambda }_{n,\mu _N},`$
where $`ϵ_{\mu _1\mathrm{}\mu _D}`$ is the Levi-Civita symbol in $`D`$ dimensions, $`_\mu `$ stands for the derivative with respect to $`x_\mu `$, and $`d^Nx`$ is the volume form on $`p(n,\mu _1,\mathrm{},\mu _N)`$ .
To show that no new condition other than the cocycle conditions (2.6) is necessary we prove it by mathematical induction. Let us assume that
$`Q`$ $`=`$ $`(1)^{\frac{k(k1)}{2}}C_N{\displaystyle \underset{n}{}}{\displaystyle \underset{\mu _1,\mathrm{},\mu _k}{\overset{}{}}}{\displaystyle _{p(n,\mu _1,\mu _2,\mathrm{},\mu _k)}}𝑑\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k,\mu _1}𝑑\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k,\mu _2}`$ (2.13)
$`\times \mathrm{}\times d\mathrm{\Lambda }_{n,\mu _k}F^{Nk}`$
holds true up to some integer $`k<N`$. Then by carrying out the manipulation from (2.8) to (2.9) we have
$`Q`$ $`=`$ $`(1)^{\frac{k(k+1)}{2}}C_N{\displaystyle \underset{n}{}}{\displaystyle \underset{\mu _1,\mathrm{},\mu _{k+1}}{\overset{}{}}}{\displaystyle _{p(n,\mu _1,\mu _2,\mathrm{},\mu _k,\mu _{k+1})}}`$
$`\times (\Delta _{\mu _{k+1}}^{}(d\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k,\mu _1}d\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k,\mu _2}\mathrm{}d\mathrm{\Lambda }_{n,\mu _k})A^{(n)}`$
$`+d\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k\widehat{\mu }_{k+1},\mu _1}d\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k\widehat{\mu }_{k+1},\mu _2}\mathrm{}d\mathrm{\Lambda }_{n\widehat{\mu }_{k+1},\mu _k}d\mathrm{\Lambda }_{n,\mu _{k+1}})F^{Nk1}.`$
By noting the Leibnitz rule on the lattice $`\Delta _\mu ^{}(f_ng_n)=\Delta _\mu ^{}f_ng_n+f_{n\widehat{\mu }}\Delta _\mu ^{}g_n`$, we get
$`\Delta _{\mu _{k+1}}^{}(d\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k,\mu _1}d\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k,\mu _2}\mathrm{}d\mathrm{\Lambda }_{n,\mu _k})`$
$`=d\Delta _{\mu _{k+1}}^{}\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k,\mu _1}d\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k,\mu _2}\mathrm{}d\mathrm{\Lambda }_{n,\mu _k}`$
$`+d\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k\widehat{\mu }_{k+1},\mu _1}d\Delta _{\mu _{k+1}}^{}\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k,\mu _2}\mathrm{}d\mathrm{\Lambda }_{n,\mu _k}`$
$`+\mathrm{}+d\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_k\widehat{\mu }_{k+1},\mu _1}d\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_k\widehat{\mu }_{k+1},\mu _2}\mathrm{}d\Delta _{\mu _{k+1}}^{}\mathrm{\Lambda }_{n,\mu _k}.`$ (2.15)
The $`j`$-th term of the rhs of this expression is symmetric in $`\mu _j`$ and $`\mu _{k+1}`$ for $`j=1,\mathrm{},k`$ by the cocycle condition (2.6), hence the first term of the integrand of (2) does not contribute to the topological charge due to the antisymmetrized sum. The resulting expression for $`Q`$ is just (2.13) with $`k`$ replaced by $`k+1`$, implying that (2.13) holds true for any $`kN`$. This completes the proof of (2.12).
## 3 Lattice abelian gauge field on the periodic lattice
The link variables $`U_\mu (n)`$ ($`n\mathrm{\Lambda },\mu =1,\mathrm{},D`$) are subject to the periodic boundary conditions
$`U_\mu (n+L\widehat{\nu })=U_\mu (n),(\nu =1,\mathrm{},D),`$ (3.1)
and are assumed to be parametrized by
$`U_\mu (n)`$ $`=`$ $`\mathrm{e}^{ia_\mu (n)},(\pi a_\mu (n)<\pi ),`$ (3.2)
where $`a_\mu (n)`$ is a vector field on $`\mathrm{\Lambda }`$. The field strength is defined by
$`F_{\mu \nu }(n)={\displaystyle \frac{1}{i}}\mathrm{ln}U_\mu (n)U_\nu (n+\widehat{\mu })U_\mu (n+\widehat{\nu })^1U_\nu (n)^1,|F_{\mu \nu }(n)|<\pi .`$ (3.3)
We exclude the exceptional field configurations with $`|F_{\mu \nu }(n)|=\pi `$, where the plaquette variable equals $`1`$ and $`(1)^y`$ ($`0y1`$) becomes ambiguous.
The field strengths can be written in terms of $`a_\mu `$ as
$`F_{\mu \nu }(n)=\Delta _\mu a_\nu (n)\Delta _\nu a_\mu (n)+2\pi n_{\mu \nu }(n),`$ (3.4)
where $`\Delta _\mu `$ is the forward difference operator and $`n_{\mu \nu }`$ is an integer-valued anti-symmetric tensor field on $`\mathrm{\Lambda }`$. The $`n_{\mu \nu }(n)`$ must be chosen so that the field strength $`F_{\mu \nu }(n)`$ lies within the principal branch of the logarithm in (3.3). Obviously, it satisfies $`|n_{\mu \nu }(n)|2`$. Furthermore, the link variables are assumed to be restricted so that the field strengths always satisfy the Bianchi identity $`\Delta _{[\lambda }F_{\mu \nu ]}(n)=0`$ in order to ensure the existence of gauge potential $`A_\mu `$ as
$`F_{\mu \nu }(n)`$ $`=`$ $`\Delta _\mu A_\nu (n)\Delta _\nu A_\mu (n).`$ (3.5)
This requirement is automatically satisfied by the admissibility condition of ref. that the field strengths satisfy
$`\underset{n,\mu ,\nu }{sup}|F_{\mu \nu }(n)|`$ $`<`$ $`ϵ`$ (3.6)
for a fixed constant $`0<ϵ<\pi /3`$.<sup>§</sup><sup>§</sup>§As noted in ref. , $`ϵ`$ should be replaced with $`ϵ/a^2`$ for the lattice spacing $`a1`$ and, hence, the restriction on the gauge field configuration disappears in the classical continuum limit. The exceptional filed configurations mentioned above are also excluded by this condition.
Gauge transformations on $`\mathrm{\Lambda }`$ are defined by
$`U_\mu (n)`$ $``$ $`V(n)U_\mu (n)V(n+\widehat{\mu })^1,`$ (3.7)
where $`V`$ is a $`U(1)`$-valued function on $`\mathrm{\Lambda }`$. The field strengths (3.3) is obviously gauge invariant. Note, however, that neither $`\Delta _\mu a_\nu (n)\Delta _\nu a_\mu (n)`$ nor $`n_{\mu \nu }(n)`$ are separately gauge invariant. To see this we parameterize the gauge transformations $`V`$ by a function $`\lambda `$ on $`\mathrm{\Lambda }`$ as $`V(n)=\mathrm{e}^{i\lambda (n)}`$, where $`\lambda `$ is assumed to satisfy $`\lambda (n+L\widehat{\mu })=\lambda (n)`$ and $`|\lambda (n)|\pi `$. Then $`a_\mu (n)`$ and $`n_{\mu \nu }(n)`$ are transformed as
$`a_\mu (n)a_\mu (n)\Delta _\mu \lambda (n)+2\pi N_\mu (n),n_{\mu \nu }(n)n_{\mu \nu }(n)\Delta _\mu N_\nu (n)+\Delta _\nu N_\mu (n),`$ (3.8)
where $`N_\mu `$ is an integer-valued vector field on $`\mathrm{\Lambda }`$ and must be chosen to satisfy
$`\pi a_\mu (n)\Delta _\mu \lambda (n)+2\pi N_\mu (n)<\pi .`$ (3.9)
If the field strength satisfies the Bianchi identities, so does $`n_{\mu \nu }`$. Hence it is always possible to find an integer-valued vector field $`m_\mu `$ on $`𝐙^D`$ satisfying
$`A_\mu (n)=a_\mu (n)+2\pi m_\mu (n),\Delta _\mu m_\nu (n)\Delta _\nu m_\mu (n)=n_{\mu \nu }(n).`$ (3.10)
To show this we note that $`m_\mu (n)`$ is only determined up to integer-valued gauge transformations $`m_\mu (n)m_\mu (n)\Delta _\mu \Lambda (n)`$ ($`\Lambda (n)𝐙`$) on $`𝐙^D`$ and we can always work in the axial gauge $`m_D(n)=0`$. In this gauge $`m_\mu (n)`$ ($`\mu D`$) satisfies $`\Delta _Dm_\mu (n)=n_{\mu D}(n)`$, which can be integrated to
$`m_\mu (n)`$ $`=`$ $`{\displaystyle \underset{y_D=0}{\overset{n_D1}{}}}n_{\mu D}(n_1,\mathrm{},n_{D1},y_D)+m_\mu (n_1,\mathrm{},n_{D1},0).`$ (3.11)
The sum on $`y_D`$ in the rhs must make sense for arbitrary integer $`n_D`$. This can be achieved by defining the sum as
$`{\displaystyle \underset{y=a}{\overset{b1}{}}}f(y)=\{\begin{array}{cc}f(a)+\mathrm{}+f(b1)\hfill & (b>a)\hfill \\ 0\hfill & (b=a)\hfill \\ f(b)\mathrm{}f(a1)\hfill & (b<a)\hfill \end{array}`$ (3.12)
for arbitrary functions $`f`$ on $`𝐙`$ and $`a,b𝐙`$. It is just an analogue of one-dimensional integral on the discrete space $`𝐙`$ and satisfies the following properties
$`{\displaystyle \underset{y=a}{\overset{b1}{}}}f(y)={\displaystyle \underset{y=b}{\overset{a1}{}}}f(y),{\displaystyle \underset{y=a}{\overset{b1}{}}}f(y)={\displaystyle \underset{y=a}{\overset{c1}{}}}f(y)+{\displaystyle \underset{y=c}{\overset{b1}{}}}f(y),`$
$`\Delta _x\left({\displaystyle \underset{y=a}{\overset{x1}{}}}f(y)\right)=f(x),{\displaystyle \underset{y=a}{\overset{b1}{}}}\Delta _yf(y)=f(b)f(a),`$ (3.13)
where $`\Delta _z`$ stands for the difference operator with respect to $`z𝐙`$. In (3.11) $`m_\mu (n)|_{n_D=0}=m_\mu (n_1,\mathrm{},n_{D1},0)`$ ($`\mu D`$) are still to be determined. In order for (3.10) to be consistent they must satisfy the following set of equations
$`\Delta _\mu m_\nu (n_1,\mathrm{},n_{D1},0)\Delta _\nu m_\mu (n_1,\mathrm{},n_{D1},0)=n_{\mu \nu }(n_1,\mathrm{},n_{D1},0).`$ (3.14)
We see that the problem of finding $`m_\mu (n)`$ in $`D`$ dimensions is reduced to the problem of solving the original equations dimensionally reduced to $`D1`$ dimensions. Again we can choose the axial gauge $`m_{D1}(n_1,\mathrm{},n_{D1},0)=0`$ and solve (3.14) as before. Obviously, this reduction process can be continued until all the $`m_\mu `$’s are completely determined. We thus obtain
$`m_\mu (n)`$ $`=`$ $`{\displaystyle \underset{\nu >\mu }{}}{\displaystyle \underset{y_\nu =0}{\overset{n_\nu 1}{}}}n_{\mu \nu }(n_1,\mathrm{},n_{\nu 1},y_\nu ,0,\mathrm{},0).`$ (3.15)
It is integer-valued as announced.
From (3.10) the gauge potential $`A_\mu `$ also serves as the local coordinates for the link variables by the relations $`U_\mu (n)=\mathrm{e}^{iA_\mu (n)}`$ and the gauge transformation (3.8) simply becomes
$`A_\mu (n)`$ $``$ $`A_\mu (n)\Delta _\mu \lambda (n)2\pi \Delta _\mu \Lambda (n),`$ (3.16)
where $`\Lambda `$ is an arbitrary integer-valued function on $`𝐙^D`$. We emphasize that the gauge potential $`A_\mu (n)`$ defined by (3.10) and (3.15) is continuous at $`U_\mu (n)=1`$ though $`a_\mu (n)`$ exhibits a discontinuity.
Note that the periodicity of the link variables and the field strengths only implies that of the gauge potentials up to gauge trasnformations by $`2\pi \Lambda `$ with $`\Lambda (n)𝐙`$. In order to obtain nonvanishing topological charge it is necessary to have nonperiodic gauge potentials. For $`m_\mu (n)`$ given by (3.15) the periodicity of $`A_\mu (n)`$ is completely determined by $`n_{\mu \nu }(n)`$ as
$`A_\mu (n+L\widehat{\nu })A_\mu (n)`$ $`=`$ $`2\pi m_\mu (n+L\widehat{\nu })2\pi m_\mu (n)`$ (3.17)
$`=`$ $`\{\begin{array}{cc}2\pi {\displaystyle \underset{n_\nu =0}{\overset{L1}{}}}n_{\mu \nu }(n_1,\mathrm{},n_\nu ,0,\mathrm{},0)\hfill & \text{for }\mu <\nu \hfill \\ 0\hfill & \text{for }\mu \nu \hfill \end{array}`$
## 4 Interpolated transition functions
We now turn to the construction of the transition functions from the gauge fields on the lattice $`\mathrm{\Lambda }`$ by using the interpolation technique given in refs. . The main concerns of the authors of refs. were the construction of the topological charges for nonabelian theories in four dimensions. The interpolated transition functions become more and more complicated as the dimensions increase in the nonabelian case so that the general expression cannot be available in arbitrary dimensions. In the abelian case, however, all the complications related to the noncommutativity of the transition functionsIn the nonabelian theories the transition functions are in fact matrix-valued in some representation. disappear and it is possible to give transition functions in closed form as we will show.
We first define parallel transport functions $`w^n(\overline{x})`$ by
$`w^n(\overline{x})=U_1(n)^{\sigma _1}U_2(n+\sigma _1\widehat{1})^{\sigma _2}\mathrm{}U_D(n+\sigma _1\widehat{1}+\sigma _2\widehat{2}+\mathrm{}+\sigma _{D1}\widehat{D1})^{\sigma _D}`$ (4.1)
for points $`\overline{x}`$ of the corners of $`c(n)`$ given by
$`\overline{x}=n+{\displaystyle \underset{\mu =1}{\overset{D}{}}}\sigma _\mu \widehat{\mu },(\sigma _\mu =\{0,1\}).`$ (4.2)
At this point they are only defined on the corners of $`c(n)`$. In what follows we define iteratively interpolations of $`w^n(\overline{x})`$ over the boundary $`c(n)`$.
Before going into mathematical detail, we mention the subtleties in defining arbitrary exponents $`(w^n(\overline{x}))^y`$ ($`0y1`$). We define exponents for link variables by $`(U_\mu (n))^y\mathrm{e}^{iya_\mu (n)}`$. But $`(w^n(\overline{x}))^y`$ does not coincide with $`(U_1(n)^{\sigma _1})^y\mathrm{}(U_D(n+\sigma _1\widehat{1}+\mathrm{}+\sigma _{D1}\widehat{D1})^{\sigma _D})^y`$ in general. The subtlety concerning the violation of such a naive distribution law of exponents can be avoided if we assume that $`|a_\mu (n)|`$ is sufficiently small for any link variable so that the naive distribution law is applicable. We assume this for the time being to justify the mathematical manipulations and remove such restrictions from the final expression of the transition functions.
Let $`\mu _1,\mathrm{},\mu _{D1}=\{1,\mathrm{},D\}\backslash \{\mu \}`$ be the $`D1`$ indices satisfying $`\mu _1<\mathrm{}<\mu _{D1}`$. Then the interpolation along the $`\widehat{\mu }_{D1}`$ direction is defined by
$`w^m\left(n+{\displaystyle \underset{k=1}{\overset{D2}{}}}\sigma _k\widehat{\mu }_k+y_{D1}\widehat{\mu }_{D1}\right)`$ $``$ $`w^m\left(n+{\displaystyle \underset{k=1}{\overset{D2}{}}}\sigma _k\widehat{\mu }_k+\widehat{\mu }_{D1}\right)^{y_{D1}}`$ (4.3)
$`\times w^m\left(n+{\displaystyle \underset{k=1}{\overset{D2}{}}}\sigma _k\widehat{\mu }_k\right)^{1y_{D1}},`$
where $`m=n`$ or $`m=n\widehat{\mu }`$ and $`y_{D1}`$ ($`0y_{D1}1`$) is the interpolation parameter and can be regarded as the coordinate of $`c(n)`$ in the $`\widehat{\mu }_{D1}`$ direction. We use (4.3) to define the second interpolation in the $`\widehat{\mu }_{D2}`$ direction as
$`w^m\left(n+{\displaystyle \underset{k=1}{\overset{D3}{}}}\sigma _k\widehat{\mu }_k+y_{D2}\widehat{\mu }_{D2}+y_{D1}\widehat{\mu }_{D1}\right)`$
$`w^m\left(n+{\displaystyle \underset{k=1}{\overset{D3}{}}}\sigma _k\widehat{\mu }_k+\widehat{\mu }_{D2}+y_{D1}\widehat{\mu }_{D1}\right)^{y_{D2}}w^m\left(n+{\displaystyle \underset{k=1}{\overset{D2}{}}}\sigma _k\widehat{\mu }_k+y_{D1}\widehat{\mu }_{D1}\right)^{1y_{D2}},`$
where $`y_{D2}`$ ($`0y_{D2}1`$) is the new parameter for the interpolation.
We can carry out such interpolation procedure step by step until all the points in $`p(n,\mu )`$ are covered. After $`l`$ steps, we obtain
$`w^m\left(n+{\displaystyle \underset{k=1}{\overset{Dl1}{}}}\sigma _k\widehat{\mu }_k+{\displaystyle \underset{k=Dl}{\overset{D1}{}}}y_k\widehat{\mu }_k\right)`$
$`w^m\left(n+{\displaystyle \underset{k=1}{\overset{Dl1}{}}}\sigma _k\widehat{\mu }_k+\widehat{\mu }_{Dl}+{\displaystyle \underset{k=Dl+1}{\overset{D1}{}}}y_k\widehat{\mu }_k\right)^{y_{Dl}}`$
$`\times w^m\left(n+{\displaystyle \underset{k=1}{\overset{Dl1}{}}}\sigma _k\widehat{\mu }_k+{\displaystyle \underset{k=Dl+1}{\overset{D1}{}}}y_k\widehat{\mu }_k\right)^{1y_{Dl}}.`$ (4.5)
In this way we can construct the parallel transport matrix $`w^m(x)`$ as
$`w^m(x)={\displaystyle \underset{\{\sigma _k=0,1\}_{k=1,\mathrm{},D1}}{}}w^m\left(n+{\displaystyle \underset{k=1}{\overset{D1}{}}}\sigma _k\widehat{\mu }_k\right)^{_{k=1}^{D1}(\sigma _ky_k+(1\sigma _k)(1y_k))}`$ (4.6)
for any point $`xp(n,\mu )`$ given by
$`x=n+{\displaystyle \underset{k=1}{\overset{D1}{}}}y_k\widehat{\mu }_k,(0y_k1).`$ (4.7)
In deriving (4.6) we have made a heavy use of the naive distribution law of exponents mentioned above. It is very important to note that the parallel transport function $`\{w^n(x)\}_{xc(n)}`$ defines a continuous mapping $`c(n)U(1)`$ as can be verified directly from (4.6).
In two and four dimensions $`w^m(x)`$ are explicitly given by
$`D=2:w^m(x)`$ $`=`$ $`w^m(n)^{1y_1}w^m(n+\widehat{\mu }_1)^{y_1}`$ (4.8)
$`D=4:w^m(x)`$ $`=`$ $`w^m(n)^{(1y_1)(1y_2)(1y_3)}w^m(n+\mu _1)^{y_1(1y_2)(1y_3)}w^m(n+\widehat{\mu }_2)^{(1y_1)y_2(1y_3)}`$ (4.9)
$`\times w^m(n+\widehat{\mu }_3)^{(1y_1)(1y_2)y_3}w^m(n+\widehat{\mu }_2+\widehat{\mu }_3)^{(1y_1)y_2y_3}`$
$`\times w^m(n+\widehat{\mu }_1+\widehat{\mu }_3)^{y_1(1y_2)y_3}w^m(n+\widehat{\mu }_1+\widehat{\mu }_2)^{y_1y_2(1y_3)}`$
$`\times w^m(n+\widehat{\mu }_1+\widehat{\mu }_2+\widehat{\mu }_3)^{y_1y_2y_3}`$
Following ref. , we define the transition functions $`v_{n,\mu }(x)`$ interpolated over $`p(n,\mu )`$ by
$`v_{n,\mu }(x)w^{n\widehat{\mu }}(x)w^n(x)^1.`$ (4.10)
With this definition one can easily show that the transition functions indeed satisfy the cocycle conditions (2.5). It is not so difficult to compute the transition functions explicitly in arbitrary dimensions. We first note that the interpolated transition functions $`v_{n,\mu }(x)`$ ($`xp(n,\mu )`$) are given by
$`v_{n,\mu }(x)={\displaystyle \underset{\{\sigma _k=0,1\}_{k=1,\mathrm{},D1}}{}}v_{n,\mu }\left(n+{\displaystyle \underset{k=1}{\overset{D1}{}}}\sigma _k\widehat{\mu }_k\right)^{_{k=1}^{D1}(\sigma _ky_k+(1\sigma _k)(1y_k))}.`$ (4.11)
Again use has been made of the naive manipulation on the exponents. For the corner points of $`p(n,\mu )`$ it is straightforward to show that the transition functions are given by
$`v_{n,\mu }\left(n+{\displaystyle \underset{k=1}{\overset{D1}{}}}\sigma _k\widehat{\mu }_k\right)=v_{n,\mu }(n)\mathrm{exp}\left[i{\displaystyle \underset{\genfrac{}{}{0pt}{}{k=1}{\mu _k<\mu }}{\overset{D1}{}}}\sigma _kF_{\mu _k\mu }(n\widehat{\mu }+\sigma _1\widehat{\mu }_1+\mathrm{}+\sigma _{k1}\widehat{\mu }_{k1})\right],`$ (4.12)
where the field strengths are given by (3.3) and $`v_{n,\mu }(n)=w^{n\widehat{\mu }}(n)w^n(n)^1=U_\mu (n\widehat{\mu })`$ is consistent with (4.10). Putting this into the rhs of (4.11), we obtain $`v_{n,\mu }(x)`$ explicitly as
$`v_{n,\mu }(x)`$ $`=`$ $`v_{n,\mu }(n)\mathrm{exp}[i{\displaystyle \underset{\genfrac{}{}{0pt}{}{k=1}{\mu _k<\mu }}{\overset{D1}{}}}{\displaystyle \underset{\{\sigma _j=0,1\}_{j=1,\mathrm{},k1}}{}}F_{\mu _k\mu }(n\widehat{\mu }+{\displaystyle \underset{l=1}{\overset{k1}{}}}\sigma _l\widehat{\mu }_l)`$ (4.13)
$`\times {\displaystyle \underset{l=1}{\overset{k1}{}}}(\sigma _ly_l+(1\sigma _l)(1y_l))y_k]`$
So far we have assumed that $`|a_\mu (n)|`$ is sufficiently small for any link variable so that all the mathematical manipulations concerning the distribution law of exponents can be justified. The final expression (4.13), however, satisfies all the desired properties expected for the transition functions such as the gauge covariance and the cocycle conditions even for general configurations as far as the exceptional configurations containing $`F_{\mu \nu }(n)=\pm \pi `$ for some $`\mu `$, $`\nu `$ and $`n`$ is excluded and the field strengths satisfy the Bianchi identities. Henceforth, we consider general field configurations not necessarily restricted to be small and regard (4.13) as the definition of the transition functions.
In the classical continuum limit we may retain only the leading terms as
$`F_{\mu _k\mu }\left(n\widehat{\mu }+{\displaystyle \underset{l=1}{\overset{k1}{}}}\sigma _l\widehat{\mu }_l\right)`$ $`=`$ $`F_{\mu _k\mu }(n)+\mathrm{}`$ (4.14)
The transition function (4.13) reduces to a simple form in the classical continuum limit as
$`v_{n,\mu }(x)`$ $`=`$ $`v_{n,\mu }(n)\mathrm{exp}\left[i{\displaystyle \underset{\genfrac{}{}{0pt}{}{k=1}{\mu _k<\mu }}{\overset{D1}{}}}F_{\mu _k\mu }(n)y_k+\mathrm{}\right].`$ (4.15)
Concrete expressions of the transition functions in some lower dimensions can be easily found form (4.13);
$`D=2:`$
$`v_{n,1}(x)`$ $`=`$ $`v_{n,1}(n),(xp(n,1)),`$
$`v_{n,2}(x)`$ $`=`$ $`v_{n,2}(n)\mathrm{e}^{iy_1F_{12}(n\widehat{2})},(xp(n,2)),`$ (4.16)
$`D=4:`$
$`v_{n,1}(x)`$ $`=`$ $`v_{n,1}(n),(xp(n,1)),`$
$`v_{n,2}(x)`$ $`=`$ $`v_{n,2}(n)\mathrm{e}^{iy_1F_{12}(n\widehat{2})},(xp(n,2)),`$
$`v_{n,3}(x)`$ $`=`$ $`v_{n,3}(n)\mathrm{e}^{i(y_1F_{13}(n\widehat{3})+(1y_1)y_2F_{23}(n\widehat{3})+y_1y_2F_{23}(n+\widehat{1}\widehat{3}))},(xp(n,3)),`$
$`v_{n,4}(x)`$ $`=`$ $`v_{n,4}(n)\mathrm{exp}i[y_1F_{14}(n\widehat{4})+(1y_1)y_2F_{24}(n\widehat{4})+(1y_1)(1y_2)y_3F_{34}(n\widehat{4})`$ (4.17)
$`+y_1y_2F_{24}(n+\widehat{1}\widehat{4})+y_1(1y_2)y_3F_{34}(n+\widehat{1}\widehat{4})`$
$`+(1y_1)y_2y_3F_{24}(n+\widehat{2}\widehat{4})+y_1y_2y_3F_{34}(n+\widehat{1}+\widehat{2}\widehat{4})],`$
$`(xp(n,4)).`$
It is instructive to see in four dimensions that the Bianchi identities is necessary for the cocycle conditions (2.5) to be fulfilled. Let us consider the case $`\mu =2`$ and $`\nu =3`$, then we have for $`xp(n,2,3)`$
$`v_{n,2}(x)=v_{n,2}(n)\mathrm{e}^{iy_1F_{12}(n\widehat{2})},v_{n,3}(x)=v_{n,3}(n)\mathrm{e}^{iy_1F_{13}(n\widehat{3})},`$
$`v_{n\widehat{3},2}(x)=v_{n,2}(x)|_{y_3=1,nn\widehat{3}}=v_{n\widehat{3},2}(n\widehat{3})\mathrm{e}^{iy_1F_{12}(n\widehat{2}\widehat{3})},`$
$`v_{n\widehat{2},3}(x)=v_{n,3}(x)|_{y_2=1,nn\widehat{2}}=v_{n\widehat{2},3}(n\widehat{2})\mathrm{e}^{i(y_1F_{13}(n\widehat{2}\widehat{3})+(1y_1)F_{23}(n\widehat{2}\widehat{3})+y_1F_{23}(n+\widehat{1}\widehat{2}\widehat{3}))}.`$
The cocycle condition in the present case follows from the relations
$`v_{n\widehat{2},3}(n\widehat{2})v_{n,2}(n)\mathrm{e}^{iF_{23}(n\widehat{2}\widehat{3})}=v_{n\widehat{3},2}(n\widehat{3})v_{n,3}(n),`$
$`\Delta _{[1}F_{23]}(n\widehat{2}\widehat{3})=0.`$ (4.19)
## 5 Topological charge of lattice abelian gauge theory on the periodic lattice
We have constructed a set of transition functions (4.13) from the link variables on the lattice $`\mathrm{\Lambda }`$. The transition functions in turn define a fiber bundle over $`T^D`$, for which the topological charge $`Q`$ can be computed unambiguously by (2.12). In order to compute $`Q`$, let us define 1-form $`d\mathrm{\Lambda }_{n,\mu }(x)`$ on $`p(n,\mu )`$ by
$`d\mathrm{\Lambda }_{n,\mu }(x)=iv_{n,\mu }(x)dv_{n,\mu }(x)^1,`$ (5.1)
where $`d`$ is the ordinary exterior derivative with respect to the continuous coordinates $`x_\mu `$. Then $`Q`$ is given by
$`Q`$ $`=`$ $`{\displaystyle \underset{n\mathrm{\Lambda }}{}}q(n),`$ (5.2)
where $`q`$ is the topological charge density satisfying $`q(n+L\widehat{\mu })=q(n)`$ for $`\mu =1,\mathrm{},D`$ and can be chosen to be
$`q(n)`$ $`=`$ $`(1)^{\frac{N(N1)}{2}}C_Nϵ_{\mu _1\mu _2\mathrm{}\mu _D}{\displaystyle _{p(n,\mu _1,\mu _2,\mathrm{},\mu _N)}}d^Nx_{\mu _{N+1}}\mathrm{\Lambda }_{n\widehat{\mu }_2\mathrm{}\widehat{\mu }_N,\mu _1}_{\mu _{N+2}}\mathrm{\Lambda }_{n\widehat{\mu }_3\mathrm{}\widehat{\mu }_N,\mu _2}`$ (5.3)
$`\times \mathrm{}\times _{\mu _D}\mathrm{\Lambda }_{n,\mu _N}.`$
From (4.13) and (5.1) one can see that $`q(n)`$ is invariant under the gauge transformations (3.16), ultralocal in the gauge potential and a sum of products of $`N`$ field strengths. Furthermore, the classical continuum limit can be obtained from (4.15) as
$`q(n)={\displaystyle \frac{1}{2^N}}C_Nϵ_{\mu _1\nu _1\mathrm{}\mu _N\nu _N}F_{\mu _1\nu _1}(n)\mathrm{}F_{\mu _N\nu _N}(n)+\mathrm{}.`$ (5.4)
These properties together with the topological invariance of $`Q`$, i.e., the invariance under arbitrary local variations of the gauge potential as in (1.2), which is obvious by construction, in fact determine the structure of $`q(n)`$ by the theorem (1.3). In the present case we have $`\beta _{\mu _1\nu _1\mathrm{}\mu _l\nu _l}=0`$ for $`l<N`$ and $`\beta _{\mu _1\nu _1\mathrm{}\mu _N\nu _N}=C_N/2^Nϵ_{\mu _1\nu _1\mathrm{}\mu _N\nu _N}`$ as can be seen from (5.4). Furthermore, the current divergence term in (1.3) does not contribute to the topological charge $`Q`$ due to the periodic boundary conditions. We thus obtain the topological charge
$`Q`$ $`=`$ $`{\displaystyle \frac{1}{2^N}}C_N{\displaystyle \underset{n\mathrm{\Lambda }}{}}ϵ_{\mu _1\nu _1\mathrm{}\mu _N\nu _N}F_{\mu _1\nu _1}(n)F_{\mu _2\nu _2}(n+\widehat{\mu }_1+\widehat{\nu }_1)`$ (5.5)
$`\times \mathrm{}\times F_{\mu _N\nu _N}(n+\widehat{\mu }_1+\widehat{\nu }_1+\mathrm{}+\widehat{\mu }_{N1}+\widehat{\nu }_{N1}).`$
The summand of this expression is nothing but the Chern character obtained solely from the noncommutative differential calculus on the lattice. This establishes the connection between the lattice analogue of the Chern character and the topology of the fiber bundle over the discrete lattice as in continuum theory.
It is possible to express the topological charge in terms of the magnetic fluxes $`\varphi _{\mu \nu }(x)`$ through the $`\mu \nu `$-plane . They are defined by
$`\varphi _{\mu \nu }(n)`$ $`=`$ $`{\displaystyle \underset{s,t=0}{\overset{L1}{}}}F_{\mu \nu }(n+s\widehat{\mu }+t\widehat{\nu })`$ (5.6)
$`=`$ $`2\pi {\displaystyle \underset{s,t=0}{\overset{L1}{}}}n_{\mu \nu }(n+s\widehat{\mu }+t\widehat{\nu }),`$
where use has been made of (3.4). We immediately see that they are gauge invariant and integer multiples of $`2\pi `$. Furthermore, they can be shown to be constants by virtue of the Bianchi identities . Since the field strengths are continuous with respect to continuous changes of link variables as far as the exceptional gauge field configurations are excluded, the fluxes $`\varphi _{\mu \nu }`$ must be invariant under such continuous changes of link variables. This implies that the sets of link variables $`\{U_\mu (n)\}_{n\mathrm{\Lambda },\mu =1,\mathrm{},D}`$ with distinct fluxes $`\{\varphi _{\mu \nu }\}_{\mu ,\nu =1,\mathrm{},D}`$ are topologically disjoint one another.
For later convenience, we introduce a set of integers $`m_{\mu \nu }=m_{\nu \mu }`$ by
$`m_{\mu \nu }={\displaystyle \underset{s,t=0}{\overset{L1}{}}}n_{\mu \nu }(n+s\widehat{\mu }+t\widehat{\nu }).`$ (5.7)
If we define $`\rho _{\mu \nu }(n)`$ by
$`\rho _{\mu \nu }(n)`$ $`=`$ $`2\pi \left(n_{\mu \nu }(n){\displaystyle \frac{1}{L^2}}m_{\mu \nu }\right),`$ (5.8)
then we have
$`{\displaystyle \underset{s,t=0}{\overset{L1}{}}}\rho _{\mu \nu }(n+s\widehat{\mu }+t\widehat{\nu })=0.`$ (5.9)
Furthermore, $`\rho _{\mu \nu }(n)`$ satisfies the Bianchi identities. Now using the lemma given in the next section, it is always possible to find a periodic vector field $`\lambda _\mu `$ on $`\mathrm{\Lambda }`$ satisfying
$`\Delta _\mu \lambda _\nu (n)\Delta _\nu \lambda _\mu (n)`$ $`=`$ $`\rho _{\mu \nu }(n).`$ (5.10)
This implies that the field strength $`F_{\mu \nu }(n)`$ can always be written as
$`F_{\mu \nu }(n)`$ $`=`$ $`\Delta _\mu \stackrel{~}{a}_\nu (n)\Delta _\nu \stackrel{~}{a}_\mu (n)+{\displaystyle \frac{2\pi }{L^2}}m_{\mu \nu },`$ (5.11)
where we have introduced a vector field on $`\mathrm{\Lambda }`$ by $`\stackrel{~}{a}_\mu (n)a_\mu (n)+\lambda _\mu (n)`$. It is possible to find a gauge potential $`\stackrel{~}{A}_\mu (n)`$ in terms of $`\stackrel{~}{a}_\mu (n)`$ and $`m_{\mu \nu }`$ as
$`\stackrel{~}{A}_\mu (n)`$ $`=`$ $`\stackrel{~}{a}_\mu (n){\displaystyle \frac{2\pi }{L^2}}{\displaystyle \underset{\nu >\mu }{}}m_{\mu \nu }n_\nu .`$ (5.12)
This satisfies the periodicity differing from (3.17) as
$`\stackrel{~}{A}_\mu (n+L\widehat{\nu })\stackrel{~}{A}_\mu (n)`$ $`=`$ $`\{\begin{array}{cc}2\pi m_{\mu \nu }/L\hfill & (\mu <\nu )\hfill \\ 0\hfill & (\mu \nu )\hfill \end{array}.`$ (5.13)
Putting (5.11) into (5.5) and noting that the periodic vector field $`\stackrel{~}{a}_\mu (n)`$ does not contribute to the topological charge, we obtain
$`Q`$ $`=`$ $`{\displaystyle \frac{1}{2^NN!}}ϵ_{\mu _1\nu _1\mathrm{}\mu _N\nu _N}m_{\mu _1\nu _1}\mathrm{}m_{\mu _N\nu _N},`$ (5.14)
where use has been made of the explicit form of $`C_N`$ given in (2.7). The rhs of this expression is manifestly an integer and topological invariant as it should be.
## 6 Lemma
In this section we describe the Lemma used in the previous section. It turns out to be useful in finding link variables for given field strengths. We now state the lemma in the following form. Lemma: Let $`\rho _{\mu \nu }`$ be an antisymmetric tensor field on the $`D`$ dimensional periodic lattice $`\mathrm{\Lambda }`$ satisfying
$`\Delta _{[\lambda }\rho _{\mu \nu ]}(n)=0,\rho _{\mu \nu }(n+L\widehat{\lambda })=\rho _{\mu \nu }(n),`$
$`{\displaystyle \underset{s,t=0}{\overset{L1}{}}}\rho _{\mu \nu }(n+s\widehat{\mu }+t\widehat{\nu })=0,(\lambda ,\mu ,\nu =1,\mathrm{},D)`$ (6.1)
then there exists a vector fields $`\lambda _\mu `$ satisfying
$`\Delta _\mu \lambda _\nu (n)\Delta _\nu \lambda _\mu (n)=\rho _{\mu \nu }(n),\lambda _\mu (n+L\nu )=\lambda _\mu (n),(\mu ,\nu =1,\mathrm{},D).`$ (6.2)
Proof: The proof will proceed as in the case of solving (3.10) for $`m_\mu (n)`$. In the present case $`\lambda _\mu (n)`$ must be periodic in the lattice coordinates. This makes the solution somewhat complicated.
We first assume the existence of $`\lambda _\mu `$ satisfying (6.2) and note that the Bianchi identity and the periodicity in (6) imply the relation
$`\Delta _\mu \left({\displaystyle \underset{s=0}{\overset{L1}{}}}\rho _{\lambda \nu }(n+s\widehat{\lambda })\right)=\Delta _\nu \left({\displaystyle \underset{s=0}{\overset{L1}{}}}\rho _{\lambda \mu }(n+s\widehat{\lambda })\right).`$ (6.3)
Now let us denote a sum of vector field $`V_\mu `$ along a lattice path $`C`$ from $`a\mathrm{\Lambda }`$ to $`b\mathrm{\Lambda }`$ through the lattice points $`a+\widehat{\mu }`$, $`a+\widehat{\mu }+\widehat{\nu }`$, $`\mathrm{}`$, $`b\widehat{\rho }`$ in this order by
$`{\displaystyle \underset{C}{}}V_\mu (y)\Delta y_\mu `$ $`=`$ $`V_\mu (a)+V_\nu (a+\widehat{\mu })+\mathrm{}+V_\rho (b\widehat{\rho }),`$ (6.4)
then (6.3) guarantees that the vector field $`\alpha _\lambda `$ defined by the sum along any lattice path $`C`$ from $`y=0`$ to $`y=n`$
$`\alpha _\lambda (n)=\alpha _\lambda (0)+{\displaystyle \underset{C}{}}{\displaystyle \underset{s=0}{\overset{L1}{}}}\rho _{\lambda \mu }(y+s\widehat{\lambda })\Delta y_\mu `$ (6.5)
is independent of the path chosen. It is also periodic as can be seen from the third property of (6). From (6.2) and (6.5) we get
$`\Delta _\mu \left({\displaystyle \underset{s=0}{\overset{L1}{}}}\lambda _\nu (n+s\widehat{\nu })\right)={\displaystyle \underset{s=0}{\overset{L1}{}}}\rho _{\mu \nu }(n+s\widehat{\nu })=\Delta _\mu \alpha _\nu (n).`$ (6.6)
Hence we may choose $`\alpha _\mu (n)`$ without loss of generality to satisfy
$`\alpha _\mu (n)={\displaystyle \underset{s=0}{\overset{L1}{}}}\lambda _\mu (n+s\widehat{\mu })`$ (6.7)
We next define $`\psi _\mu (n)`$ and $`\omega _{\mu \nu }(n)`$by
$`\psi _\mu (n)=\lambda _\mu (n)+{\displaystyle \frac{1}{L}}\alpha _\mu (n),\omega _{\mu \nu }(n)=\rho _{\mu \nu }(n)+{\displaystyle \frac{1}{L}}(\Delta _\mu \alpha _\nu (n)\Delta _\nu \alpha _\mu (n)).`$ (6.8)
From (6.2) and (6.7) one can easily show that these fulfill the following relations
$`\Delta _\mu \psi _\nu (n)\Delta _\nu \psi _\mu (n)=\omega _{\mu \nu }(n),\Delta _{[\lambda }\omega _{\mu \nu ]}(n)=0,`$
$`{\displaystyle \underset{s=0}{\overset{L1}{}}}\psi _\mu (n+s\widehat{\mu })={\displaystyle \underset{s=0}{\overset{L1}{}}}\omega _{\mu \nu }(n+s\widehat{\mu })=0.`$ (6.9)
It is now possible to solve these with respect to $`\psi _\mu (n)`$ just like (3.10) by working with the axial gauge $`\psi _D(n)=0`$. In this gauge $`\psi _\mu (n)`$ ($`\mu =1,\mathrm{},D1`$) are given by
$`\psi _\mu (n)={\displaystyle \underset{y_D=0}{\overset{n_D1}{}}}\omega _{\mu D}(n_1,\mathrm{},n_{D1},y_D)+\psi _\mu (n_1,\mathrm{},n_{D1},0),`$ (6.10)
where $`\psi _\mu (n_1,\mathrm{},n_{D1},0)`$ is periodic in the lattice coordinates and still to be determined. It must satisfy
$`\Delta _\mu \psi _\nu (n_1,\mathrm{},n_{D1},0)\Delta _\nu \psi _\mu (n_1,\mathrm{},n_{D1},0)=\omega _{\mu \nu }(n_1,\mathrm{},n_{D1},0),`$
$`{\displaystyle \underset{n_\mu =0}{\overset{L1}{}}}\psi _\mu (n_1,\mathrm{},n_{D1},0)=0.`$ (6.11)
These are the equations (6) restricted to $`n_D=0`$, and can be solved again by choosing $`\psi _{D1}(n_1,\mathrm{},n_{D1},0)=0`$. This procedure can be continued until all the $`\psi _\mu (n)`$ are found. This completes the proof of the existence of $`\psi _\mu `$ and, hence, $`\lambda _\mu `$ by (6.8).
To illustrate all these ideas in a concrete example, let us consider a constant fieldGeneral constant fields of arbitrary magnetic fluxes are explicitly given in ref. . $`F_{\mu \nu }(n)=ϵ_{\mu \nu }B`$ in two dimensions.<sup>\**</sup><sup>\**</sup>\**In two dimensions the Bianchi identities $`\Delta _{[\lambda }F_{\mu \nu ]}=0`$ are trivially satisfied for any anti-symmetric tensor fields. If we parameterize the link variables as in (2.3), $`a_\mu (n)`$ ($`\mu =1,2`$) satisfy
$`F_{12}(n)=\Delta _1a_2(n)\Delta _2a_1(n)+2\pi n_{12}(n)=B,(\pi a_\mu (n)<\pi ,|n_{12}(n)|2).`$ (6.12)
The magnetic flux $`\varphi 2\pi \varphi _{12}`$ is given by
$`\varphi =BL^2={\displaystyle \underset{n\mathrm{\Lambda }}{}}2\pi n_{12}(n).`$ (6.13)
This implies that $`BL^2`$ must be an integer multiple of $`2\pi `$. For $`BL^2=2\pi `$ we may choose $`n_{\mu \nu }(n)`$ as
$`n_{\mu \nu }(n)`$ $`=`$ $`\delta _{\stackrel{~}{n}_1,[L/2]}\delta _{\stackrel{~}{n}_2,[L/2]},`$ (6.14)
where we have introduced periodic lattice coordinates
$`\stackrel{~}{n}_\mu =n_\mu Lϵ\left({\displaystyle \frac{n_\mu }{L}}+{\displaystyle \frac{1}{2}}\right),([L/2]<\stackrel{~}{n}_\mu [L/2])`$ (6.15)
with $`ϵ(x)`$ being the stair-step function defined by
$`ϵ(x)=n\text{for}n<xn+1,(n𝐙).`$ (6.16)
The $`a_\mu (n)`$’s now satisfy the following equation
$`\Delta _1a_2(n)\Delta _2a_1(n)=B2\pi n_{12}(n)=B2\pi \delta _{\stackrel{~}{n}_1,[L/2]}\delta _{\stackrel{~}{n}_2,[L/2]}.`$ (6.17)
We can solve this equation by following the procedure described in the proof of the lemma. We thus obtain
$`a_1(n)={\displaystyle \frac{2\pi }{L}}\stackrel{~}{n}_2\delta _{\stackrel{~}{n}_1,[L/2]},a_2(n)={\displaystyle \frac{2\pi }{L^2}}\stackrel{~}{n}_1.`$ (6.18)
The gauge potential $`A_\mu (n)`$ can be found by solving (3.10) in the present case. The $`m_\mu (n)`$’s are easily obtained from (3.15) as
$`m_1(n)=\delta _{\stackrel{~}{n}_1,[L/2]}ϵ\left({\displaystyle \frac{n_2}{L}}+{\displaystyle \frac{1}{2}}\right),m_2(n)=0.`$ (6.19)
In deriving this use has been made of the relation
$`\Delta _\mu \stackrel{~}{n}_\mu =1L\delta _{\stackrel{~}{n}_\mu ,[L/2]},(\text{no sum on}\mu ).`$ (6.20)
We thus find the gauge potential as
$`A_1(n)=a_1(n)+2\pi m_1(n)=LB\delta _{\stackrel{~}{n}_1,[L/2]}n_2,A_2(n)=B\stackrel{~}{n}_1.`$ (6.21)
The nontrivial periodicity property of the gauge potential is
$`A_1(n+L\widehat{2})A_1(n)=L^2B\delta _{\stackrel{~}{n}_1,[L/2]}=2\pi \delta _{\stackrel{~}{n}_1,[L/2]}.`$ (6.22)
It is also possible to write the gauge potential in the form (5.12) as
$`\stackrel{~}{A}_1(n)={\displaystyle \frac{2\pi }{L^2}}n_2=Bn_2,\stackrel{~}{A}_2(n)=0.`$ (6.23)
Finally the topological charge is given by
$`Q={\displaystyle \frac{1}{4\pi }}{\displaystyle \underset{n\mathrm{\Lambda }}{}}ϵ_{\mu \nu }F_{\mu \nu }(n)={\displaystyle \frac{L^2B}{2\pi }}=1.`$ (6.24)
## 7 Summary and discussion
We have argued that the topological charge obtained from the lattice generalization of the Chern character is indeed an integer related to the winding number of a U(1) bundle constructed from the link variables by a smooth interpolation. It picks up the topological structure of the underlying lattice originating from the periodicity. The configuration space of the link variables that is topologically trivial and connected to the trivial one becomes topologically disjoint by excising the exceptional field configurations. No two belonging to different connected components can be continuously deformed into each other without crossing the exceptional configurations where the topological charge may jump due to the discontinuities of the field strengths at the exceptional plaquette variables. Link variables with different magnetic fluxes belong to different connected components of the field configurations. We have established an explicit relation between the topological charge and the magnetic fluxes.
The construction presented in this paper is rather indirect. One might think that the interpolation to a smooth fiber bundle played an essential role and there might exist different interpolations leading to different topological charges. But such is not the case. The topological information of the underlying lattice is carried by the gauge potentials and the value of the topological charge is completely unique for a given gauge field configuration. In order to see that the topological charge (5.5) is indeed an integer given by (5.14) no interpolation to a smooth bundle is necessary. This way of understanding, however, makes the topological meaning of (5.5) obscure. So it is more desirable to have a formalism of topological invariants of fiber bundles over discrete lattices without leaning upon the interpolation method.
What is the implication of the present analysis on the index theorem of the Ginsparg-Wilson Dirac operator on the lattice? We infer that the topological charge (5.5) coincides with the index of the Ginsparg-Wilson Dirac operator on the periodic lattice $`\mathrm{\Lambda }`$ as the index theorem suggests. In the case of infinite lattice the chiral anomaly coincides with the Chern character. However, both the index and the topological charge are not well-defined in general. On the other hand both of them are well-defined in the case of finite lattice. Unfortunately, we have very little known about the precise connection between them. The situation gets even worse for nonabelian theories.<sup>††</sup><sup>††</sup>††As a related study, see ref. . The topological charges obtained so far in the literature by interpolations are too complicated to manipulate and no differential geometric construction of Chern characters leading to topological invariants seems to be available on the lattice. We must extend the theorem given in the Introduction to nonabelian theories in order to classify the topological invariants on the lattice. It is a challenging problem to establish the lattice extension of the index theorem.
One of us (K.W.) is very grateful to F. Sakata for the warm hospitality extended to him and to Faculty of Science of Ibaraki University for the financial support during his visit at Ibaraki University. |
warning/0001/gr-qc0001011.html | ar5iv | text | # Attractive and Repulsive Gravity*footnote **footnote *gr-qc/0001011, December 31, 1999
## I Introduction
Few observational facts appear to be as well established in physics as the attractive nature of gravity. However, despite this, recent cosmological observations have raised the possibility that under certain circumstances gravity might actually contain an effective repulsive component, to thus invite consideration of the degree to which, and of the specific set of conditions under which, gravity actually need be strictly attractive in the first place. In fact, familiar as attractive gravity is, its actual attractiveness stems from the a priori assumption that Newton’s constant $`G`$ be chosen (purely by hand) to actually be positive. Then with this positive $`G`$ being treated as a fundamental input parameter by both Newtonian gravity and its relativistic Einstein gravity generalization (and even by their quantum-mechanical string theory generalization as well for that matter), the universal attractive nature of gravity is then posited on all distance scales and for all possible gravitational field strengths. As such, this actually constitutes a quite severe extrapolation of observationally established gravitational information from the kinematic solar system distance scale weak gravity regime where it was expressly obtained in the first place. And indeed, when the standard Newton-Einstein gravitational theory is actually extended beyond its solar system origins, notwithstanding the successes that are then encountered, nonetheless, disturbingly many difficulties are also encountered, in essentially every single such type of extrapolation that is in in fact made. Thus its extrapolation to larger distances such as galactic leads to the need for as yet unestablished galactic dark matter, its extrapolation to strong gravity leads to singularities and the development of event horizons and trapped surfaces in the fabric of spacetime, its extrapolation to the high temperature early universe leads to a cosmology with fine tuning and cosmological dark matter problems as well as to the notorious cosmological constant problem, and its extrapolation to large quantum field theoretic momenta far off the mass shell leads to uncontrollable renormalization infinity problems. Now while all of these issues may ultimately be resolved in favor of the standard theory, it is important to emphasize that all of them essentially arise from using just the first few measured weak, perturbative terms in a series (such as the first few terms in the Schwarzschild metric, the only such terms in that metric which have so far been observationally tested in fact) to try to guess the rest of the series. With there thus being many possible extrapolations of the standard Schwarzschild weak gravity solar system wisdom, extrapolations which can be just as covariant as the standard one, there are thus many possible departures from the standard gravitational intuition when gravity is extended to altogether different conditions, and in this paper we shall explore the possibility that, even as it acts attractively on solar system distance scales, gravity nonetheless gets to act repulsively on the much larger one associated with cosmology.
## II How the Standard Intuition Came About
In developing a fundamental theory of planetary motion Newton found that an inverse square force law with a universal gravitational constant $`G`$ not merely accounted for the Keplerian elliptical orbital motions of the planets around the sun, but equally, it also described the gravitational motions of objects near the surface of the earth, i.e. it described gravity not only on solar system distance scales but also on altogether smaller terrestrial distance scales as well. In this way Newton’s law of gravity thereby acquired a universal character causing it to come to be regarded as a universal law which was then to also be valid on altogether larger distance scales as well even though it had not in fact been tested on them. However, with the advent of galactic astronomy eventually then permitting such testing on these altogether larger distance scales, it was actually found (see e.g. for recent reviews) that the detailed orbits of stars and galactic gas in galaxies did not in fact conform to the ones expected on the basis of the known luminous matter content of the galaxies. In fact long ago Zwicky had already noted an analogous problem in (presumed virialized) clusters of galaxies, with the discrepancy he found between the measured mean kinetic energy of the visible galaxies within the cluster and the mean gravitational potential energy generated by those selfsame galaxies leading him to conclude that the continuing applicability of Newton’s law of gravity entailed that there would have to be far more matter in the cluster than he was able to detect. Then, once it became clear that discrepancies such as these were in fact common and even widespread on these large distance scales, a dark matter paradigm was then adopted, a paradigm in which any detected discrepancy between the known luminous matter Newtonian expectation and observation was to then be accounted for by whatever amount of non-luminous matter would then (i.e. only after the fact) specifically be required. Moreover, with there not as yet being a single application or test of Newtonian gravity on these large distance scales (in either galaxies or clusters of galaxies) which does not involve an appeal to this as yet unestablished and still poorly understood dark matter, we thus see the complete circularity of the chain of reasoning which leads to dark matter in the first place, since one assumes the continuing validity of Newton’s law of gravity and then posits the presence of just the appropriate amount of dark matter needed in order to maintain the validity of the presupposed Newtonian law. Apart from not being a particularly satisfying prescription, this galactic dark matter paradigm does yet even qualify as being a falsifiable theory (the sine qua non of physical theory) since it has not yet been brought to the point where it can actually make definitive and expressly falsifiable galactic predictions, i.e. given a detected amount of luminous matter in a given galaxy, dark matter theory should be able to predict both the amount and spatial distribution of the dark matter in the galaxy in advance of any measurement of a galactic rotation curve. Moreover, even more disturbing than this is the concomitant need of having to now retroactively explain why dark matter is not also required in order to fit solar system planetary orbits, i.e. the need to explain why luminous matter alone should in fact be capable of providing a complete accounting of solar system dynamics in the first place. A model in which dark matter is there when needed and not when not hardly qualifies as being a model at all, with its most conservative characterization actually being that in fact dark matter is nothing more than a phenomenological parameterization of the detected departure of the luminous Newtonian expectation from observation. This then is the galactic dark matter problem, and while it may eventually be resolved by the actual direct detection of dark matter, nonetheless it could equally well be signaling a failure of the standard Newtonian wisdom and intuition on these very large distance scales.
While some Newtonian wisdom was in fact supplanted by the subsequent development of Einstein relativity, it is curious that its development in fact actually served to reinforce the particular Newtonian intuition that we described above. Specifically, even while the general relativistic curved spacetime Einstein gravitational theory replaced the strictly non-relativistic Newtonian gravitational one, nonetheless the Einstein theory still recovered the Newtonian theory in the non-relativistic limit, even as it prescribed relativistic corrections to it. The actual observational confirmation of these relativistic corrections not merely served to establish the validity of Einstein gravity, it also served to reinforce the validity of Newtonian gravity whenever the non-relativistic limit could appropriately be taken. However, since these relativistic corrections were themselves only established on solar system distance scales (cf. the first few terms in a perturbative solar system Schwarzschild metric expansion or the first few perturbative terms in the metric of a binary pulsar, a similarly sized such system), the extrapolation of Einstein gravity (and of its universal character) to galactic distance scales and beyond was then no more secure than had been the extrapolation of Newtonian gravity to those same distance scales (with the Einstein equations thus only being as secure as dark matter). Thus again a new intuition was acquired, one in which the very presence of Newton’s constant as an a priori fundamental coupling constant in the Einstein-Hilbert action then endowed $`G`$ with an even more fundamental and universal status than it had actually previously possessed. Moreover, in giving Newton’s constant such a status, a dichotomy is immediately set up between gravity and the standard $`SU(2)\times U(1)`$ electroweak theory, a theory where another dimensionful phenomenological parameter, Fermi’s constant $`G_F`$, is not in fact elevated into a fundamental parameter at all, a theory which can then be made completely renormalizable precisely because this is not in fact done, with $`G_F`$ itself then emerging solely as an effective parameter which is only of relevance at low energies.
That Einstein gravity was taken to be universal was hardly surprising given its very general geometric character. However, it is important to distinguish between the geometric and dynamical aspects of the Einstein theory, and even while these two issues are logically distinct, nonetheless the Einstein theory is ordinarily regarded as being one integral and indivisible package. However, the standard theory both geometrizes gravity by identifying the spacetime metric $`g^{\mu \nu }`$ as the gravitational field, and then determines its dynamics by imposing the second order Einstein equations
$$R^{\mu \nu }g^{\mu \nu }R_\alpha ^\alpha /2=8\pi GT^{\mu \nu }$$
(1)
where $`R^{\mu \nu }`$ is the Ricci tensor associated with the geometry and $`T^{\mu \nu }`$ is the energy-momentum tensor of its gravitational source. However, even while insisting that the metric is to describe the gravitational field, nonetheless, without giving up general covariance, its dynamical equations of motions could still depart from the above second order Einstein equations, and would in fact readily do so if the gravitational equations were to be obtained from the variation of some equally covariant general coordinate scalar action other than the standard Einstein-Hilbert one ($`I_{EH}=d^4x(g)^{1/2}R_\alpha ^\alpha /16\pi G`$) which is ordinarily used. (In fact the fully covariant string gravitational theory, for instance, replaces the Einstein equations by a set of equations which contains an entire, infinite series of derivatives of the Riemann tensor.) One thus has to distinguish between the fact of curvature (viz. geometry) and the amount of curvature (viz. dynamics), and in particular one has to go over both the successes and the problems of Einstein gravity to ascertain which are due to geometry and which to dynamics, and should it turn out that the successes are predominantly due to geometry while the problems arise from a particular assumed dynamics, we would then be able to identify the extrapolation of the standard gravity equations of motion beyond their solar system origin as the root cause of the problems that standard theory currently encounters, while not at the same time needing to give up the underlying geometrical picture.
Indeed, the very cornerstone of standard gravity, viz. the equivalence principle, is completely geometrical and has nothing to do with dynamics at all. Rather it is a statement about the geometric nature of geodesics, with particles which follow such geodesics then having to uniquely couple to an external gravitational field according to
$$\frac{d^2x^\lambda }{d\tau ^2}+\mathrm{\Gamma }_{\mu \nu }^\lambda \frac{dx^\mu }{d\tau }\frac{dx^\nu }{d\tau }=0$$
(2)
regardless of the particular form of the equation of motion obeyed by the external gravitational field itself. Thus, even while the magnitudes of the Christoffel symbols are of course sensitive to the dynamics associated with the background gravitational field, nonetheless, their very presence in the geodesic equations to begin with is strictly geometric. Gravitational bending, lensing, redshifting and time delaying of light, modifications to Newtonian planetary orbits, and the decay of the orbit of a binary pulsar (a consequence of the retarded nature of the gravitational radiation reaction of each of the two stars in the binary on the other due to the finite limiting velocity with which gravitational information is communicated) will thus all be found to occur in any strictly covariant metric theory of gravity, with the dynamical equations only controlling the magnitude of these effects but not the fact of their existence.In passing it is important to note that there is actually a hidden assumption in using geodesic equations to describe gravitational phenomena, viz. the assumption that real particles can be treated as classical geodesic test particles in the first place. Now while the geodesic assumption is immediately valid for the massless rays of the eikonal approximation to wave theory (since the rays are already geodesic in flat spacetime, with a covariantizing of their motion then making them geodesic in a background gravitational field as well), the situation regarding material particles is not at all as straightforward, with there actually being no justification for associating real particles (i.e. the excitations of the quantum fields of elementary particle physics) with the classical test particle action $`I_T=mc𝑑\tau `$, even though the variation of this action would enforce geodesic motion. It is thus of interest to note, that in a recent study of the extension of the equivalence principle to the propagation of quantum-mechanical matter waves in a background classical gravitational field (with a curved space Schrodinger equation explicitly being shown to be equivalent to an accelerating coordinate frame flat space one), it was shown that it is only because there is such a quantum extension that the equivalence principle actually gets to hold for real classical particles at all. Thus it was shown that gravity, itself a field theory, couples first and foremost to fields rather than to particles (i.e. first and foremost to wavelength rather than to mass - with the eikonal rays associated with the minimally coupled scalar field wave equation $`S_{;\mu }^{;\mu }(mc/\mathrm{})^2S=0`$ indeed being geodesic), and then, only upon second quantization ($`\lambda =\mathrm{}/mv`$) of those rays, to mass. It is thus only because of quantum mechanics that massive classical particles get to be geodesic at all.
Moreover, as regards the uniqueness of any possible underlying dynamics, it was noted by Eddington in the very early days of relativity that the familiar standard gravity exterior $`R^{\mu \nu }=0`$ Schwarzschild solution (the one used in the standard solar system relativistic tests) is just as equally a solution to higher derivative gravitational theories as well, since the vanishing of the Ricci tensor entails the vanishing of its derivatives as well, with solar system tests thus not in fact being able to definitively exclude gravitational actions other than the Einstein-Hilbert one after all. Moreover, since such higher derivative theories turn out to then have different continuations to larger distances , the possibility then emerges that the need for galactic dark matter is only an artifact of using the Einstein gravity continuation. And should that be the case, the further continuation to cosmology (a regime which is to a good degree controlled by the high symmetry of geometries such as Robertson-Walker and de Sitter rather than by the structure of the explicit dynamical evolution equations themselves) would then potentially become unreliable too.
In order to illustrate the above remarks in an explicit example it is convenient to consider a particular alternate gravitational theory, viz. conformal gravity, a general coordinate invariant pure metric theory of gravity which possesses an explicit additional and highly restrictive symmetry (invariance of the geometry under any and all local conformal stretchings $`g_{\mu \nu }(x)\mathrm{\Omega }^2(x)g_{\mu \nu }(x)`$) as well. In fact, so restrictive is this symmetry that it allows only one unique gravitational action (an action which is to thus replace the standard $`I_{EH}`$), viz.
$$I_W=\alpha _gd^4x(g)^{1/2}C_{\lambda \mu \nu \kappa }C^{\lambda \mu \nu \kappa }$$
(3)
where $`C^{\lambda \mu \nu \kappa }`$ is the conformal Weyl tensor and where the gravitational coupling constant $`\alpha _g`$ introduced here is a universal dimensionless one, to thereby endow gravity with a structure similar to that found in the electroweak interaction case described above, with such a gravitational theory actually then being power counting renormalizable. For this theory variation of the action leads to the equations of motion
$$(g)^{1/2}\delta I_W/\delta g_{\mu \nu }=2\alpha _gW^{\mu \nu }=T^{\mu \nu }/2$$
(4)
where $`W^{\mu \nu }`$ is given by
$`W^{\mu \nu }=g^{\mu \nu }(R_\alpha ^\alpha )_{;\beta }^{;\beta }/2+R_{;\beta }^{\mu \nu ;\beta }R_{;\beta }^{\mu \beta ;\nu }R_{;\beta }^{\nu \beta ;\mu }2R^{\mu \beta }R_\beta ^\nu +g^{\mu \nu }R_{\alpha \beta }R^{\alpha \beta }/2`$ (5)
$`2g^{\mu \nu }(R_\alpha ^\alpha )_{;\beta }^{;\beta }/3+2(R_\alpha ^\alpha )^{;\mu ;\nu }/3+2R_\alpha ^\alpha R^{\mu \nu }/3g^{\mu \nu }(R_\alpha ^\alpha )^2/6,`$ (6)
so that we can immediately confirm that the Schwarzschild $`R^{\mu \nu }=0`$ solution is indeed an exterior solution to the theory just as Eddington had warned us. Standard gravity is thus seen to be only sufficient to give the standard Schwarzschild phenomenology but not at all necessary, with it thus indeed being possible to bypass the Einstein-Hilbert action altogether as far as low energy phenomena are concerned.
Further insight into the structure of this alternate gravity theory is obtained by noting that for a static, spherically symmetric source such as a star, the conformal symmetry allows one to set $`g_{rr}=1/g_{00}`$ without any loss of generality, with the field equations of Eq. (4) then being found to reduce (without any approximation at all, i.e. for strong and weak gravitational fields alike) to
$$^4g_{00}=3(T_0^0T_r^r)/4\alpha _gg_{00}f(r),$$
(7)
i.e. to reduce to a fourth order Poisson equation rather than to the second order one familiar in the standard theory. The general solution to Eq. (7) exterior to a star of radius $`R`$ is then given by
$$g_{00}(r>R)=12\beta ^{}/r+\gamma ^{}r$$
(8)
with the coefficients being given by
$$\beta ^{}=_0^R𝑑rf(r)r^4/12,\gamma ^{}=_0^R𝑑rf(r)r^2/2,$$
(9)
i.e. by two different moments of the source. We thus see that the Newtonian potential need not be associated with either the second order Einstein equations or with their non-relativistic second order Poisson equation limit, and that the standard theory is thus only sufficient to give Newton but not at all necessary. (In order to show the lack of necessity of the standard theory it is sufficient to construct just one alternative.) The Newtonian potential can thus just as readily be generated in higher order gravitational theories as well, theories which can then have a very different behavior on altogether larger distance scales. And indeed, through the use of the linear $`\gamma ^{}r`$ potential term, a term which actual dominates over Newton at large enough distances, conformal gravity was actually found capable of providing for a complete accounting of galactic rotation curves without the need to invoke dark matter at all. (In fact, the value for the coefficient $`\gamma ^{}`$ required by galactic data then entailed the numerical irrelevance of the $`\gamma ^{}r`$ term on the much smaller solar system distance scales, to thus yield a solution of the galactic rotation curve problem which naturally leaves standard solar system physics intact.) However, regardless of the specific merits of any particular alternate gravitational theory such as conformal gravity itself, our analysis here does serve to underscore the risks inherent in extrapolating solar system wisdom beyond the confines of the solar system, with the need for galactic dark matter perhaps being symptomatic of the lack of applicability of one particular such extrapolation.
With regard to this conformal gravity alternative, we note further, that in it, with the coefficient $`\beta ^{}`$ of a star being given as the energy-momentum tensor moment integral of Eq. (9), an identification of this same moment integral in the case of a single proton or neutron with one half of the $`2\beta _p`$ Schwarzschild radius of a single nucleon, then enables us to identify the stellar coefficient $`\beta ^{}`$ in conformal gravity as $`N\beta _p`$ for a weak gravity star containing $`N`$ such nucleons. This relation is completely identical to the one obtained by solving the standard theory second order Poisson equation for a weak gravity source composed of $`N`$ fundamental elementary particle sources each contributing a potential $`\beta _p/r`$, to thus enable us to recover the familiar extensive property of the Newtonian potential of weak gravity bulk matter, while also seeing that it need not be tied exclusively to the second order theory. Moreover, as far as gravitational sources are concerned, the coefficients of their $`1/r`$ potentials are actually radii, specifically their gravitational Schwarzschild radii, so that an identification of $`\beta _p`$ with $`Gm_p/c^2`$ where $`m_p`$ is the mass of a proton (to thus define $`G`$ once and for all as $`c^2\beta _p/m_p`$, a purely microscopic quantityIn a conformal invariant theory dimensionful microscopic parameters such as $`\beta _p`$ and $`m_p`$ can only be explicitly generated when the conformal symmetry is spontaneously broken, with their values then being fixed by the details of the symmetry breaking mechanism. However, regardless of any specific dynamics that may be needed to explicitly do this, the dependence of the solution of Eq. (8) on the radial coordinate $`r`$ is already the most general allowable one possible.) then entails that for a weak gravity bulk matter star $`\beta ^{}`$ is given as $`GM^{}/c^2`$ where $`M^{}=Nm_p`$ is the mass of the star. We thus see (i) that the universality of $`G`$ need not be tied to the second order Poisson equation, (ii) that, just like Boltzmann’s constant, Newton’s constant $`G`$ need not itself be fundamental (only the product $`MG/c^2`$ is ever measurable gravitationally and never $`G`$ itself - just as only the product $`kT`$ is measurable in statistical mechanics), and (iii) that $`G`$ need not have any applicability at all in the strong gravity limit where the energy-momentum tensor moment integrals no longer scale linearly with the number of fundamental sources. Thus not only might our standard gravitational intuition not necessarily be generalizable to large distance scales, it might also not be generalizable to strong gravity either.<sup>§</sup><sup>§</sup>§In fact, for an appropriate gravitational self-energy contribution, the moment integrals of Eq. (9) might even have differing signs in the weak and strong gravity limits.
## III When an Attractive Potential is Repulsive
Even though it is generally thought that the attractive or repulsive nature of a potential is determined once and for all by its overall sign, in this section we show that this turns out to not in fact necessarily always be the case, with this then being another piece of the standard intuition which would appear to require reappraisal. To explicitly illustrate this specific point it is convenient to consider the geometry near the surface of a static, spherically symmetric gravitational source of radius $`R`$ and mass $`M`$ with exterior geometrical line element of the form
$$d\tau ^2=B(r)c^2dt^2dr^2/B(r)r^2d\mathrm{\Omega }$$
(10)
where the metric coefficient $`B(r)`$ is given not just by the usual Schwarzschild form, but rather by the more general
$$B(r)=12MG/c^2r+\gamma r$$
(11)
form we introduced earlier. If we erect a Cartesian coordinate system $`x=r`$sin$`\theta `$cos$`\varphi `$, $`y=r`$sin$`\theta `$sin$`\varphi `$, $`z=r`$cos$`\theta R`$ at the surface of the source, then, with $`z`$ being normal to the surface, to lowest order in $`x/R,y/R,z/R,MG/c^2R(=gR/c^2)`$ and $`\gamma R`$ the line element is then found to take the form
$$d\tau ^2=[1a(z)]c^2dt^2dx^2dy^2[1+a(z)]dz^2b(xdx+ydy)dz$$
(12)
where $`a(z)=2g(Rz)/c^2\gamma (R+z)`$ and $`b=4g/c^22\gamma `$. For trajectories for which the initial velocity $`v`$ is in the horizontal $`x`$ direction, the geodesic equations associated with Eq. (12) take the form (the dot denotes differentiation with respect to the proper time $`\tau `$ for massive particles or with respect to an affine parameter for massless ones)
$$\ddot{t}=0,\ddot{x}=0,\ddot{y}=0,\ddot{z}+2(g\gamma c^2/2)v^2/c^2+g+\gamma c^2/2=0.$$
(13)
Thus in the non-relativistic $`v=0`$ case the motion is described by
$$\ddot{z}+g+\gamma c^2/2=0,$$
(14)
while in the relativistic $`v=c`$ case it is described by
$$\ddot{z}+3g\gamma c^2/2=0$$
(15)
instead. Thus we see that the effect of the $`\gamma `$ term is actually opposite in these two limits, with positive $`\gamma `$ leading to attractive bending for slow moving particles but to repulsive deflection for fast moving ones.This deflection of light away from a source also holds for the exact $`B(r)=1+\gamma r`$ geodesic as well . Thus the fact that a potential may be attractive for non-relativistic motions does not in and of itself mean that it must therefore also be attractive for light, with the $`v^2/c^2`$ type terms not only modifying the magnitude of the effect of gravity (something already the case even in the standard theory where $`\ddot{z}+g(1+2v^2/c^2)=0`$), but even being able to modify the sign of the effect as well. Thus in general we see that even after appropriately fixing the sign of the coefficient of a gravitational potential term once and for all, such a potential need not always lead to attraction, with a potential which would ordinarily be considered to be attractive (as defined by non-relativistic binding) still being able to lead to repulsion in other kinematic regimes. Caution thus needs to be exercised before one can conclude that attraction in one kinematic regime entails attraction in all others as well.
As regards the discussion of the metric given in Eq. (12) some further comment is in order. Specifically, noting that the transformation
$`t^{}=t[1g(Rz)/c^2+\gamma (R+z)/2],x^{}=x,y^{}=y,`$ (16)
$`z^{}=z[1+g(2Rz)/2c^2\gamma (2R+z)/4]+(g+\gamma c^2/2)t^2/2+(g/c^2\gamma /2)(x^2+y^2)`$ (17)
brings the metric of Eq. (12) to the flat coordinate form $`d\tau ^2=c^2dt^2dx^2dy^2dz^2`$, we see that the unprimed system origin obeys
$$z^{}=(g+\gamma c^2/2)t^2/2$$
(18)
in the primed Cartesian coordinate system. We thus provide a direct demonstration of the equivalence principle with gravity indeed being found to act the same way as an acceleration in flat spacetime, and with the equivalence principle indeed being seen to have validity beyond the standard second order Einstein theory.
Now while we have just seen that the weak gravity metric of Eq. (12) is equivalent to an acceleration in flat spacetime, this result is initially somewhat puzzling since the full starting metric of Eqs. (10) and (11) is not only not at all flat, it even possesses a Riemann tensor which is explicitly non-zero (and thus explicitly not flat) even in this very same lowest order in $`g`$ (or $`\gamma `$) under which Eq. (12) was actually derived. The answer to this puzzle is that while the Christoffel symbols are first order derivative functions of the metric, the Riemann tensor itself is a second order such derivative. Thus to get the lowest non-trivial term in the Riemann tensor we need to expand the metric to second order in $`x/R,y/R,z/R`$. Since a first order expansion suffices for the Christoffel symbols, we thus see that there is actually a mismatch between orders of expansion of the Christoffel symbols and the Riemann tensor. Hence a first order study of the geodesics is simply not sensitive to the curvature, and thus we not only see why the equivalence principle works for weak gravity near the surface of a system such as the earth, we even see why it has to do so.With $`g_{;\nu }^{\mu \nu }`$ vanishing identically, Riemannian geometries possess no non-trivial covariant first order derivative function of the metric at all. Consequently, lowest order trajectories can only be described by non-tensors such as the coordinate dependent Christoffel symbols, with only the sum of the two quantities $`d^2x^\lambda /d\tau ^2`$ and $`\mathrm{\Gamma }_{\mu \nu }^\lambda (dx^\mu /d\tau )(dx^\nu /d\tau )`$ actually transforming as a contravariant vector. In lowest order then, the coupling of a particle to gravity has to be purely inertial. Moreover, on recognizing this fact, we immediately realize that a typical non-geodesic but still fully covariant particle equation of motion such as<sup>\**</sup><sup>\**</sup>\**This (purely illustrative) equation of motion can actually be derived by a stationary variation of the action $`I=mc𝑑\tau \kappa _1𝑑\tau R_\beta ^\beta `$ where $`\kappa _1`$ is an appropriate constant.
$`m\left({\displaystyle \frac{d^2x^\lambda }{d\tau ^2}}+\mathrm{\Gamma }_{\mu \nu }^\lambda {\displaystyle \frac{dx^\mu }{d\tau }}{\displaystyle \frac{dx^\nu }{d\tau }}\right)=\kappa _1R_\beta ^\beta \left({\displaystyle \frac{d^2x^\lambda }{d\tau ^2}}+\mathrm{\Gamma }_{\mu \nu }^\lambda {\displaystyle \frac{dx^\mu }{d\tau }}{\displaystyle \frac{dx^\nu }{d\tau }}\right)`$ (19)
$`\kappa _1R_{\beta ;\alpha }^\beta \left(g^{\lambda \alpha }+{\displaystyle \frac{dx^\lambda }{d\tau }}{\displaystyle \frac{dx^\alpha }{d\tau }}\right)`$ (20)
will just as readily satisfy the same weak gravity tests as the Eq. (2) geodesic itself. Moreover, since Eq. (2) and Eq. (20) both reduce to the Cartesian coordinate Newtonian law $`md^2x^\lambda /d\tau ^2=0`$ in the absence of curvature, both equations represent valid curved space generalizations of Newton’s second law of motion, with Eq. (20) possessing both inertial (the Christoffel symbols) and non-inertial (the Ricci scalar) contributions.<sup>††</sup><sup>††</sup>††In passing it is perhaps worth stressing that some relativists regard the equivalence principle as the statement that gravitational effects are strictly inertial, with Eq. (2) then being the only possible allowable coupling of a particle to gravity. However, we take the far more cautious view here that while the Christoffel symbols certainly do possess the nice, purely geometric inertial property of being simulatable by an acceleration in flat spacetime, that does not, and in fact cannot, preclude there being a non-inertial, truly coordinate independent, coupling to gravity as well, with this issue actually being a dynamical rather than a geometrical one which is only decidable by consideration of the curved spacetime field equations with which real (as opposed to test) particles are associated. With Eqs. (2) and (20) having very different strong gravity continuations, we thus see that the weak gravity successes of the equivalence principle reveal nothing about how Eq. (2) would in fact fare in strong gravity, with an Eotvos experiment near the surface of a black hole not being at all guaranteed to give a null result. Since the presence of such non-inertial terms is actually expected to be the general rule rather than the exception in field theory,<sup>‡‡</sup><sup>‡‡</sup>‡‡The electromagnetic vector potential equation of motion $`g^{\alpha \beta }A_{\mu ;\alpha ;\beta }A_{;\alpha ;\mu }^\alpha +A^\alpha R_{\mu \alpha }=0`$ contains an explicit non-inertial piece in curved spacetime, with a similar situation being found for both the Dirac field (vierbeins being non-inertial) and for the non-minimally coupled scalar field $`S(x)`$ with equation of motion $`S_{;\mu }^{;\mu }(mc/\mathrm{})^2S+\xi R_\alpha ^\alpha S/6=0`$ where $`\xi `$ is a dimensionless parameter. it would appear that particle motions may not in fact be controlled by the strong gravity event horizons and trapped surfaces associated with Schwarzschild metric geodesic motion after all, with gravity itself possibly being able to protect particles from such phenomena in the strong gravity limit,<sup>\**</sup><sup>\**</sup>\**Not only could non-inertial effects such as those exhibited in Eq. (20) become significant in strong gravitational fields, there could even be a switch over to the effective repulsion associated with $`B(r)=1+\gamma r`$ type metrics as particles are accelerated to high velocities. with the strong gravity extrapolation of standard weak gravity thus being another potentially unreliable extrapolation.
## IV The Case for Repulsive Gravity
Recently, through study of type 1A supernovae at very high redshift , it has become possible to explore the attractive or repulsive character of gravity on cosmological distance scales, with it now being possible to determine whether the universe itself might indeed be slowing down or whether it might perhaps actually be speeding up. In terms of the standard Robertson-Walker cosmological line element
$$d\tau ^2=c^2dt^2R^2(t)[(1kr^2)^1dr^2+r^2d\mathrm{\Omega }]$$
(21)
with associated scale factor $`R(t)`$ and spatial curvature $`k`$, the new high $`z`$ data have made it possible to extend Hubble plot measurements of the temporal behavior of $`R(t)`$ beyond lowest order in time, with the current era value of the $`q(t_0)=\ddot{R}(t_0)R(t_0)/\dot{R}^2(t_0)`$ deceleration parameter having now become as amenable to observation as the current value of the $`H(t_0)=\dot{R}(t_0)/R(t_0)`$ Hubble parameter itself. Moreover, the actual observations themselves have now provided the first direct evidence that gravity might actually contain an explicit repulsive component. Specifically, in terms of the standard Einstein-Friedmann cosmological evolution equations, viz.
$$\dot{R}^2(t)+kc^2=\dot{R}^2(t)(\mathrm{\Omega }_M(t)+\mathrm{\Omega }_\mathrm{\Lambda }(t))$$
(22)
and
$$q(t)=(n/21)\mathrm{\Omega }_M(t)\mathrm{\Omega }_\mathrm{\Lambda }(t)=(n/21)(1+kc^2/\dot{R}^2(t))n\mathrm{\Omega }_\mathrm{\Lambda }(t)/2$$
(23)
where $`\mathrm{\Omega }_M(t)=8\pi G\rho _M(t)/3c^2H^2(t)`$ is due to ordinary matter (viz. matter for which $`\rho _M(t)=A/R^n(t)`$ where $`A>0`$ and $`3n4`$) and where $`\mathrm{\Omega }_\mathrm{\Lambda }(t)=8\pi G\mathrm{\Lambda }/3cH^2(t)`$ is due to a possible cosmological constant $`\mathrm{\Lambda }`$, it was found that the data constrained the allowable current (n=3) era values of the parameters $`\mathrm{\Omega }_M(t_0)`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ to a quite small region in which $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)\mathrm{\Omega }_M(t_0)+1/2`$ or so, with (the presumed positive) $`\mathrm{\Omega }_M(t_0)`$ being limited to the range $`(0,1)`$ or so and with $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ being limited to the range $`(1/2,3/2)`$ or so, to thus yield a current era deceleration parameter $`q(t_0)`$ which had to lie within the expressly negative $`(1/2,1)`$ interval. While these data thus appear to point toward a universe which is actually currently accelerating, systematic effects (such as an apparent evolutionary effect between high and low $`z`$ supernovae ) are actually large enough that the data could still support a positive value for $`q(t_0)`$, albeit one which would however still have to be substantially smaller than the $`q(t_0)=1/2`$ value expected in the standard $`\mathrm{\Omega }_k(t_0)=kc^2/\dot{R}^2(t_0)0`$ flat inflationary universe paradigm in which $`\mathrm{\Omega }_M(t_0)=1`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)=0`$. Thus even if the universe is not accelerating, there would still have to be some cosmic repulsion with respect to standard inflation (i.e. with respect to the expressly positive $`q(t_0)=(n/21)\mathrm{\Omega }_M(t_0)1/2`$ associated with normal, gravitationally attractive matter in a flat universe), even if this needed cosmic repulsion did not actually dominate over the positive $`\mathrm{\Omega }_M(t_0)`$ contribution and lead to a net overall acceleration. The data thus entail the existence of some form or other of repulsive component to gravity on cosmological distance scales, and it is to the possibility that gravity need not always be strictly attractive which we therefore now turn.
From a purely phenomenological viewpoint, Eq. (23) immediately suggests two fairly straightforward ways in which $`q(t_0)`$ could in fact be reduced below one half in standard gravity, viz. a positive $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ or a negative spatial curvature $`k`$. Of these proposals the possibility of a non-vanishing $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ was first raised by Einstein himself as long ago as the very early days of relativity. Specifically, motivated by the desire to have a static universe, he noted that if he modified the Einstein-Hilbert action by the addition of a fundamental cosmological constant term, the associated cosmology would then admit of a static solution (with $`\dot{R}(t)=0`$, $`\ddot{R}(t)=0`$ and fixed $`R(t)=R_0`$ in the Robertson-Walker language) provided $`\rho =A/R_0^3=2c\mathrm{\Lambda }=kc^4/4\pi GR_0^2`$, i.e. provided $`\mathrm{\Lambda }`$ and $`k`$ were both taken to be positive. In such a solution a positive cosmological constant term would act repulsively to counteract the attraction (i.e. deceleration) associated with positive $`\rho `$ and positive $`k`$, to thus yield a (closed) static universe. While such a cosmology quickly fell into disfavor following the discovery of the cosmic recession of the nebulae shortly thereafter, it did raise for the first time the issue of cosmic repulsion, while also raising a problem that has been with us ever since, the notorious cosmological constant problem, a problem whose modern variant is not so much one of the possible presence of a fundamental macroscopic $`\mathrm{\Lambda }`$ but rather of the possible presence of a (potentially unacceptably large) microscopically induced one instead.
As regards the second way to get cosmic repulsion, viz. negative curvature, we note that such a mechanism is not in fact tied to any specific cosmological evolution equation such as the Einstein-Friedmann one of Eqs. (22) and (23), with it actually turning out to be quite general in nature. Specifically, it was shown that, no matter what the explicit form of the gravitational field equations of motion themselves, the propagation of waves such as Maxwell waves in a general curved spacetime Robertson-Walker background is completely analogous to the propagation of flat spacetime Maxwell waves in a material medium, with vector (and also scalar) curved space waves being found to have a dispersion relation of the form $`\omega ^2/c^2=\lambda ^2+k`$ in a Robertson-Walker background. Thus we see that spatial curvature acts just like a frequency dependent refractive index of the form $`n(\omega )=c\lambda /\omega =(1kc^2/\omega ^2)^{1/2}`$, with the group velocity associated with energy transport then being given by $`v_g=d\omega /d\lambda =c(1kc^2/\omega ^2)^{1/2}`$. When $`k`$ is negative, spatial curvature is then seen to act as a tachyonic mass, to thus effectively give faster than light propagation, with a negative curvature space then acting just like a diverging dispersive medium wherein group velocities are explicitly greater than their values in empty space. Particles propagating in a negative curvature space are thus effectively accelerated (cf. a diverging lens), while those traveling in a positive curvature space are accordingly decelerated (cf. a converging lens). With the standard wisdom regarding the aftermath of the big bang being that the mutual attraction of the galaxies would serve to slow down the expansion of the universe (i.e. $`q(t)=\mathrm{\Omega }_M(t)/2>0`$), we now see that this particular wisdom only applies if the galaxies are propagating in an inert, empty space (viz. a $`k=0`$ plane lens) which itself has no dynamical consequences. However, once $`k`$ is non-zero, the galaxies would instead then propagate in a non-trivial geometric medium, a medium which can then explicitly participate dynamically, with the gravitational field itself which is then present in the medium being able to accelerate or decelerate the galaxies according to the sign of its spatial curvature. Thus we again see that the standard notion of purely attractive gravity needs to be reconsidered once the cosmological curvature $`k`$ is non-zero, with gravity itself (viz. curvature) potentially being able to generate some repulsion all on its own.
While both of the above cosmic repulsion mechanisms can readily be utilized in standard gravity in order to provide a purely phenomenological fit to the supernovae data, nonetheless the specific values explicitly required of the parameters $`\mathrm{\Omega }_M(t_0)`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ actually pose a somewhat severe theoretical problem for the standard theory. Specifically, given the radically differing temporal behaviors of $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ and $`\mathrm{\Omega }_M(t)`$, the apparent current closeness to one of their ratio entails that in the early universe this same $`\mathrm{\Omega }_\mathrm{\Lambda }(t)/\mathrm{\Omega }_M(t)`$ ratio would have had to have been fantastically small (typically of order as small as $`10^{60}`$), with a Friedmann universe only being able to evolve into the currently observed one if this ratio had been extremely fine-tuned by fixing the initial conditions in the early universe to incredible accuracy. Moreover, even if we take advantage of the systematic uncertainties identified in which might potentially (but not necessarily) permit us to set the current value of $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ to zero, since $`\mathrm{\Omega }_M(t_0)`$ would still be required to be less than one even in such a case, Eq. (22) would then oblige the current value of $`\mathrm{\Omega }_k(t_0)`$ to be necessarily different from zero. Then, given the different temporal behaviors of $`\mathrm{\Omega }_M(t)`$ and $`\mathrm{\Omega }_k(t)`$, this time it would be the $`\mathrm{\Omega }_k(t)/\mathrm{\Omega }_M(t)`$ ratio which would need early universe fine-tuning. With Eq. (22) being writable as $`\mathrm{\Omega }_M(t)+\mathrm{\Omega }_\mathrm{\Lambda }(t)+\mathrm{\Omega }_k(t)=1`$, we thus see that current era non-vanishing of any two of these three quantities entails some form of early universe fine-tuning problem. Since the current data do not support the one point (viz. the $`\mathrm{\Omega }_M(t_0)=1`$, $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)=0`$, $`\mathrm{\Omega }_k(t_0)=0`$ inflationary universe) which could be reached without a fine-tuning of Eq. (22), we see that, even with the current observational uncertainties, the new high $`z`$ data will not support a Friedmann cosmology without some form or other of fine-tuning problem, with the non-vanishing of $`1\mathrm{\Omega }_M(t)`$ entailing the explicit presence of some cosmic repulsion.
Moreover, apart from the above macroscopic fine-tuning problems, microscopic quantum physics presents cosmology with yet more problems, with the current value of the $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)/\mathrm{\Omega }_M(t_0)`$ ratio actually being expected to be absolutely enormous - potentially of order $`10^{120}`$ if generated by quantum gravity, and of typical order $`10^{60}`$ if generated by elementary particle physics phase transitions such as the electroweak one.<sup>\*†</sup><sup>\*†</sup>\*†Associating a typical temperature scale $`T_V`$ with the electroweak vacuum breaking phase transition and a temperature $`T(t)`$ with the ordinary matter in the universe leads to $`\mathrm{\Omega }_\mathrm{\Lambda }(t)/\mathrm{\Omega }_M(t)=T_V^4/T^4(t)`$, a ratio which is of order $`10^{60}`$ today. Microscopic physics thus leads to an expectation which is nowhere near the current data at all, with this then being the modern variant of the cosmological constant problem to which we referred earlier.<sup>\*‡</sup><sup>\*‡</sup>\*‡Moreover, should it turn out that there is no phenomenological need for a non-zero $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ after all, even then we would still have to explain why $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ is not as big as its theoretical expectation, with the disappearance of the need not entailing the disappearance of the problem. In all then we thus identify an uncomfortably large number of problems for current cosmology (even if explicitly less than one, an $`\mathrm{\Omega }_M(t_0)`$ of order one would still require an enormous amount of as yet totally undetected non-luminous, expressly non-baryonic, cosmological dark matter), and see that they all appear to have one common ingredient, namely the use of the evolution equation of Eq. (22) in the first place. Thus it is highly suggestive that the problems that the standard cosmology currently faces might all derive from the lack of reliability of the extrapolation of standard gravity wisdom beyond its solar system origins. In the following then we shall thus relax this assumption, and in particular we shall show that all of the above problems can readily be resolved if gravity acquires one further form of cosmic repulsion, namely that due to an effective cosmological Newton constant which is expressly taken to be negative.
In trying to identify the root cause of the above problems we note that the macroscopic Friedmann equation fine-tuning problem arises because of mismatch between the early and current universes, while the microscopic cosmological constant problem arises because of a clash between elementary particle physics and gravitational physics, with particle physics wanting a large $`\mathrm{\Lambda }`$ and gravity a small one. Since this clash is between different branches of physics, we should not immediately assume that it is the particle physics which is at fault. Rather, the indications of particle physics might well be correct, with its contribution to $`\mathrm{\Lambda }`$ actually being as big as it would appear to be. Indeed, the very failure to date of attempts to quench the particle physics $`\mathrm{\Lambda }`$ from so large an expected value might even be an indicator that it is not in fact quenched, with it being reasonable to then ask what the implications for cosmology are if $`\mathrm{\Lambda }`$ really is big, and whether cosmology could actually accommodate a large $`\mathrm{\Lambda }`$. To this end we note that the quantity which is of relevance to cosmological evolution is not in fact $`\mathrm{\Lambda }`$ itself but rather $`\mathrm{\Omega }_\mathrm{\Lambda }(t)=8\pi G\mathrm{\Lambda }/3cH^2(t)`$, i.e. not the energy of the vacuum itself, but rather, its contribution to gravitational evolution, with only this latter quantity being observable. With this latter contribution depending not just on $`\mathrm{\Lambda }`$ but also on $`G`$, we see that a quenching of $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ could potentially be achieved not by quenching $`\mathrm{\Lambda }`$ but by quenching $`G`$ instead. Thus again we are led to consider $`G`$ as being only an effective parameter, one whose cosmological coupling might be altogether smaller than that relevant to the solar system. We shall thus explore the possibility that the effective cosmological $`G`$ is both small and negative, first as a general effect and then in an explicit solvable model.
As regards explicitly trying to find a solution to the Friedmann universe fine-tuning problem, we note that since the standard cosmology has a big bang, the early universe $`\dot{R}(t)`$ would have to be divergent at $`t=0`$ (or at least be very large), with Eq. (22) then requiring the quantity $`\mathrm{\Omega }_M(t=0)+\mathrm{\Omega }_\mathrm{\Lambda }(t=0)`$ to be equal to one at $`t=0`$, regardless of what particular value the spatial curvature $`k`$ might take. Then, given the radically different temporal behaviors of $`\mathrm{\Omega }_M(t)`$, $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ and $`\mathrm{\Omega }_k(t)`$, we see directly that no cosmology which obeys this initial constraint, be it flat or non-flat, could ever evolve into one in which $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)\mathrm{\Omega }_M(t_0)O(1)`$ today (or into one in which $`\mathrm{\Omega }_M(t_0)O(1)`$ should $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ just happen to be zero) without extreme fine tuning. Hence it is the big bang itself which is creating the fine-tuning problem, with it being very difficult for a Friedmann universe to evolve from a singular early state into the one currently observed. In order to eliminate such an incompatibility we are thus led to consider removing the big bang singularity from cosmology altogether, and have the universe expand from some initial (but still very hot) state characterized by $`\dot{R}(t=0)=0`$ instead.Once $`\dot{R}(t=0)`$ is non-singular, the natural definition of the initial time in a universe which expands is then the one where $`R(t)`$ is at a minimum, with its minimum value not needing to be zero in a non-singular cosmology. Since the big bang singularity itself derives from the assumption that gravity continues to be attractive even when it is strong, we are thus led to consider the possibility that the effective cosmological $`G`$ actually be repulsive, with cosmology then being able to protect itself from its own singularities and potentially rid itself of fine-tuning problems.That the origin of the flatness problem could be traced to the positivity of $`G`$ was already noted quite some time ago in , where it was also pointed out that a negative effective $`G`$ repulsive cosmology would actually have no fine-tuning flatness problem at all, with quantities which would have had to cancel to very high accuracy in Eq. (22) no longer having to do so.
Continuing in this same vein, we note further that if the universe turns out to ultimately be an accelerating one (either even already, or at some time later in the future<sup>\*∥</sup><sup>\*∥</sup>\*∥Even if $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ turns out to be negligible today, as long as it is greater than zero, then no matter by how little, there will eventually come a time when $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ will ultimately come to dominate the expansion of the universe.), $`\dot{R}(t)`$ will then become arbitrarily large in the late (rather than the early) universe, with the quantity $`\mathrm{\Omega }_M(t)+\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ tending to one at late times, again independent of the value of $`k`$. Then, in such a late universe it would be $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ which would have to tend to one, no matter what may or may not have happened in the early universe, and regardless of how big $`\mathrm{\Lambda }`$ itself might actually be. Thus at very late times the cosmological constant problem would not only get solved, it would get solved by cosmology itself, with an accelerating universe always being able to quench $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ once given time enough to do so. Thus we see that the very same problem which is generated by the very existence of a cosmological constant is then made solvable by the very cosmic acceleration which it simultaneously produces; with the key question for cosmology thus being whether the universe is already sufficiently late for this quenching to have already taken place. However, in order for this to actually be the case, it is necessary that the current era contribution of ordinary matter to the evolution of the universe be cosmologically insignificant (i.e. that $`\mathrm{\Omega }_M(t)=8\pi G\rho _M(t)/3c^2H^2(t)`$ already be close to an asymptotically expected value of zero). Thus again we find ourselves led to considering the possibility that the effective cosmological $`G`$ be altogether smaller than the one relevant in the solar system, so that, no matter how big $`\rho _M(t)`$ itself might be, ordinary matter would then no longer be of relevance to the current expansion, though, given its temporal behavior, $`\rho _M(t)`$ would still be highly significant at altogether earlier times.<sup>\***</sup><sup>\***</sup>\***Thus we do not seek to change the matter content of the universe, but rather only to modify how it impacts on cosmological evolution. Moreover, with the ratio $`\mathrm{\Omega }_\mathrm{\Lambda }(t)/\mathrm{\Omega }_M(t)=T_V^4/T^4(t)`$ actually being independent of $`G`$, we see that the requirement that $`T_V/T(t_0)`$ be large and that the current era $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ be of order one together entail that $`\mathrm{\Omega }_M(t_0)`$ be highly suppressed, something readily achievable if the effective cosmological $`G`$ is then very small. Thus to conclude, we see that many outstanding cosmological puzzles are readily addressable if gravity were indeed to be describable cosmologically by a repulsive and very small effective $`G`$, and thus motivated we shall, in the following, present a specific alternate gravitational theory, actually the conformal gravity theory to which we referred above, where this is precisely found to be the case.
## V Repulsive Gravity
Since we are not currently aware of how it might be possible to implement the above general ideas within standard gravity (a $`G`$ which evolves with temperature from the early universe until today is certainly conceivable within standard gravity though perhaps not one which might also change sign as it evolves), we shall instead turn to an alternate gravity theory, viz. conformal invariant gravity, a theory which is immediately suggested since gravity then becomes a theory with dimensionless coupling constants and no intrinsic mass scales just like the three other fundamental interactions. Thus gravity becomes a theory which is power counting renormalizable, with the absence of any intrinsic mass scale immediately obliging any fundamental cosmological constant term to be set to zero in it,<sup>\*††</sup><sup>\*††</sup>\*††The absence of any intrinsic quantum gravity scale not merely eliminates the Planck length from quantum relevance, it also provides for a theory of gravity in which quantum gravity fluctuations themselves cannot then generate a cosmological constant term at all; with any induced microscopic $`\mathrm{\Lambda }`$ then only being generatable by elementary particle physics phase transitions. to thus yield a theory in which the cosmological constant term is actually controlled by an underlying symmetry, an objective long sought in the standard model. With the conformal theory also, as noted earlier, possessing a good Newtonian limit despite the absence of any intrinsic Einstein-Hilbert term (the conformal gravitational action is uniquely given by the conformal $`I_W`$ action of Eq. (3)), we see that a possible reason why the standard gravity cosmological constant problem has resisted solution for so long could be that was always being sought was a symmetry which would eliminate a fundamental $`\mathrm{\Lambda }`$ but not $`I_{EH}`$ rather than one which would in fact eliminate both. Since solar system information had never in fact required the presence of $`I_{EH}`$ in the first place, we see that its removal actually opens up a new line of attack on the cosmological constant problem; and as we shall now see, the conformal symmetry not merely controls any fundamental cosmological constant term, even after the conformal symmetry is spontaneously broken (something needed to generate particle masses in the conformal symmetry), the continuing tracelessness of the energy-momentum tensor sharply constrains the magnitude of any induced one in a manner which is then found to automatically implement the $`\mathrm{\Omega }_\mathrm{\Lambda }(t)`$ quenching mechanism we introduced above.
The cosmology associated with conformal gravity was first presented in where it was shown, well in advance of the recent discovery of cosmic repulsion, to be one with an effective repulsive cosmological $`G`$ just as desired above. To discuss conformal cosmology it is convenient to consider the conformal matter action
$$I_M=\mathrm{}d^4x(g)^{1/2}[S^\mu S_\mu /2S^2R_\mu ^\mu /12+\lambda S^4+i\overline{\psi }\gamma ^\mu (x)(_\mu +\mathrm{\Gamma }_\mu (x))\psi gS\overline{\psi }\psi ]$$
(24)
where we take the elementary particle physics matter fields to be generically represented by massless fermions for definitiveness and simplicity, with a conformally coupled massless scalar field being introduced to serve as the order parameter associated with spontaneous breakdown of the scale symmetry. With dynamical symmetry breaking being needed in order to generate elementary particle masses in the scaleless interacting massless fermion and gauge boson theories now standard in particle physics, the scalar field $`S(x)`$ introduced here should be thought of not as a fundamental scalar field but as the expectation value of an appropriate fermion multilinear in an appropriate coherent fermionic state. $`S(x)`$ is thus to serve as a cosmological analog of the Cooper pair of superconductivity theory, with the action of Eq. (24) serving as an analog of the Ginzburg-Landau phenomenological superconductivity Lagrangian. In such a case the vacuum energy would be zero above the critical temperature where the order parameter $`S(x)`$ would then vanish, and would be expressly negative below it. Simulating the vacuum energy below the critical point by the effective $`\mathrm{}\lambda S^4(x)\mathrm{\Lambda }`$ term, then entails (as noted in ) that in a theory which is scale invariant above the critical point, the effective $`\mathrm{\Lambda }`$ which is induced below the critical point is then expressly negative. Thus unlike the situation in the standard theory, in the conformal case the sign of the induced cosmological constant term is explicitly determined.
For the above matter action the matter field equations of motion take the form
$`i\gamma ^\mu (x)[_\mu +\mathrm{\Gamma }_\mu (x)]\psi gS\psi =0`$ (25)
$`S_{;\mu }^\mu +SR_\mu ^\mu /64\lambda S^3+g\overline{\psi }\psi =0`$ (26)
with the matter energy-momentum tensor being given by
$`T^{\mu \nu }=\mathrm{}\{i\overline{\psi }\gamma ^\mu (x)[^\nu +\mathrm{\Gamma }^\nu (x)]\psi +2S^\mu S^\nu /3g^{\mu \nu }S^\alpha S_\alpha /6SS^{\mu ;\nu }/3`$ (27)
$`+g^{\mu \nu }SS_{;\alpha }^\alpha /3S^2(R^{\mu \nu }g^{\mu \nu }R_\alpha ^\alpha /2)/6g^{\mu \nu }\lambda S^4\}.`$ (28)
Thus, when the scalar field acquires a non-zero vacuum expectation value $`S_0`$, the energy-momentum tensor then takes the form (for a perfect matter fluid $`T_{kin}^{\mu \nu }`$ of the fermions)
$$T^{\mu \nu }=T_{kin}^{\mu \nu }\mathrm{}S_0^2(R^{\mu \nu }g^{\mu \nu }R_\alpha ^\alpha /2)/6g^{\mu \nu }\mathrm{}\lambda S_0^4.$$
(29)
Since the Weyl tensor $`C^{\lambda \mu \nu \kappa }`$ and the quantity $`W^{\mu \nu }`$ of Eq. (6) both vanish identically in the highly symmetric Robertson-Walker geometry, the complete cosmological solution to the joint scalar, fermionic, and gravitational field equations of motion then reduces to just one relevant equation, viz.
$$T^{\mu \nu }=0,$$
(30)
a remarkably simple condition which immediately fixes the zero of energy. Thus in its spontaneously broken phase conformal cosmology is described by the equation
$$\mathrm{}S_0^2(R^{\mu \nu }g^{\mu \nu }R_\alpha ^\alpha /2)/6=T_{kin}^{\mu \nu }g^{\mu \nu }\mathrm{}\lambda S_0^4,$$
(31)
an equation which we recognize as being none other than that of standard gravity save only that the quantity $`\mathrm{}S_0^2/12`$ has replaced the familiar $`c^3/16\pi G`$. Conformal cosmology thus acts exactly like a standard gravity theory in which $`G`$ is effectively negative, with its magnitude actually becoming smaller the larger $`S_0`$ gets to be, i.e. the same particle physics mechanism which makes the cosmological constant large serves to make the effective cosmological $`G`$ small. Thus, just as desired, conformal cosmology is controlled by an effective $`G`$ which is both negative and small. Noting further that the Weyl tensor vanishes in high symmetry, cosmologically relevant geometries such as the homogeneous Robertson-Walker one, but not in low symmetry ones such as Schwarzschild (viz. geometries which are generated by the presence of local spatial inhomogeneities in an otherwise homogeneous cosmological background), we thus see that the gravitational coupling constant $`\alpha _g`$, viz. the one which according to Eq. (9) explicitly controls the geometry outside of a local static source, simply decouples from cosmology. Thus in conformal gravity (inhomogeneous) locally attractive and (homogeneous) globally repulsive gravity can readily coexist, with solar system physics indeed not then being a good guide to the behavior of gravity in altogether different circumstances.
Given Eq. (31) the conformal cosmology evolution equations immediately take the form
$$\dot{R}^2(t)+kc^2=3\dot{R}^2(t)(\mathrm{\Omega }_M(t)+\mathrm{\Omega }_\mathrm{\Lambda }(t))/4\pi S_0^2L_{PL}^2\dot{R}^2(t)(\overline{\mathrm{\Omega }}_M(t)+\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t))$$
(32)
and
$$q(t)=(n/21)\overline{\mathrm{\Omega }}_M(t)\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)$$
(33)
(Eq. (32) serves to define $`\overline{\mathrm{\Omega }}_M(t)`$ and $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)`$), and are thus remarkably similar to the standard evolution equations of Eqs. (22) and (23); with our entire earlier general discussion on the need to replace the cosmological $`G`$ by a small negative effective one now finding explicit realization in a specific alternate gravitational theory. With both the effective $`G`$ and the vacuum breaking $`\mathrm{\Lambda }`$ being negative in conformal gravity, we thus see that $`\overline{\mathrm{\Omega }}_M(t)`$ is necessarily negative while $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)`$ is positive. Consequently, the ordinary matter energy density and the vacuum energy density both naturally lead to cosmological repulsion, with the conformal $`q(t)`$ always having to be less than or equal to zero. Conformal cosmologies thus never decelerate with each epoch seeing some measure of cosmic repulsion no matter what the explicit magnitudes of $`\overline{\mathrm{\Omega }}_M(t)`$ and $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)`$ might be.
In order to determine how much repulsion there might be in any given epoch it is necessary to determine the value of the spatial curvature $`k`$. To this end we note while we cannot immediately fix $`k`$ from study of the broken symmetry phase itself ($`\overline{\mathrm{\Omega }}_M(t)`$ and $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)`$ make contributions of opposite sign in Eq. (32)), it is possible to extract information about $`k`$ from the high temperature phase above all phase transitions, a phase where the order parameter vanishes, a phase which can be modeled entirely by the presence of just a perfect fluid, viz. one in which the entire $`T^{\mu \nu }`$ is given by $`T_{kin}^{\mu \nu }`$. In such a high temperature Robertson-Walker phase the gravitational equations of motion of Eq. (4) reduce to the condition
$$T_{kin}^{\mu \nu }=0,$$
(34)
an equation which quite remarkably actually has a non-trivial solution in curved space.
To explore such a solution, we note that while we have generically identified the fields in $`T_{kin}^{\mu \nu }`$ to be perfect fluid fermions, for calculational purposes it is more convenient to consider them to be non-interacting scalar fields instead. We thus populate the non-spontaneously broken high temperature universe by a perfect fluid built out of the modes of a normal (i.e. not spontaneously broken) scalar field with equation of motion
$$S_{;\mu }^\mu +SR_\mu ^\mu /6=0$$
(35)
and energy-momentum tensor
$$T^{\mu \nu }=2S^\mu S^\nu /3g^{\mu \nu }S^\alpha S_\alpha /6SS^{\mu ;\nu }/3+g^{\mu \nu }SS_{;\alpha }^\alpha /3S^2(R^{\mu \nu }g^{\mu \nu }R_\alpha ^\alpha /2)/6.$$
(36)
With the Ricci scalar being given by the spatially independent $`R_\alpha ^\alpha =6(k+R(t)\ddot{R}(t)+\dot{R}^2(t))/R^2(t)`$, the scalar field equation of motion is found to be separable. Thus on introducing the conformal time $`p=c^t𝑑t/R(t)`$ and on setting $`S(x)=f(p)g(r,\theta ,\varphi )/R(t)`$, Eq. (35) is found to reduce to ($`\gamma `$ denotes the determinant of the spatial metric)
$$\frac{1}{f(p)}\left[\frac{d^2f}{dp^2}+kf(p)\right]=\frac{1}{g(r,\theta ,\varphi )}\gamma ^{1/2}_i[\gamma ^{1/2}\gamma ^{ij}_jg(r,\theta ,\varphi )]=\lambda ^2,$$
(37)
where we have introduced a separation constant $`\lambda ^2`$. With the dependence on $`f(p)`$ on $`p`$ being harmonic, the frequencies thus have to satisfy $`\omega ^2=\lambda ^2+k`$, a relation we actually presented earlier. For the spatial dependence we can further set $`g(r,\theta ,\varphi )=g_\lambda ^{\mathrm{}}(r)Y_{\mathrm{}}^m(\theta ,\varphi )`$ where $`g_\lambda ^{\mathrm{}}(r)`$ obeys the radial equation
$$\left[(1kr^2)\frac{^2}{r^2}+\frac{(23kr^2)}{r}\frac{}{r}\frac{\mathrm{}(\mathrm{}+1)}{r^2}+\lambda ^2\right]g_\lambda ^{\mathrm{}}(r)=0,$$
(38)
with the radial solutions being given by $`j_{\mathrm{}}(\omega r)/\omega `$ Bessel functions when $`k=0`$, by associated Legendre functions when $`k=1`$, viz.
$$g_\lambda ^{\mathrm{}}(r)=[\pi \omega ^2(\omega ^2+1^2)\mathrm{}(\omega ^2+\mathrm{}^2)/2]^{1/2}sinh^{\mathrm{}}\chi \left[\frac{d}{dcosh\chi }\right]^{\mathrm{}+1}\frac{cos\omega \chi }{\omega }$$
(39)
where $`r=sinh\chi `$, and by Gegenbauer polynomials when $`k=1`$, viz.
$$g_\lambda ^{\mathrm{}}(r)=[\pi \omega ^2(\omega ^21^2)\mathrm{}(\omega ^2\mathrm{}^2)/2]^{1/2}sin^{\mathrm{}}\chi \left[\frac{d}{dcos\chi }\right]^{\mathrm{}+1}\frac{cos\omega \chi }{\omega }$$
(40)
where $`r=sin\chi `$. With these normalizations (which differ slightly from those given in ), an incoherent averaging of the energy-momentum tensor of Eq. (36) over all the available spatial states associated with a given frequency $`\omega `$ is then found to lead directly for every allowed $`\omega `$ to the traceless kinematic perfect fluid
$$T_{kin}^{\mu \nu }=\frac{\omega ^2(4U^\mu U^\nu +g^{\mu \nu })}{6\pi ^2R^4(t)}$$
(41)
in all three of the spatial geometries.<sup>\*‡‡</sup><sup>\*‡‡</sup>\*‡‡On purely general geometric grounds the most general rank two tensor in a Robertson-Walker geometry is writable as $`T^{\mu \nu }=(A(t)+B(t))U^\mu U^\nu +B(t)g^{\mu \nu }`$, with the coefficients $`A(t)`$ and $`B(t)`$ being otherwise unconstrained, so that their ratio $`w(t)=B(t)/A(t)`$ need not be time independent in general. However, for a tensor which is both traceless and covariantly conserved, it further follows that $`3B(t)=A(t)=C/R^4(t)`$ where $`C`$ is a pure constant; with the calculation leading to Eq. (41) thus being a calculation of the value of this constant. While on this point, it is worth noting in passing that it is not automatically the case that $`w(t)`$ is necessarily time independent or that $`A(t)`$ and $`B(t)`$ necessarily have the same dependence on time in the general Robertson-Walker cosmology case. Relations between $`A(t)`$ and $`B(t)`$ only follow under specific dynamical assumptions. If, for example, $`A(t)`$ and $`B(t)`$ contain two or more separate components, then even if the separate components are related via time independent $`w_1=B_1/A_1`$, $`w_2=B_2/A_2`$, it does not necessarily follow that $`B_1+B_2`$ is proportional to $`A_1+A_2`$. Further, even for the restricted case of the kinematic $`T_{kin}^{\mu \nu }`$ in which $`A(t)`$ and $`B(t)`$ are both associated with just one single completely standard perfect fluid, it turns out that even in that case they are still not in fact proportional to each other at all temperatures. To illustrate this point, consider an ideal $`N`$ particle classical gas of particles of mass $`m`$ in a volume $`V`$ at a temperature $`T`$. For this system the Helmholtz free energy $`A(V,T)`$ is given as exp$`[A(V,T)/NkT]=Vd^3p`$ exp$`[(p^2+m^2)^{1/2}/kT]`$, so that the pressure takes the simple form $`P=(A/V)_T=NkT/V`$, while the internal energy $`U=AT(A/T)_V`$ evaluates in terms of Bessel functions as $`U=3NkT+NmK_1(m/kT)/K_2(m/kT)`$. In the two limits $`m/kT0`$, $`m/kT\mathrm{}`$ we then find that $`U3NkT`$, $`UNm+3NkT/2`$. Thus only at these two extreme temperature limits does it follow that the energy density and the pressure are in fact proportional, with their relation in intermediate regimes such as the transition region from the radiation to the matter era being far more complicated. With the condition $`T_{kin}^{\mu \nu }=0`$ of Eq. (34) then only permitting the soft $`\omega =0`$ modes, and with only the $`k=1`$ radial equation admitting of a non-trivial radial solution in such a case (viz. the entire infinite set<sup>†\*</sup><sup>†\*</sup>†\*In passing we note that it might prove interesting should there be a group under which this infinite tower of states transforms irreducibly. of all $`\mathrm{}`$ modes built on the $`\mathrm{}=0`$ solution $`g_1^0(r)\chi /sinh\chi `$), we see that it is possible to satisfy the condition $`T_{kin}^{\mu \nu }=0`$ non-trivially in the negative spatial curvature conformal cosmology case, with the very high temperature universe then being composed of a perfect fluid bath of soft modes which non-trivially support a $`k<0`$ universe. Cosmology thus fixes the curvature itself, and does so before the onset of any symmetry breaking phase transition at all.<sup>††</sup><sup>††</sup>††To show that there actually will be a phase transition as the temperature drops requires the development of a detailed dynamical model for the coherent correlations that are not present in the incoherently averaged perfect fluid limit. However, without knowing the details of how long range order actually sets in, nonetheless, use of the perfect fluid model at temperatures far above the critical region is sufficient to fix the sign of $`k`$ once and for all.
It is also illuminating to derive this result in a slightly different fashion. Since $`T_{kin}^{\mu \nu }`$ is a conformal invariant tensor, we could also evaluate it by first making the conformal transformation $`g_{\mu \nu }(x)R^2(t)g_{\mu \nu }(x)`$ on the Robertson-Walker metric, a transformation which brings the geometric line element to the form $`d\tau ^2=dp^2dr^2/(1kr^2)r^2d\mathrm{\Omega }=dp^2\gamma _{ij}dx^idx^j`$. For this transformed metric the Ricci scalar is given by $`R_\alpha ^\alpha =6k`$, with the incoherently averaged soft mode contribution to the energy density then being given by
$$T_{kin}^{00}=\frac{1}{6}\underset{\mathrm{},m}{}\left[\underset{i=1}{\overset{3}{}}\gamma ^{ii}|_i(g_1^{\mathrm{}}Y_{\mathrm{}}^m(\theta ,\varphi ))|^2+k|g_1^{\mathrm{}}Y_{\mathrm{}}^m(\theta ,\varphi )|^2\right],$$
(42)
a quantity which can actually vanish non-trivially provided $`k`$ is negative. And indeed, through use of the various completeness relations which these modes obey, explicit evaluation of $`T_{kin}^{00}`$ is then found to yield
$`T_{kin}^{00}={\displaystyle \frac{1}{6}}{\displaystyle \underset{\mathrm{},m}{}}{\displaystyle \frac{(2\mathrm{}+1)(\mathrm{}m)!}{4\pi (\mathrm{}+m)!}}[(1kr^2)P_{\mathrm{}}^m(cos\theta )^2({\displaystyle \frac{dg_1^{\mathrm{}}}{dr}})^2+{\displaystyle \frac{1}{r^2}}({\displaystyle \frac{dP_{\mathrm{}}^m(cos\theta )}{d\theta }})^2(g_1^{\mathrm{}})^2`$ (43)
$`+{\displaystyle \frac{m^2}{r^2sin^2\theta }}P_{\mathrm{}}^m(cos\theta )^2(g_1^{\mathrm{}})^2+kP_{\mathrm{}}^m(cos\theta )^2(g_1^{\mathrm{}})^2]`$ (44)
$`={\displaystyle \frac{1}{24\pi }}{\displaystyle \underset{\mathrm{}}{}}(2\mathrm{}+1)\left[(1kr^2)({\displaystyle \frac{dg_1^{\mathrm{}}}{dr}})^2+{\displaystyle \frac{\mathrm{}(\mathrm{}+1)}{r^2}}(g_1^{\mathrm{}})^2+k(g_1^{\mathrm{}})^2\right]`$ (45)
$`={\displaystyle \frac{1}{24\pi }}\left[{\displaystyle \frac{2k}{3\pi }}{\displaystyle \frac{4k}{3\pi }}+{\displaystyle \frac{2k}{\pi }}\right]=0.`$ (46)
Thus we see that when the spatial curvature is negative, the negative energy density then present in the gravitational field completely cancels the positive energy density of the matter fields, with gravity itself then being able to fix the spatial curvature of the universe.
Having now shown that $`k`$ is uniquely negative in conformal cosmology (we will show below how galactic rotation curve data actually enable us to measure $`k`$ to find that it is indeed negative), we can now proceed to study its cosmological implications as the universe cools. Thus, with the signs of $`k`$ and $`\lambda `$ now being fixed (we simulate the negativity of $`\mathrm{\Lambda }=\mathrm{}\lambda S_0^4`$ by a negative $`\lambda `$), Eq. (32) is then found (in the simpler to treat high temperature era where $`cT_{kin}^{00}=\rho _M(t)=A/R^4=\sigma T^4`$) to admit of the unique solution
$$R^2(t,\alpha >0,k<0)=k(\beta 1)/2\alpha k\beta sinh^2(\alpha ^{1/2}ct)/\alpha $$
(47)
where have introduced the positive parameter $`\alpha =2\lambda S_0^2`$ and the parameter $`\beta =(116A\lambda /k^2\mathrm{}c)^{1/2}`$ which is greater than one. In this solution the deceleration parameter is found to take the requisite non-positive form
$$q(t,\alpha >0,k<0)=tanh^2(\alpha ^{1/2}ct)+2(1\beta )cosh(2\alpha ^{1/2}ct)/\beta sinh^2(2\alpha ^{1/2}ct).$$
(48)
We thus see that the cosmology is non-singular, having a finite minimum radius and an initial $`\dot{R}(t=0)`$ which is zero rather than infinite. Since the cosmology has a minimum radius, it also has a finite maximum temperature $`T_{max}`$ in terms of which Eq. (47) can be rewritten as
$$T_{max}^2(\alpha >0,k<0)/T^2(t,\alpha >0,k<0)=1+2\beta sinh^2(\alpha ^{1/2}ct)/(\beta 1),$$
(49)
with the Hubble parameter then being given as
$$H(t)=\alpha ^{1/2}c(1T^2(t)/T_{max}^2)/tanh(\alpha ^{1/2}ct).$$
(50)
In terms of the convenient effective temperature $`T_V`$ defined via $`c\mathrm{\Lambda }=c\mathrm{}\lambda S_0^4=\sigma T_V^4`$, we find that the parameter $`\beta `$ can be expressed as
$$\beta =(1+T_V^4/T_{max}^4)/(1T_V^4/T_{max}^4)$$
(51)
(with $`T_V`$ thus being less than the maximum temperature $`T_{max}`$), with the temporal evolution of the theory then being given by
$$T_{max}^2/T^2(t)=1+(1+T_{max}^4/T_V^4)sinh^2(\alpha ^{1/2}ct)$$
(52)
and with the energy density terms then being given by
$`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t)=(1T^2(t)/T_{max}^2)^1(1+T^2(t)T_{max}^2/T_V^4)^1,`$ (53)
$`\overline{\mathrm{\Omega }}_M(t)=(T^4(t)/T_V^4)\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t),`$ (54)
$`\mathrm{\Omega }_k(t)=kc^2/\dot{R}^2(t)=1\overline{\mathrm{\Omega }}_M(t)\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t).`$ (55)
With $`(1+T^2(t)T_{max}^2/T_V^4)^1`$ being a quantity which is always necessarily bounded between zero and one (no matter what the magnitude of $`T_V`$), we see that the single, simple requirement that $`T_{max}`$ be very much greater than $`T(t_0)`$ then entails that $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ must lie somewhere between zero and an upper bound of one today, with its current value then being given by $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)=(1+T^2(t_0)T_{max}^2/T_V^4)^1=tanh^2(\alpha ^{1/2}ct_0)`$ according to Eq. (52). Thus, no matter how big $`T_V`$ might be, $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ must not only be of order one today, it must also be approaching its asymptotically expected value of one from below; to thus yield a completely natural current era quenching of $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ regardless of the numerical value of any cosmological parameter. Moreover, the larger $`T_V`$, the further $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ must lie below its asymptotic bound of one, with it taking a typical value of one-half in the event that the current temperature is given by $`T(t_0)T_V^2/T_{max}`$, a condition which is readily realizable if the conditions $`T_{max}T_V`$ and $`T_VT(t_0)`$ both hold.<sup>†‡</sup><sup>†‡</sup>†‡There are three possible ways in which the condition $`T_{max}T(t_0)`$ can be realized in Eq. (52). If $`T_V`$ is of order $`T(t_0)`$, $`sinh^2(\alpha ^{1/2}ct_0)`$ would be very much less than one. Similarly, if $`T_V`$ is of order $`T_{max}`$, $`sinh^2(\alpha ^{1/2}ct_0)`$ would be very much greater than one. However, if $`T_V`$ is of intermediate order $`(T(t_0)T_{max})^{1/2}`$, $`sinh^2(\alpha ^{1/2}ct_0)`$ would then be of order one. Thus we see that $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ could actually be appreciably below one today, and that it would actually fall further below one the larger rather than the smaller the cosmological constant, with conformal cosmology thus having no difficulty handling a large $`T_V/T(t_0)`$. Thus it is precisely in the event that the cosmological constant is in fact large that $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ is then able to be of a phenomenologically acceptable magnitude long before becoming fully asymptotic.<sup>†§</sup><sup>†§</sup>†§Since $`T_{max}`$ is greater than $`T_V`$, a large $`T_V/T(t_0)`$ entails a large $`T_{max}/T(t_0)`$, and thus a universe old enough for the current era $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ to be given by the bounded $`tanh^2(\alpha ^{1/2}ct_0)`$. For large $`T_V`$ conformal cosmology thus solves the cosmological constant problem simply by living a long time. Beyond this, we note additionally, that given the fact that there is an upper bound on $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$, we see that a large $`T_V/T(t_0)`$ will then completely quench the current era $`\overline{\mathrm{\Omega }}_M(t_0)`$ altogether. Thus in conformal gravity a large cosmological constant naturally leads to current era suppression of $`\overline{\mathrm{\Omega }}_M(t_0)`$ just as was desired above.<sup>†¶</sup><sup>†¶</sup>†¶Phenomenologically, this quenching of $`\overline{\mathrm{\Omega }}_M(t_0)=3\mathrm{\Omega }_M(t_0)/4\pi S_0^2L_{Pl}^2`$ requires that the scale factor $`S_0`$ be altogether greater than $`L_{Pl}^1`$, a condition which is readily realizable not only by associating a fundamental temperature with a fundamental $`S_0`$ which is altogether larger than the Planck temperature, but instead, and preferentially, by identifying $`S_0`$ as a macroscopically occupied order parameter , with $`S_0`$ then being proportional to the (very large) number, $`N`$, of occupied positive energy perfect fluid modes, modes whose coherent correlations cause phase transitions to occur in the early universe in the first place. Then, because of this suppression, the current era evolution equation of Eq. (32) reduces to $`1=\mathrm{\Omega }_k(t_0)+\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$, with negative curvature explicitly forcing $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ to lie below one. It is thus the negative spatial curvature of the universe which bounds and tames the contribution of the cosmological constant to cosmology,<sup>†∥</sup><sup>†∥</sup>†∥In a $`\lambda <0`$ cosmology with $`k>0`$, the current era $`\overline{\mathrm{\Omega }}_M(t_0)`$ is still found to be completely suppressed by large $`T_V/T(t_0)`$, but in such a cosmology $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ is no longer bounded by one from above, but only from below, with the $`k>0`$ case $`T_{max}`$ being found to be less than $`T_V`$. Thus we see that in the $`k<0`$ case it is precisely the contribution of negative curvature itself which produces a cosmology in which the highest temperature $`T_{max}`$ is greater than the vacuum energy temperature $`T_V`$ (with there thus being an in principle difference between flat space and curved space phase transitions), with curvature itself controlling the value of $`T_{max}`$. Moreover, with there being an explicit maximum temperature $`T_{max}^2=k\mathrm{}S_0^2c/2(\sigma A)^{1/2}`$ in a $`k<0`$ cosmology even in the absence of any $`\lambda S_0^4`$ term at all, we see that the magnitudes of $`T_{max}`$ and $`T_V`$ are fixable independently, with the imposition of the condition $`T_{max}T_V`$ which we use thus being readily naturally achievable in the $`\lambda 0`$ case without the need for any fine-tuning of parameters. with it thus indeed being possible to construct a phenomenologically acceptable cosmology in which $`\mathrm{\Lambda }`$ can still be as large as particle physics suggests.
From a phenomenological viewpoint, once $`\overline{\mathrm{\Omega }}_M(t_0)`$ is suppressed, the deceleration parameter is then given by $`q(t_0)=tanh^2(\alpha ^{1/2}ct_0)`$ (so that it then has to lie between zero and minus one), while the curvature contribution is given by $`\mathrm{\Omega }_k(t_0)=sech^2(\alpha ^{1/2}ct_0)`$. Thus in conformal cosmology first matter dominates the expansion rate (in the early universe), then curvature, and finally vacuum energy; and even if the current era is not yet vacuum energy dominated, nonetheless $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)`$ will still be under control no matter what the value of $`\alpha ^{1/2}ct_0`$. As regards the actual supernovae data themselves, we note that even though the phenomenological fitting allowed for solutions with $`\mathrm{\Omega }_M(t_0)=0`$ (typically with $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)=q(t_0)=1/2`$), and even though such solutions would not be expected to occur in the standard theory, we see that in the conformal theory<sup>†\**</sup><sup>†\**</sup>†\**With $`\mathrm{\Omega }_M(t_0)`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)`$ being treated as free parameters in supernovae data fitting based on the standard Eq. (22), those fits are just as equally phenomenological fits to the conformal Eq. (32). solutions with $`\overline{\mathrm{\Omega }}_M(t_0)=0`$ and $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)=1/2`$ are right in the region allowed by Eq. (55). Moreover, in fits with $`\mathrm{\Omega }_M(t_0)=0`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)=0`$ were found, and even though those fits were of quality comparable with the best reported $`\mathrm{\Omega }_M(t_0)0`$, $`\mathrm{\Omega }_\mathrm{\Lambda }(t_0)0`$ fits, such fits were not considered further since in standard gravity they would correspond to an empty universe. However, we now see that in conformal gravity not only would such fits (fits in which $`T_V^2/T(t_0)T_{max}1`$, so that $`\overline{\mathrm{\Omega }}_M(t_0)0`$ $`\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)0`$) be quite acceptable, such fits would not entail an empty universe since it is $`\overline{\mathrm{\Omega }}_M(t_0)=0`$ which is suppressed and not $`\rho _M(t_0)`$ itself. Thus with no fine-tuning of parameters at all conformal cosmology leads us right into the $`\overline{\mathrm{\Omega }}_M(t_0)=0`$, $`0\overline{\mathrm{\Omega }}_\mathrm{\Lambda }(t_0)1`$ region favored by the supernovae data. Moreover, with the suppression of $`\overline{\mathrm{\Omega }}_M(t_0)=0`$ being achievable without any constraint being put on $`\rho _M(t_0)`$, its value is thus not constrained to be of order the critical density $`\rho _c=3c^2H^2(t_0)/8\pi G`$, with conformal cosmology thus being released from the need to contain copious amounts of cosmological dark matter.<sup>†††</sup><sup>†††</sup>†††Like $`\overline{\mathrm{\Omega }}_M(t)`$, in the conformal case the quantity $`\mathrm{\Omega }_M(t)`$ itself starts off being infinite at $`t=0`$ (no matter what the numerical values of the parameters) and finishes up being zero at $`t=\mathrm{}`$. Thus, $`\mathrm{\Omega }_M(t)`$ has to pass through one in some epoch without the need for any fine tuning (and even has to pass through one quite slowly since the cosmology associated with $`R(t)`$ of Eq. (47) is a very slow coasting one) while being far from one in other epochs, to thus not constrain the value of $`\rho _M(t_0)`$ at all. Moreover, if dark matter is not invoked, known explicitly visibly established matter alone leads to an $`\mathrm{\Omega }_M(t_0)`$ of order $`10^2`$ or so, so $`\mathrm{\Omega }_M(t_0)`$ may not be so close to one today as to require any particular explanation in the first place (i.e. the value of $`\mathrm{\Omega }_M(t_0)`$ would not be special unless it was actually equal to one to some incredibly high degree of accuracy); and indeed, even in an $`\mathrm{\Omega }_M(t_0)<1`$ standard theory, ongoing expansion will eventually lead to an $`\mathrm{\Omega }_M(t)`$ which will be nowhere near to one, with its current closeness to one then being an accidental consequence of the fact that it is this particular epoch in which we just happen to be making observations. Conformal gravity thus not only gets rid of the need for galactic dark matter, it eliminates the need for cosmological dark matter as well.
With conformal gravity having thus eliminated the need for cosmological dark matter, it is of interest to see just exactly how the conformal gravity evolution equation of Eq. (32) itself actually manages to avoid any flatness fine tuning problem.<sup>†‡‡</sup><sup>†‡‡</sup>†‡‡The author is indebted to Dr. M. Turner for asking a helpful question in this regard. Thus in the illustrative $`\lambda =0`$ case where the evolution equation is given by
$$\dot{R}^2(t)+3\dot{R}^2(t)\mathrm{\Omega }_M(t)/4\pi S_0^2L_{Pl}^2=kc^2$$
(56)
with solution
$$R^2(t,\alpha =0,k<0)=2A/\mathrm{}kS_0^2ckc^2t^2,$$
(57)
we see the two terms on the left hand side of Eq. (56) have radically different time behaviors, even as their sum remains constant (=$`kc^2`$). Specifically, $`\dot{R}^2`$ begins at zero and slowly goes to $`kc^2`$ at late times, while the $`3\dot{R}^2(t)\mathrm{\Omega }_M(t)/4\pi S_0^2L_{Pl}^2`$ term does the precise opposite as it goes to zero from an initial value of $`kc^2`$. Moreover, while these two terms turn out to be of the same magnitude at some time in the early universe, simply because the scale factor $`S_0`$ is so much larger than $`L_{Pl}^1`$, these two terms are nowhere near the same order of magnitude today, and yet their sum remains constant. As such this behavior differs radically from that found in the standard model, since there the very fact that $`L_{Pl}`$ is its natural scale forces the analogous two terms (terms which are of opposite sign in the standard model) to be of the same order of magnitude at all times right up to the present and to thus have to cancel to an extraordinary degree. Since such fine-tuning is not required in the conformal case, we see that it is the changing of the effective $`G`$ which explicitly enables us to resolve the flatness problem.
With the general $`\alpha 0`$ Hubble parameter obeying Eq. (50), we see that its current value obeys $`q(t_0)H^2(t_0)=\alpha c^2`$, with the current age of the universe then being given by $`H(t_0)t_0=arctanh[(q(t_0))^{1/2}]/(q(t_0))^{1/2}`$. Thus we see that $`t_0`$ is necessarily greater or equal to $`1/H(t_0)`$ ($`t_0=1/H(t_0)`$ when $`q(t_0)=0`$, and $`t_0=1.25/H(t_0)`$ when $`q(t_0)=1/2`$). Thus we see that conformal cosmology readily resolves another problem which has troubled the standard theory, viz. the universe age problem. Further, in conformal cosmology the (dimensionless) ratio of the particle horizon size $`d(t)`$ to the spatial radius of curvature $`R_{curv}(t)`$ (=$`(6/R^{(3)})^{1/2}`$ where $`R^3`$ is the modulus of the Ricci scalar of the spatial part of the metric) is given by
$$\frac{d(t)}{R_{curv}(t)}=(k)^{1/2}c_0^t\frac{dt}{R(t)}=(k)^{1/2}c_0^t\frac{dt}{[k(\beta 1)/2\alpha k\beta sinh^2(\alpha ^{1/2}ct)/\alpha )]^{1/2}}.$$
(58)
Thus, in the illustrative $`\lambda =0`$ case where $`q(t_0)=0`$ and where $`T_{max}^2=T^2(t_0)(11/\overline{\mathrm{\Omega }}_M(t_0))`$, Eq. (58) evaluates as
$$\frac{d(t)}{R_{curv}(t)}=log\left[\frac{T_{max}+(T_{max}^2T^2(t))^{1/2}}{T(t)}\right],$$
(59)
with the horizon size thus being altogether greater than one at recombination. Similarly in the $`q(t_0)=1/2`$, $`T_{max}T_VT(t_0)`$ case where the current time obeys $`sinh^2(\alpha ^{1/2}ct_0)=1`$ and where $`\beta 1`$, we then find for all times up to recombination that $`R(t,\alpha >0,k<0)`$ can be approximated by $`k(\beta 1)/2\alpha kc^2t^2`$, with the recombination time horizon size then being found to again be given by Eq. (59) (with $`T_{max}`$ being the $`\alpha >0`$ one this time). Conformal cosmology thus readily resolves the horizon problem, and leads to a naturally causally connected cosmology.<sup>‡\*</sup><sup>‡\*</sup>‡\*For comparison, we recall that in the typical $`k<0`$, $`\mathrm{\Lambda }=0`$ standard gravity case, this same $`d(t)/R_{curv}(t)`$ ratio is given by $`log[(T_{ref}+(T_{ref}^2+T^2(t))^{1/2})/T(t)]`$ (where $`T_{ref}^2=T^2(t_0)(1/\mathrm{\Omega }_M(t_0)1)`$) and is thus much smaller than one at recombination. Thus with the conformal gravity $`T_{max}`$ being much larger than the standard gravity $`T_{ref}`$ simply because the effective conformal $`G`$ is so much smaller than the standard one, we see that it is the changing of the effective $`G`$ which explicitly enables us to resolve the horizon problem. Thus, to sum up, we see that simply by modifying the effective cosmological $`G`$, conformal gravity is then able to resolve a whole variety of current cosmological problems, viz. the flatness, horizon, dark matter, universe age, cosmic acceleration and cosmological constant problems, and should thus be seen as a potentially viable candidate cosmological theory (for its current overall status and for the challenges that it itself still faces see ).
With conformal cosmology being rendered singularity free through the negative sign effective Einstein-Hilbert action present in the conformal matter action of Eq. (24), we see that in conformal gravity it is gravity itself which can protect itself from its own singularities. Conformal gravity thus provides a possible (though currently far from guaranteed) mechanism by which the collapse of a star can still possibly be prevented when all conventional mechanisms (such as radiation pressure or Pauli degeneracy) have failed, viz. that a scalar field condensate could be generated inside the star (perhaps when the star reaches nuclear density) which would then induce some repulsive gravity (cf. negative energy density<sup>‡†</sup><sup>‡†</sup>‡†Such a negative contribution is simply not considered in models in which positivity of the energy density is assumed, and even while the energy density of a standard kinematic perfect fluid is indeed positive, we thus see that the extension of such positivity to the entire energy-momentum tensor which serves as the source of gravity is actually questionable in general (even if absent in the flat space limit, non-inertial explicitly curvature dependent terms are not forbidden in the curved space case), as would then be the gravitational collapse theorems which rely on such positivity and assume that no negative component is ever generated during any such collapse.) and arrest the collapse (to then either generate a rebound or produce a stable configuration with radius greater than the Schwarzschild radius of the star).<sup>‡‡</sup><sup>‡‡</sup>‡‡While recent data on the velocities of stars in inner galactic regions have indicated the presence of large (black hole candidate) mass concentrations at the centers of galaxies such as M87 and the Milky Way, those data are currently unable to ascertain whether any such large mass is actually confined to a radius smaller than its Schwarzschild radius or determine whether any event horizon has actually been formed (at 400 km s<sup>-1</sup> or so the measured velocities, while large by galactic standards, are still well below the velocity of light). Thus with gravity being able to generate its own repulsion, the whole issue of gravitational singularities needs to be reconsidered, with the standard strong gravity picture possibly being another piece of the standard weak gravity intuition whose generalization might be unreliable.
Having thus presented the cosmological case for a negative spatial curvature universe, we now present some additional, quite direct observational evidence in its support, evidence from an at first somewhat unlikely source, namely the systematics of galactic rotation curves. While these curves provided the first clear evidence of the need in standard gravity for dark matter, beyond the fact that these curves show that there actually is a departure from the luminous Newtonian expectation in the outskirts of spiral galaxies, explicit study of the systematics of such departure has revealed the presence of an apparent cosmological imprint on the data, with it being cosmology itself which will actually enable us to eliminate the need for any galactic dark matter at all. Indeed, with the potentials associated with static sources in conformal gravity being ones which actually grow with distance according to Eq. (8), the very fact that they do so entails that in calculating the motions of individual particles within a given galaxy, one is now no longer able to ignore the contributions of the potentials due to distant matter sources outside of that galaxy. Thus in going to a higher order theory such as conformal gravity, we immediately transit into a world where we have to consider effects due to matter not only inside but also outside of individual systems, and thus we are led to look for both local and global imprints on galactic rotation curve data, this being a quite radical (and quite Machian) conceptual departure from the standard second order, purely local Newtonian world view.<sup>‡§</sup><sup>‡§</sup>‡§With it being only for spherically symmetrically distributed matter with $`1/r`$ potentials that exterior sources decouple locally, the very detection of any global cosmological imprint within galaxies would then argue against gravitational potentials being of a pure $`1/r`$ form.
To isolate such possible global imprints, it is instructive to look at the centripetal accelerations of the data points farthest from the centers of individual galaxies. In particular, for a set of 11 particular galaxies whose rotation curves are regarded as being particularly characteristic of the pattern of deviation from the luminous Newtonian expectation that has so far been obtained, it was found that the farthest centripetal accelerations in these galaxies could all be parameterized by the universal three component relation
$$(v^2/R)_{last}=\gamma _0c^2/2+\gamma ^{}N^{}c^2/2+\beta ^{}N^{}c^2/R^2$$
(60)
where $`\gamma _0=3.06\times 10^{30}`$ cm<sup>-1</sup>, $`\gamma ^{}=5.42\times 10^{41}`$ cm<sup>-1</sup>, $`\beta ^{}=1.48\times 10^5`$ cm, and where $`N^{}`$ is the total amount of visible matter (in solar mass units) in each galaxy. Since the luminous Newtonian contribution is decidedly non-leading at the outskirts of galaxies, we thus uncover the existence of two linear potential terms which together account for the entire measured departure from the luminous Newtonian expectation,<sup>‡¶</sup><sup>‡¶</sup>‡¶With the luminous Newtonian contribution falling with distance and with the rotation curves of the prominent bright spirals being flat, the departure from the luminous Newtonian expectation must itself thus be growing with distance, and according to Eq. (60) even be growing universally in fact. with one of these two terms depending on the number, $`N^{}`$, of stars within each given galaxy, and with the other, the $`\gamma _0c^2/2`$ term, not being dependent on the mass content of the individual galaxies at all. Moreover, since numerically $`\gamma _0`$ is found to have a magnitude of order the inverse Hubble radius, we can thus anticipate that it must represent a universal global effect generated by the matter outside of each galaxy (viz. the rest of the matter in the universe), and thus not be associated with any local dynamics within individual galaxies at all.
As regards the $`N^{}`$ dependent contribution of the matter within the individual galaxies, the integration of the non-relativistic stellar potentials $`V^{}(r)=\beta ^{}c^2/r+\gamma ^{}c^2r/2`$ over an infinitesimally thin galactic optical disk with luminous surface matter distribution $`\mathrm{\Sigma }(R)=\mathrm{\Sigma }_0`$exp$`(R/R_0)`$ and total number of stars $`N^{}=2\pi \mathrm{\Sigma }_0R_0^2`$ yields the centripetal acceleration
$$v^2/R=g_{gal}^{lum}=g_\beta ^{lum}+g_\gamma ^{lum}$$
(61)
where
$$g_\beta ^{lum}=(N^{}\beta ^{}c^2r/2R_0^3)[I_0(r/2R_0)K_0(r/2R_0)I_1(r/2R_0)K_1(r/2R_0)]$$
(62)
and where
$$g_\gamma ^{lum}=(N^{}\gamma ^{}c^2r/2R_0)I_1(r/2R_0)K_1(r/2R_0),$$
(63)
to thus yield a net acceleration which behaves asymptotically as $`(v^2/R)_{last}=\gamma ^{}N^{}c^2/2+\beta ^{}N^{}c^2/R^2`$. The conformal gravity local galactic potentials associated with the matter within any given galaxy thus nicely generate the $`N^{}`$ dependent terms exhibited in Eq. (60).
As regards the remaining $`N^{}`$ independent $`\gamma _0c^2/2`$ term, in order to be able to identify it as being of cosmological origin, we need to rewrite the comoving Hubble flow of the rest of the universe in the rest frame coordinate system of any given galaxy of interest. To determine just how a comoving geometry might look in a Schwarzschild coordinate system, we note that the general coordinate transformation
$$r=\rho /(1\gamma _0\rho /4)^2,t=𝑑\sigma /R(\sigma )$$
(64)
effects the metric transformation
$`d\tau ^2=(1+\gamma _0r)c^2dt^2{\displaystyle \frac{dr^2}{(1+\gamma _0r)}}r^2d\mathrm{\Omega }`$ (65)
$`{\displaystyle \frac{(1+\rho \gamma _0/4)^2}{R^2(\sigma )(1\rho \gamma _0/4)^2}}\left(c^2d\sigma ^2{\displaystyle \frac{R^2(\sigma )(d\rho ^2+\rho ^2d\mathrm{\Omega })}{(1\rho ^2\gamma _0^2/16)^2}}\right).`$ (66)
Thus with metrics conformal to a Robertson-Walker one also being allowed cosmological solutions in a conformal invariant theory, we see that in conformal gravity a static, Schwarzschild coordinate linear potential metric is coordinate equivalent to conformal cosmologies in which the spatial curvature $`k=\gamma _0^2/4`$ is expressly negative,<sup>‡∥</sup><sup>‡∥</sup>‡∥Positive $`k`$ would lead to a complex $`\gamma _0`$ in Eq. (66), with it not being possible for the topologically open $`d\tau ^2=(1+\gamma _0r)c^2dt^2dr^2/(1+\gamma _0r)r^2d\mathrm{\Omega }`$ metric to ever be equivalent to anything other than other topologically open ones. with a universal linear potential thus being the local manifestation of global negative curvature, and with the local $`\gamma _0`$ and the global $`k`$ thus having a common connection. As such, this connection is actually geometrically quite natural, since not only does negative $`k`$ lead to repulsion, but, as had been noted earlier, so does positive $`\gamma _0`$. However, while a positive $`\gamma _0r`$ metric term does indeed lead to gravitational deflection of light, nonetheless, for non-relativistic systems this same positive $`\gamma _0r`$ term acts attractively, with cosmological negative spatial curvature thus generating an attractive gravitational effect for non-relativistic motions within galaxies, an effect which the standard theory in essence tries to simulate locally by the introduction of local galactic dark matter. Now while we had initially identified the static $`\gamma _0`$ term of Eq. (60) as being of cosmological origin since its phenomenologically measured value was found to be of order the inverse of the Hubble radius, such an identification could at best have only been heuristic since the Hubble parameter itself is a time dependent one which varies from one epoch to the next. However, cosmology actually possesses a second scale beyond that associated with its expansion rate, namely that associated with its spatial curvature, with it being this latter, epoch independent one, with which $`\gamma _0`$ is then nicely identified, with the $`\gamma _0r`$ term thus serving as a time independent universal potential term no matter what the epoch.
Further, for galaxies which have no peculiar velocities with respect to the Hubble flow, i.e. for any galaxy whose center can precisely serve as the coordinate origin for the radial coordinate $`r`$ in the transformation of Eq. (64), the metrics of Eqs. (61) and (66) can simply be added in the weak gravity limit, to yield as the net weak gravity centripetal acceleration
$$v^2/R=g_{tot}=g_\beta ^{lum}+g_\gamma ^{lum}+\gamma _0c^2/2,$$
(67)
an acceleration whose asymptotic limit precisely yields Eq. (60) for $`(v^2/R)_{last}`$. Moreover, not only does Eq. (67) yield this requisite asymptotic formula, its use in the sub-asymptotic region as well is then found to provide acceptable parameter free (and thus dark matter free) fitting to all of the (in excess of 250) data points in the 11 galaxy sample, with Eq. (67) thus capturing the essence of the data.<sup>‡\**</sup><sup>‡\**</sup>‡\**For comparison, standard dark matter halo fitting uses two free parameters per halo and thus no less than 22 additional free parameters for the same 11 galaxy sample. Thus we identify an explicit imprint of cosmology on galactic rotation curves, recognize that it is its neglect which may have led to the need for dark matter, and for our purposes here confirm that $`k`$ is indeed negative, just as had been required in the cosmological study which we presented above, with the phenomenological formula of Eq. (60) actually providing an explicit measurement of the curvature of the universe which test particles sample as they orbit in galaxies.
Moreover, not only does Eq. (67) provide for an acceptable accounting of galactic rotation curve data, the magnitude obtained for the stellar $`\gamma ^{}`$ is found to be so small (of order $`10^{41}`$ cm<sup>-1</sup>) that the linear potential term then makes a completely negligible contribution on solar system distance scales, with the metric of Eq. (8) thus reducing to the standard Schwarzschild one within the solar system. The strength of the linear potential terms in Eq. (67) thus serve as the scale which is to parameterize departures from the luminous Newtonian expectation (for the bright spirals $`\gamma ^{}N^{}`$ and $`\gamma _0`$ are of the same order of magnitude), with such a scale nicely explaining why no dark matter is needed on solar system distance scales, with the solar system simply being too small to be sensitive to any cosmologically relevant scale. With the linear potential term first becoming competitive with the Newtonian one on none other than galactic distance scales, we thus explain not only why solar system physics is unaffected by conformal gravity, but also, we identify at exactly what point departures from the luminous Newtonian contribution are to first set in. The (negative) spatial curvature of the universe thus sets the scale at which standard gravity needs to introduce dark matter in order to avoid failing to fit data.
Our uncovering of a universal acceleration in conformal gravity immediately recalls the presence of a similar one in the MOND theory , one also of a cosmologically significant magnitude. Specifically, Milgrom had suggested that if a universal acceleration $`a_0`$ did exist, then Newton’s law of gravity could possibly be phenomenologically modified into a form such as $`v^2/R=\nu (a_0/g_N)g_N`$ where $`g_N=g_\beta ^{lum}`$ in the galactic case. The candidate functional form $`\nu (x)=(1/2+(4x^2+1)^{1/2}/2)^{1/2}`$ would then yield
$$v^2/R=g_N\{1/2+(g_N^2+4a_0^2)^{1/2}/2g_N\}^{1/2},$$
(68)
an expression which, despite the absence so far of any deeper underlying theory, is nonetheless found to perform extremely well phenomenologically. While Eqs. (67) and (68) have different underlying motivations, it is of interest to note that Eq. (67) would in fact fall into the general MOND approach if the MOND formula were to be reinterpreted as
$$v^2/R=\nu (\gamma _0c^2/2g_{loc})g_{loc}$$
(69)
where $`g_{loc}`$ is the entire local luminous galactic contribution $`g_{gal}^{lum}`$ given in Eq. (61), and if the function $`\nu (x)`$ were to instead take the form $`\nu (x)=1+x`$. Conformal gravity thus not only provides a rationale for why there is in fact a universal acceleration in the first place (something simply assumed in MOND), but also it even yields an explicit form for the function $`\nu (x)`$, albeit not the one commonly utilized in the standard MOND studies (where $`g_{loc}`$ is taken to be the purely Newtonian $`g_N`$). However, despite such differences, both theories have in common the recognition that there is a universal scale which is to parameterize departures from the standard luminous Newtonian expectation, and that its magnitude is a cosmologically significant one.
Further support for the existence of such a scale has been presented by McGaugh in a study of the behavior of the quantity $`M_{dyn}(R)/M_{lum}(R)`$ as a function of the measured orbital acceleration $`v^2(R)/R`$ at points $`R`$ within galaxies. ($`M_{dyn}(R)=Rv^2(R)/G`$ is the amount of matter interior to $`R`$ as would be required by Newtonian gravity, while $`M_{lum}(R)`$ is the amount of luminous matter detected in the same region). In this study McGaugh found that mass discrepancies (viz $`M_{dyn}(R)/M_{lum}(R)>1`$) systematically occurred in galaxies whenever the measured $`v^2(R)/R`$ fell below a universal value of $`10^8`$ cm s<sup>-2</sup> or so, a value which is immediately recognized as being close to the values of the $`a_0`$ and $`\gamma _0^2c^2/2`$ acceleration parameters which were respectively phenomenologically obtained in the MOND and conformal gravity theories. While MOND and conformal gravity might differ as to how the centripetal accelerations should behave in regions where there are measured mass discrepancies, both theories (and the data) thus agree that there is a universal scale which determines when such discrepancies should first set in.
Hence, independent of the merits of alternate theories such as conformal gravity or MOND themselves, it would appear that the data possess a cosmological imprint, an imprint which heralds when dark matter is first needed in the standard theory, with the very existence of such an imprint enabling Eqs. (67) and (68) to organize the data in an extremely economical fashion. Thus even if one does not want to contemplate going beyond standard gravity, the galactic data themselves seem to be insisting that dark matter theories should be parameter free (i.e. that they should be formulatable without the (extravagant) need for two free parameters per galactic halo), and that there is a cosmological imprint in the data which dark matter theories must be able to produce. Moreover, apart from the fact that dynamical dark matter models have not yet produced such a scale (say by studying the growth of galaxy fluctuations in cosmology), it would appear difficult for them to ever be able to do so in the standard cold dark matter flat inflationary universe cosmological model, since simply by virtue of being flat such a cosmology then lacks the one key ingredient which leads to a universal galactic scale in the conformal theory, namely a non-zero spatial curvature, with the standard flat classical cosmological model simply not possessing any intrinsic such universal scale at all.<sup>ࠠ</sup><sup>ࠠ</sup>ࠠSee for an interesting attempt to generate such a scale quantum-mechanically as a renormalization group correlation length associated with a quantum gravity fixed point, a point at which scale invariance (such as that in conformal gravity) is realized via anomalous dimensions. Given the above observational support for the existence of such a scale, galactic data thus seem to be supporting the notion that the universe has a non-zero spatial curvature, a non-trivial curvature which conformal gravity (and for the moment only conformal gravity apparently) can readily and naturally produce, a curvature which releases gravity from having to always be attractive.
Thus to conclude, we believe that it is not yet justified to assert that gravity is always attractive, and that in fact a repulsive cosmological component immediately allows one to resolve a whole host of problems which currently beset the standard theory. And even if conformal gravity itself should not turn out to be the correct extrapolation of the standard solar system wisdom (as a cosmology conformal gravity is not without challenges of its own ), that would in no way constitute evidence in favor of the correctness of the standard extrapolation. Since our study of conformal gravity has shown that the problems with which the standard theory is currently afflicted are not in fact generic to cosmology, their very existence could be a warning that the extrapolation of standard gravity beyond the confines of the solar system might be a lot less reliable than is commonly believed, with gravity not always being as attractive as Newton initially took it to be.
## VI Author’s Note
During the last few years I have written articles on gravitational theory for special issues of Foundations of Physics in honor of my longtime colleagues and friends Fritz Rohrlich and Larry Horwitz. It is a great pleasure for me to dedicate this third article in that series to another equally close colleague and friend Kurt Haller. This work has been supported in part by the Department of Energy under grant No. DE-FG02-92ER4071400. |
warning/0001/astro-ph0001358.html | ar5iv | text | # Modeling the Photoionized Interface in Blister HII Regions
## 1 Introduction
A blister H II region is formed when a massive star is born near the edge of a molecular cloud. The radiation from the newborn star rapidly ionizes the surrounding material and the ionization front breaks through the surface of the cloud creating a cavity. The photoionized gas in the cavity is exposed to view and can be observed as a blister on the surface of the molecular cloud. The concept of a blister was introduced by Zuckerman (1973) to describe the Orion Nebula, one of the nearest and definitely the best studied H II region. The term “blister model” was introduced by Israel (1978) who studied a sample of about 30 galactic H II regions and, based on their positional relationship to CO emitting molecular clouds, concluded that almost every optically observable H II region is a blister H II region.
Blister H II regions serve as useful probes of current interstellar abundances in the Galaxy, since the stars responsible for maintaining the ionization are less than about 10 million years old. Classical techniques of nebular analysis (described in several textbooks, such as Aller (1984) and Osterbrock (1989)), based mainly on optical spectra, have been used to determine abundances in several H II regions and map abundances in the Galaxy (e.g. Hawley (1978), Shaver et al. (1983)). These methods have two well known uncertainties. First, lines from all the ionization stages of an element present in an H II region are usually not observed. This requires prescriptions to correct for unseen stages of one element based on ionic ratios of a different element. Second, temperatures derived from lines of one species (such as \[O III\]) are assumed when obtaining the abundance of another species (such as \[S II\]) which is formed in a different zone. These uncertainties are discussed by French & Grandi (1981).
An alternative method of determining abundances is by using photoionization models to interpret H II region spectra. This approach has been applied in some detail to the Orion Nebula (Baldwin et al. (1991), Rubin et al. (1991)) and has the advantage that the electron temperature and ionization structure are calculated self-consistently. Furthermore, it becomes unnecessary to implement ad-hoc ionization correction schemes. However a major limitation of this method (as emphasized by Rubin et al. (1998)) is that the input density and geometry need to be specified in order to calculate models, and these in general are not well known.
In a blister H II region, the density profile of the interface between the molecular cloud and the ionized gas is determined by the ionizing radiation driving a photoevaporative flow off the surface of the molecular cloud. The sharply stratified ionization structure of this interface was observed in the Orion Nebula by Hester et al. (1991) in Hubble Space Telescope (HST) Wide Field Camera narrow band images. The ionization stratification was discussed in the broader context of interpreting H II region spectra by Hester (1991), who also suggested that most of the \[S II\] emission arises in a very narrow transition zone between the H II region interior and the photodissociation region of the molecular cloud. The transition zone occurs beyond the hydrogen ionization edge, where photons with energies higher than 10.4 eV (the ionization potential of S<sup>0</sup>) keep the sulphur singly ionized. More recently, Hester et al. (1996) presented HST Wide Field and Planetary Camera - 2 (WFPC2) images of the “elephant trunk” structures in the blister H II region M 16. In these images, the photoionized interface is seen in tangency and the ionization stratification is clearly resolved. The \[S II\] emitting zone is very narrow – it has a width of about $`8\times 10^{15}`$ cm ($``$ 0$`\stackrel{}{\mathrm{.}}`$3 at the assumed distance of 2000 pc). In that work, we used an empirical density profile derived from the H$`\alpha `$ emission profile and presented a photoionization model that successfully reproduced the main features of the emission from the interface.
In this paper, we use photoionization models in conjunction with high spatial resolution data to develop a framework for interpreting H II region spectra. Arbitrary assumptions about the density structure are not made, since the models are constrained by the structure observed in the high resolution images. Once such a framework is established, it will be a powerful method for obtaining the physical properties of the emitting gas in H II regions.
We first present a grid of models for the interface in blister H II regions using the density profiles calculated by Bertoldi (1989) for photoionized, photoevaporative flows. (These calculations are a significant improvement over earlier work because they allow for an ionization front of finite width, rather than treating it as a discontinuity, and they treat non-equilibrium ionization and energy deposition). We vary the incident stellar continuum, the ionizing flux and the elemental abundances and examine the dependence of nebular properties on these input parameters. We focus on the diagnostics provided by spatially resolved line strengths (such as can be obtained by HST for several nearby H II regions). We then consider the M 16 data in detail and present a photoionization model for the emission. We compare the model results to published spectra and show how knowledge of the structure of the emitting region is crucial for interpreting ground based observations. We conclude with a summary and a consideration of the implications of our work, and future directions.
## 2 Photoionization Models
### 2.1 Input Parameters
The basic input parameters for the photoionization models are the shape and intensity of the incident ionizing continuum and the elemental abundances. The interface between the molecular cloud and the H II region interior is the result of a photoevaporative flow driven by the incident stellar continuum (which is also responsible for ionizing the gas). The solutions for such a flow determine the density profile of the interface, which also needs to be specified.
Hydrodynamic models for photoionized, photoevaporative flows off spherical cloud surfaces have been calculated by Bertoldi (1989) and Bertoldi & Draine (1996). The form of the density profile they find may be written as follows:
$$n(x)=n_0(1+\frac{Rx}{r_c})^{2.5}$$
(1)
Here, $`x`$ is the distance from the ionizing source, $`R`$ is the distance from the source to the ionization edge, and $`n_0`$ is the number density at $`x=R`$. (Note that the density is a function of the distance from the interface, $`Rx`$, and the actual distance from the source is not explicitly required in the models). The radius of curvature of the evaporating surface is given by $`r_c`$ and is effectively the scale length for the flow. The values of $`R`$ and $`n_0`$ depend on the ionizing continuum. For each of our models (see below), we start by performing a calculation with $`R=5\times 10^{17}`$ cm and $`n_0=1500`$ cm<sup>-3</sup>, and then iterate using the predicted values at the ionization edge till we obtain the correct values. (We found that 3 iterations were sufficient in every case). We keep the value of $`r_c`$ fixed at $`2\times 10^{17}`$ cm for all the models, anticipating our analysis of the M 16 data in §3.
The validity of using equation 1 for the density profile is subject to the following caveat. The models calculated by Bertoldi (1989) did not include the effects of dust on the flow. It was shown by Baldwin et al. (1991) that grains absorb, and are ionized by the incident continuum and cause a radial acceleration of the gas away from the star. This phenomenon will affect the outer regions of the flow and alter the density profile. In order to asses the effect on the density profile at and near the interface (which we are concerned with here), the hydrodynamic and photoionization problem would have to be solved self consistently, taking into account the presence of dust - a task beyond the scope of this paper.
We calculate a grid of 60 models, using all combinations drawn from 4 stellar atmospheres, 3 values for the ionizing flux and 5 sets of elemental abundances. The stellar atmospheres used are PHOENIX models of B0 (T<sub>eff</sub> = 33,340 K), O8 (T<sub>eff</sub> = 38,450 K), O6 (T<sub>eff</sub> = 43,560 K), and O4 (T<sub>eff</sub> = 48,670 K) stars, all with gravities, log(g) = 3.9 (Hauschildt, Baron & Allard (1997), Aufdenberg, private communication). PHOENIX models represent an advance over earlier model stellar atmospheres since they use spherical geometry, calculate the strongest lines of important elements in non-LTE, and include line blanketing. In Figure Modeling the Photoionized Interface in Blister HII Regions these model spectra are plotted, normalized to their respective peak values. Also shown on the plot are the ionization potentials of S<sup>0</sup> (10.36 eV) and O<sup>+</sup> (35.1 eV). The values used for the incident ionizing flux are log \[$`\mathrm{\Phi }`$(H)\] = 10.5, 11.0 and 11.5, where where $`\mathrm{\Phi }`$(H) is the number of ionizing photons (energy $``$ 13.6 eV) per second per cm<sup>2</sup> incident on the gas. In all the models we keep the helium abundance at solar. The heavier element abundances are varied simultaneously and values of 0.2, 0.4, 0.6, 0.8 and 1.0 times solar (where solar values are taken from Grevesse & Anders (1989)) are used in the grid of models. We use the publicly available code, CLOUDY (Ferland et al. (1998)) to calculate the photoionization models.
One concern about the modeling procedure is that we are using an equilibrium code (CLOUDY) to model the emission from a photoevaporative flow where non-equilibrium effects may be important. At the interface, the cooling times are short (a few tens of years) compared to the dynamic time scales (a few thousand years), so thermal equilibrium is a valid assumption. However sufficiently close to the ionization front, photoionization equilibrium will not hold. The effect of non-equilibrium on the predicted emission lines was studied by Harrington (1977). He found that the ionization fraction differed only moderately between equilibrium and non-equilibrium models, and that happened within about 10<sup>15</sup> cm of the front. He also found that the lines strengths most affected were \[N I\] $`\lambda `$5199 and \[O I\] $`\lambda `$6300 which varied about 40% and 15% respectively. Both these lines are formed beyond and are narrower than the \[S II\] zone. While it would be useful to examine the effect of photoionization non-equilibrium in detail for a wider range of conditions than was done by Harrington (1977), we expect our models to be suitable representation of reality.
### 2.2 Results
The photoionization models predict the highly stratified ionization structure of the interface. The H$`\alpha `$ more or less follows the density profile as expected. The \[S II\] comes from a narrow zone near the interface that peaks sharply just beyond the H$`\alpha `$ peak, in a region where the ionization fraction of hydrogen is rapidly falling and sulphur (whose ionization potential is lower than hydrogen) is kept singly ionized. The \[O III\] emission profile is always less peaked and often quite extended compared to the H$`\alpha `$ and \[S II\]. While all models show these general characteristics of the low ionization and high ionization gas, there are significant differences in the details of the spatial profiles and line strengths among the various models. We now turn to a discussion of these differences and their dependence on input parameters.
In Figure Modeling the Photoionized Interface in Blister HII Regions we show the density and the H$`\alpha `$, \[S II\] $`\lambda \lambda `$6716,6731 and \[O III\] $`\lambda `$5007 emissivity as a function of distance from the interface for two models. Both models use an O6 star continuum and abundances of 0.6 solar for elements heavier than helium. The models shown differ in the incident ionizing fluxes – $`\mathrm{\Phi }_\mathrm{H}=10^{10.5}`$ photons s<sup>-1</sup> cm<sup>-2</sup> (top panel) and $`\mathrm{\Phi }_\mathrm{H}=10^{11.5}`$ photons s<sup>-1</sup> cm<sup>-2</sup> (bottom panel). The most dramatic difference between the two models is the density at the ionization front, which is about 3 times higher for the higher incident flux. Consequently the peak H$`\alpha `$ emission is higher by about a factor of 10 (since emissivity is proportional to the square of the density). This is a result of the ionization front being pushed into a cloud where the density increases towards the interior – a larger flux of ionizing photons will push the front further up the density ramp. The \[S II\] and \[O III\] emissivities are also correspondingly higher.
In both plots of Figure Modeling the Photoionized Interface in Blister HII Regions, the origin of the x-axis is chosen to be at the peak of the \[S II\] emission. We can see from these plots, that the width of the \[S II\] zone varies considerably with changes in the incident ionizing flux. If we define the width as being the distance at which the \[S II\] emissivity falls to half its peak value, then the widths for the two cases shown are $`11.2\times 10^{15}`$ cm for the lower ionizing flux (top panel) and $`4.4\times 10^{15}`$ cm for the higher ionizing flux (bottom panel). At a distance of 1 kpc, these widths correspond to 0$`\stackrel{}{\mathrm{.}}`$75 and 0$`\stackrel{}{\mathrm{.}}`$29 respectively.
In Figure Modeling the Photoionized Interface in Blister HII Regions, models with different shapes for the incident continuum are shown. The top and bottom panels show models using O8 and O4 stellar atmospheres. Both models use an incident ionizing flux of $`10^{11.0}`$ photons s<sup>-1</sup> cm<sup>-2</sup> and metal abundances of 0.6 times solar. The \[O III\] emission is higher by a factor of about 6 for the O4 model while H$`\alpha `$ and \[S II\] are about the same. This is because the fraction of photons produced which are capable of ionizing oxygen to O<sup>++</sup> is much lower for O8 stars than for O4 stars (Figure Modeling the Photoionized Interface in Blister HII Regions). Here again, the \[S II\] profiles are significantly different. The O8 star produces an \[S II\] zone that is $`12.3\times 10^{15}`$ cm wide (top panel) and the hotter O4 star produces a narrower \[S II\] zone ($`6.4\times 10^{15}`$ cm).
The variation of emission profiles with abundance is shown in Figure Modeling the Photoionized Interface in Blister HII Regions. Both models use an O6 stellar atmosphere and an ionizing flux of $`10^{11.0}`$ photons s<sup>-1</sup> cm<sup>-2</sup>. The models shown in the top and bottom panels have the abundances of all elements heavier than helium set to 0.2 times solar and 0.8 times solar respectively. The H$`\alpha `$ intensity is about the same in both cases, and the total \[S II\] intensity is 1.6 times higher in the model with the higher abundance. The ratio of the peak \[S II\] emissivity to the peak H$`\alpha `$ emissivity shows a stronger contrast between the models. In the case of the low abundance model this ratio is 0.27, while for the high abundance model it is 0.55, which is a factor of 2 higher.
#### 2.2.1 Stellar Atmosphere Models
The spectral energy distribution of the incident ionizing continuum is a key ingredient in HII region models. Since the ionizing photons cannot be observed for any but a handful of stars, we depend on continua predicted by theoretically computed stellar atmospheres. In this paper we are using atmospheres calculated by the PHOENIX code, which treats the atmosphere as spherically symmetric, includes line blanketing, and treats the strongest lines of several important elements in non-LTE. These models are an improvement over the widely used ATLAS models (Kurucz (1991)) which are plane-parallel and neglect non-LTE effects entirely.
Another set of theoretical non-LTE stellar atmospheres, the CoStar Models (Schaerer & de Koter (1997)) have also been used recently in H II region models (Stasińska & Schaerer (1997)). CoStar and PHOENIX models differ in significant ways. CoStar models include the effects of stellar winds on the emitted continuum, while the PHOENIX models are spherically extended but hydrostatic atmospheres. However, CoStar models do not include line blanketing in the calculation of the temperature structure whereas PHOENIX models include line blanketing in a self consistent way in determining the temperature structure of the atmosphere. Furthermore CoStar models treat only H and He in non-LTE while PHOENIX treats over 50,000 of the strongest lines of several elements in non-LTE. In this section, we examine the effect of using different model atmospheres on the predicted H II region photoionization model.
Figure Modeling the Photoionized Interface in Blister HII Regions shows the spectral energy distributions of the PHOENIX O4V model (T = 48670 K) and the CoStar E2 model (T = 48500 K). Also shown are the S<sup>0</sup> and O<sup>++</sup> ionization edges. The CoStar spectrum is flatter and has a greater flux at higher energies than the PHOENIX spectrum. This effect is due to the wind and is seen in Stasińska & Schaerer (1997) where CoStar and ATLAS spectra are compared. In order to understand the effects of using different stellar atmospheres on H II region spectra, we calculated a photoionization model using the CoStar spectrum along with an incident ionizing flux of $`10^{11.0}`$ photons s<sup>-1</sup> cm<sup>-2</sup>, and metal abundances of 0.6 times solar. In Figure Modeling the Photoionized Interface in Blister HII Regions we compare this CoStar model with a model using the PHOENIX O4V ionizing continuum with the same incident flux and abundances. The difference between these models is significant - the CoStar continuum produces twice as much \[O III\] as the PHOENIX continuum. The CoStar model also produces more \[S II\] and has a somewhat broader \[S II\] zone. The peak H$`\alpha `$ is lower in the CoStar model, so the peak \[S II\] to peak H$`\alpha `$ is higher by a factor of 2 compared to the PHOENIX model.
We wish to caution the reader that currently it is very uncertain as to which of these models (PHOENIX or CoStar) is closer to reality. However our comparison has shown that in their current forms, the ionizing continua from by PHOENIX and CoStar predict very different H II region line strengths. A thorough understanding of H II region spectra can be of value in validating atmosphere models as pointed out by Rubin, Kunze, & Yamamoto (1995). It is beyond the scope of this work to pursue this issue further and for the rest of the paper, we will discuss models using only PHOENIX stellar continua.
### 2.3 Diagnostics
The grid of models we have calculated predict several quantities that could serve as useful starting points in analyzing spectra of blister H II regions, particularly for the cases where the emitting interface is spatially resolved. We now present and discuss some of these diagnostics.
In most of the following plots, we will be visualizing the data in a somewhat specialized way which is worth describing here. The quantity of interest is plotted on the y-axis. The x-axis is simply the model number (from 1 to 60, with the tick marks suppressed). The 4 large divisions in the plot correspond to stellar atmospheres used in the model (models 1 through 15 use a B0 atmosphere and so on). Within each set of 15 models, groups of 5 (differentiated by the symbols used) correspond to different values for the ionizing flux. Finally, within each group of 5, the heavy element abundances go from 0.2 solar to 1.0 solar in steps of 0.2 with increasing model number.
In the top panel of Figure Modeling the Photoionized Interface in Blister HII Regions we show the maximum density reached in each model. This is the hydrogen density at the ionization edge defined (for convenience) to be the point where $`n_e`$/$`n_H`$ has fallen below 0.001. (We note, however that the observed lines are not emitted so far into the photodissociation region – for instance, the \[S II\] peak occurs where $`n_e0.7n_H`$). The density at the ionization edge depends strongly on the ionizing flux (for the reason mentioned in the previous section). There is, however no dependence on the stellar temperature; (note that the four large divisions in the plot correspond to different ionizing stars). For higher incident flux, there is a noticeable dependence on the abundance. A higher abundance of metals implies that more atoms compete for the ionizing photons and the maximum density reached is lower. The peak H$`\alpha `$ emission depends on the density at the ionization edge. The bottom panel of Figure Modeling the Photoionized Interface in Blister HII Regions is a scatter plot of the peak H$`\alpha `$ against the maximum density reached in the models. As expected there is a strong correlation between the H$`\alpha `$ peak emissivity and the maximum density. This in turn implies that the peak H$`\alpha `$ emissivity is a good discriminant for the incident ionizing flux.
Next, we discuss the peak emissivities and total line intensities of three important diagnostic lines in H II regions – \[O III\] $`\lambda `$5007, \[S II\] $`\lambda \lambda `$6716,6731 and \[N II\] $`\lambda `$6584. In what follows, all peak emissivities and total intensities are taken relative to H$`\alpha `$, and this is to be understood even if not explicitly mentioned. Note that the peak emissivity for different lines occurs at different locations.
Figure Modeling the Photoionized Interface in Blister HII Regions shows plots of the peak \[O III\] emissivity (top panel) and the total \[O III\] intensity (bottom panel) for the grid of models. Both these quantities are most sensitive to the stellar type. \[O III\] $`\lambda `$5007 is a high ionization line and in spectra of H II regions, it is an indicator of the hardness of the ionizing spectrum. For a given ionizing flux, either the peak \[O III\] emissivity or the total \[O III\] intensity could be used to distinguish the shape of the incident ionizing continuum. For models using a given star type and ionizing flux, the \[O III\] peak emissivity and intensities reach a maximum for abundances about 0.4 solar. The decrease in emission for lower abundances is simply because there is less oxygen. For higher abundances, the increased oxygen abundance leads to the \[O III\] 52 $`\mu `$m and 88 $`\mu `$m infrared lines dominating the cooling and lowering the electron temperature of the gas (see e.g. Henry (1993)). At lower temperatures the intensity of \[O III\], a collisionally excited line, decreases.
In contrast to \[O III\], the \[S II\] 6716,6731 Å lines trace low ionization gas. These \[S II\] lines are the only strong optical lines for any ionization stage of sulphur and are therefore have been of great importance in estimating S abundances in nebulae. However the dominant state of sulphur in most H II regions is S<sup>++</sup> and obtaining only the S<sup>+</sup> abundance leads to large uncertainties in the total sulphur abundance estimates (e.g. Dennefeld & Stasińska (1983)). In Figure Modeling the Photoionized Interface in Blister HII Regions we show plots of the peak \[S II\] emissivity (top panel) and the total \[S II\] intensity (bottom panel) for the grid of models. (Note that the y-axis scales are different on the two plots). For models with the same ionizing continuum and flux, the peak \[S II\] emissivity is more sensitive than the total \[S II\] intensity to changes in abundance, at least for lower abundances. The sensitivity gets better when the ionizing continuum is from hotter stars (O4 and O6 in our grid). Conversely, the peak \[S II\] is less sensitive than the total \[S II\] intensity to the incident ionizing flux. (Both quantities decrease with increased ionizing flux). Therefore, with spatially resolved data for an H II region interface, the peak \[S II\] emissivity can be used to estimate the sulphur abundance.
The ionization potential of N<sup>0</sup> is 14.5 eV, somewhat higher than that of hydrogen. In addition to direct ionization from the ground state, N<sup>0</sup> in the excited <sup>2</sup>D state can be ionized by photons below the Lyman limit, and in some cases the latter may dominate the ionization balance, and affect the \[N II\] emission lines. However in all our models, the \[N II\] $`\lambda `$6584 emission arises in between the \[S II\] zone and the \[O III\] zone and can be considered a tracer of intermediate ionization gas. The strength of this line is generally used to estimate the abundance of N<sup>+</sup> and via the use of an ionization correction factor, the total nitrogen abundance as well (e.g. Mathis (1982)). Figure Modeling the Photoionized Interface in Blister HII Regions shows the peak \[N II\] emissivity (top panel) and the total \[N II\] intensity (bottom panel) for the grid of models. As was the case for \[S II\], these plots show that the peak \[N II\] is more sensitive indicator of abundance than the total intensity. For models using O8, O6 and O4 stellar continua, the \[N II\] intensity decreases with increasing flux. Also, the \[N II\] intensity decreases for hotter ionizing stars. The reason is that in both these cases more nitrogen gets ionized to N<sup>++</sup> due to the larger number of energetic photons. However, the peak emissivity depends on the N<sup>+</sup> density in the zone where the emission actually reaches its maximum. This ion density is proportional to the total density, which increases with increasing ionizing flux. The peak \[N II\] emissivity is therefore higher for greater ionizing flux and for hotter stars.
We have presented a grid of models and discussed the results and diagnostics obtained from them. We now consider the specific case of the blister H II region M 16 as an example of how these model results can be applied.
## 3 The Blister H II Region M 16
HST-WFPC2 images of three elephant trunk structures in M 16 have been presented and discussed in detail by Hester et al. (1996). Figure Modeling the Photoionized Interface in Blister HII Regions shows the Planetary Camera image of the head of the second column. The images have a resolution of 0$`\stackrel{}{\mathrm{.}}`$046, corresponding to a linear distance of $`1.35\times 10^{15}`$ cm at an assumed distance of 2000 pc (e.g. Hillenbrand et al. (1993)). The H$`\alpha `$ and \[S II\] images show the ionized interface between the opaque molecular cloud and the H II region. Striations in the H$`\alpha `$ image are due to emission from the photoevaporated gas streaming away from the molecular cloud surface. The \[O III\] emission is much more extended than the H$`\alpha `$ and \[S II\] emission. The ratio map of \[S II\] to H$`\alpha `$ peaks around the edge of the column. While it may not be obvious from the image, the ratio along most of the interface is more or less uniform.
For our analysis and modeling, we will concentrate on the emission along a spatial cut across the interface at the tip of the column. The location is shown in the \[S II\] image in Figure Modeling the Photoionized Interface in Blister HII Regions. Line flux profiles were taken along the cut, and averaged over a width of 5 pixels. Background intensities were subtracted off, and in addition the \[O III\] profile was smoothed to remove noise. We note that the background intensity is mainly due to the back wall of the cavity with some contribution from the outer parts of the photoevaporative flow. Here we are concentrating on emission from a narrow region around the interface itself. Plots of the H$`\alpha `$, \[S II\] and \[O III\] intensity profiles are shown in Figure Modeling the Photoionized Interface in Blister HII Regions (top panel). The intensities shown in these plots are reddening corrected, taking E<sub>B-V</sub> = 0.7 (Chini & Krügel (1983)) and the ratio of total to selective extinction, R = 3.1, the standard value. For the photoionization model, we assume that the density profile for the emitting gas is given by equation 1 and from Figure Modeling the Photoionized Interface in Blister HII Regions, we find the radius of curvature, $`r_c`$ to be $`2\times 10^{17}`$ cm.
The peak H$`\alpha `$ surface brightness is about $`5.5\times 10^{13}`$ erg s<sup>-1</sup> cm<sup>-2</sup> arcsec<sup>-2</sup>. From the observed geometry (see Figure Modeling the Photoionized Interface in Blister HII Regions) we estimate that the path length through the emitting region lies between about $`6.0\times 10^{16}`$ cm and $`7.5\times 10^{16}`$ cm. These values give a peak H$`\alpha `$ emissivity lying between about $`4\times 10^{18}`$ ergs s<sup>-1</sup> cm<sup>-3</sup> and $`5\times 10^{18}`$ ergs s<sup>-1</sup> cm<sup>-3</sup>. The conversion of observed to intrinsic flux depends on the extinction. For M 16, Chini & Krügel (1983) have reported that R may be as high as 4.7 (rather than the standard value of 3.1 which we have so far assumed). In that case, the H$`\alpha `$ emissivity is much higher. Therefore, from Figure Modeling the Photoionized Interface in Blister HII Regions we estimate that the observed H$`\alpha `$ emission requires the density at the interface n$`{}_{\mathrm{H}}{}^{}4000`$ cm<sup>-3</sup> and correspondingly an incident ionizing flux $`\mathrm{\Phi }_\mathrm{H}>10^{11.0}`$ photons s<sup>-1</sup> cm<sup>-2</sup>.
The observed \[O III\] emission is weak relative to the H$`\alpha `$ emission (Figure Modeling the Photoionized Interface in Blister HII Regions, top panel). By integrating the observed profiles out to a distance of about $`2\times 10^{17}`$ cm, we estimate that the total \[O III\] intensity is 0.7 times the H$`\alpha `$ intensity. The ionizing flux from stars hotter than O6 or cooler than O8 cannot produce this ratio (Figure Modeling the Photoionized Interface in Blister HII Regions, bottom panel). The ratio of the peak \[S II\] emissivity to the peak H$`\alpha `$ emissivity is between 0.2 and 0.3 (Figure Modeling the Photoionized Interface in Blister HII Regions, top panel). From our diagnostic plot (top panel of Figure Modeling the Photoionized Interface in Blister HII Regions), this implies a sulphur abundance lower than about 0.4 solar. Our grid of models uses abundance sets where all the elements heavier than helium vary in lock-step. This is a simplification. For modeling the interface in M 16 we fixed all abundances (except for sulphur) based on the values reported by Hawley (1978), Shaver et al. (1983) and Dennefeld & Stasińska (1983). These values are – helium solar; nitrogen 0.5 times solar; oxygen and the other metals 0.8 times solar. The sulphur abundance was allowed to vary.
The H$`\alpha `$, \[S II\] and \[O III\] profiles, along with the density profile for our final M 16 model are presented in the bottom panel of Figure Modeling the Photoionized Interface in Blister HII Regions. The ionizing continuum used in the model is from an O7Ia supergiant, with T$`{}_{eff}{}^{}=38720`$ K and the ionizing flux at the interface is $`3\times 10^{11}`$ photons s<sup>-1</sup> cm<sup>-2</sup>. The sulphur abundance is 0.4 times solar. The photoionization model reproduces the observed emission from the interface, including the stratified ionization structure of the flow. In matching the model to the observation, we have assumed a plane-parallel geometry – clearly reality is more complex (indeed we have assumed a scale length for the flow based on the radius of curvature of the column). However since we are concentrating on the emission from close to the interface, for our purposes the plane parallel approximation is sufficient.
In Table 1 we present the observed (reddening-corrected) spectrum of M 16 measured by Hawley (1978) along with the spectrum predicted by our final model. All intensities are given relative to H$`\alpha `$ = 100. The two observed spectra were taken along slits 2$`\stackrel{}{\mathrm{.}}`$4 by 4$`\stackrel{}{\mathrm{.}}`$0 separated by 35″ along an east-west line in a region at the top of column 1 of the HST image presented by Hester et al. (1996). The spectra at the observed positions vary significantly. The \[S II\] $`\lambda \lambda `$ 6716,6731 for example is higher by a factor of 4 at position 2. The \[O III\] $`\lambda `$5007 on the contrary is lower by a factor of 2/3. Our model spectrum matches the position 1 spectrum rather well. The major exception is the \[O II\] $`\lambda `$3727 line, where the mismatch may be due to a repositioning error since the \[O II\] line was taken through a separate grating (see Hawley (1978)). Another difference is in the ratio between the two \[S II\] lines. Since the exact location of the observed spectrum is not known and since the observation and model pertain to two different regions (columns 1 and 2 respectively), the reason for the difference is not clear. The total strength of the two lines (relative to H$`\alpha `$) however is reproduced by the model. It is well known that spectra vary within an H II region. The significance here is that our model matches one of the spectra so well. This allows us to infer that the major reason for the differences in the two observed spectra are due to different orientations of the slit with respect to the highly stratified emitting interface.
## 4 Concluding Remarks
In this work, we have presented photoionization models for the emission from the photoevaporative interface between the ionized gas and the molecular cloud in blister H II regions. The density profile of the interface is a power-law which can be parameterized by the distance from the ionizing source to the ionization front and the density at the ionization front (Bertoldi & Draine (1996)). The values of these parameters depend upon the shape and strength of the ionizing continuum.
From the grid of models, which have systematically varying input parameters, we find that the H$`\alpha `$ peak emissivity is strongly dependent on the incident ionizing flux, increasing for higher values of flux. We also find that the \[O III\] emission is most sensitive to the type of star responsible for ionizing the interface. Stars cooler than O8 produce photoionized interfaces with virtually no \[O III\] emission. The low ionization \[S II\] emission is confined to a very sharp zone. We find that the peak \[S II\] emissivity is more sensitive than the total \[S II\] intensity to the sulphur abundance. The same holds true for the intermediate ionization \[N II\] line. It is worth noting that in the \[S II\] zone, the \[S II\] $`\lambda \lambda `$6716,6731 to H$`\alpha `$ ratio can get relatively high (e.g. it is $`>0.5`$ for the model shown in the bottom panel of Figure Modeling the Photoionized Interface in Blister HII Regions). Such high ratios are often considered to be a signature of shock-excited gas but our models show that they can also occur in photoionized interfaces.
We then considered narrow band HST data of the blister H II region M 16. The emission and ionization structure seen at high resolution was used to constrain the properties of the ionizing continuum and the sulphur abundance. We presented a specific model that reproduced the observations in detail. We found that the integrated spectrum predicted by the model matched one of two ground based spectra taken by Hawley (1978) at a nearby location. These data also validate our photoionization models. Our being able to match the observed emission and ionization structure with these models is strong evidence that blister H II regions can be described by a photoionized, photoevaporative flow.
The method we have presented here is important for the study of H II regions, both in our Galaxy and in other galaxies. In the case of sufficiently nearby objects (such as M 16) we can obtain spatially resolved spectra. Then, our current study indicates that we could use photoionization modeling to obtain elemental abundances without having to make assumptions about the structure of the emitting region. Carrying out such an exercise for a large sample of H II regions will be a crucial test for the applicability of the model for a range of conditions. We can apply our understanding of the emission from nearby H II regions to more distant H II regions (governed by the same physical mechanisms). Specifically, we can calculate photoionization models of photoevaporative flows in order to interpret spectra of these objects. One important class of distant emission nebulae which we may be able to study within this framework are the giant extra-galactic H II regions (GEHRs). It is promising that narrow band HST images of 30 Doradus, the nearest GEHR, show the optical emission concentrated in sharp photoevaporative interfaces, much like in M 16 (Scowen et al. (1998)). A direct application of our method to study the conditions in 30 Doradus would be an important step towards extrapolating to GEHRs in other galaxies.
RS thanks Jason Aufdenberg for discussions about stellar atmospheres and for providing the PHOENIX models used in our calculations. We thank the anonymous referee for several useful comments. Part of this work was completed while RS was a graduate student at Arizona State University. This work was partly supported by NASA/JPL contracts 959289 and 959329. |
warning/0001/cond-mat0001431.html | ar5iv | text | # The 𝑔-factors of discrete levels in nanoparticles
## Abstract
Spin-orbit scattering suppresses Zeeman splitting of individual energy levels in small metal particles. This suppression becomes significant when the spin-orbit scattering rate $`\tau _{\mathrm{so}}^1`$ is comparable with the quantum level spacing $`\delta `$. The $`g`$-factor exhibits mesoscopic fluctuations; at small $`\delta \tau _{\mathrm{so}}`$ it is distributed according to the Maxwell distribution. At $`\delta \tau _{\mathrm{so}}0`$ the average $`g`$-factor levels off at a small value $`g(l/L)^{1/2}`$ given by the ratio of the electron mean free path $`l`$ to the particle size $`L`$. On the contrary, in 2D quantum dots the $`g`$-factor is strongly enhanced by spin-orbit coupling.
In bulk metals the spin-orbit interaction is known to affect the $`g`$-factor of electrons. These corrections are relatively weak; even in such a heavy metal as gold, the $`g`$-factor is 2.1, and differs only slightly from the nominal value of 2. On the contrary, in recent experiments with Al nanoparticles the measured values of the $`g`$-factor were in the interval between 0.27 and 1.9. Later a similar effect was observed in gold nanoparticles, where the value $`g0.3`$ was measured. A systematic study of the effect of spin-orbit scattering on the $`g`$-factors of discrete levels was undertaken in Ref. . In particular, it was demonstrated that when 4% of gold impurities are added to Al particle, the $`g`$-factor drops from 2 to about 0.7.
In bulk materials the $`g`$-factor is conventionally determined from the electron spin resonance (ESR) data. ESR involves transitions between states of the electron continuum. On the contrary, the experiments with nanoparticles study the Zeeman splitting of individual electron levels $`ϵ_i`$:
$$ϵ_{i\sigma }(H)=ϵ_i\pm \frac{1}{2}g_i\mu _BH.$$
(1)
Here $`H`$ is the magnetic field, $`\mu _B=e\mathrm{}/2mc`$ is Bohr magneton; $`m`$ is the free electron mass. The splitting (1) is linear in $`H`$ as long as it is small compared to the quantum level spacing $`\delta `$. The $`g`$-factor determined by Eq. (1), in general, varies from level to level. In the absence of spin-orbit interaction $`g_i=2`$. Evidently the spin-orbit scattering affects the values of $`g_i`$.
In this paper we develop a theory of the $`g`$-factors of individual levels. Unlike its bulk value, the $`g`$-factor defined by Eq. (1) is very sensitive to even weak spin-orbit interaction. The reason is that the magnitude of the correction to $`g=2`$ value measured by ESR in the bulk, is determined by the comparison of spin-orbit interaction to the Fermi energy $`E_F`$, whereas the relevant energy scale for $`g`$-factors of individual levels is the quantum level spacing $`\delta E_F`$. Indeed, the spin-orbit interaction is usually described by the mean time of spin-orbit scattering $`\tau _{\mathrm{so}}`$. The scattering time should be compared with the time $`1/\delta `$ that electron travels along a closed trajectory corresponding to a quantum level. At $`\delta \tau _{\mathrm{so}}1`$ the electron spin flips very infrequently, and the effect of the spin-orbit scattering is weak. On the contrary, at $`\delta \tau _{\mathrm{so}}1`$ the spin flips the average of $`N=1/\delta \tau _{\mathrm{so}}`$ times during the electron motion along the closed trajectory. Thus the average spin in such a quantum state is significantly less than 1/2, and the response (1) to the magnetic field is strongly suppressed. The $`g`$-factor of this level can be estimated by assuming that electron spin-flips occur at random moments in time. The root-mean-square value of the spin is then $`1/\sqrt{N}`$, resulting in $`g_i\sqrt{\delta \tau _{\mathrm{so}}}`$.
The above estimate accounts only for the spin contribution to the $`g`$-factor. In fact, the level splitting in a magnetic field is determined by the total magnetic moment of the state
$$M_i=\mu _B^{}l_z_i+2\mu _Bs_z_i.$$
(2)
Here $`l_z`$ and $`s_z`$ are the operators of the angular momentum and spin, respectively; $`\mu _B^{}=e\mathrm{}/2m^{}c`$, with $`m^{}`$ being the effective mass of the electrons. The brackets $`\mathrm{}_i`$ denote the expectation value in $`i`$-th eigenstate of an electron confined to a nanoparticle in an infinitesimal magnetic field in $`z`$-direction. In a nanoparticle of a generic shape, the orbital levels are not degenerate. Then in the absence of the spin-orbit interaction the time reversal symmetry dictates $`l_z_i=0`$ at $`H0`$. It is important to note that in the presence of spin-orbit interaction $`l_z_i0`$, and the orbital motion contributes to the splitting of levels by magnetic field.
One can estimate the angular momentum of the electron as the (directed) area $`A`$ covered by its trajectory divided by the period of motion $`1/\delta `$, that is, $`l_z_im^{}A\delta `$. To find $`A`$ we notice that during time $`1/E_T`$ the electron travels across the grain and, therefore, covers the area $`L^2`$. (Here $`E_T=D/L^2`$ is the Thouless energy, $`L`$ is the grain size, and $`D`$ is the diffusion constant for electrons in the grain.) During the period of motion $`1/\delta `$ the electron bounces off the boundaries $`E_T/\delta `$ times. Since the direction of motion after each bounce is random, the total directed area is $`AL^2\sqrt{E_T/\delta }`$. Consequently, the root-mean-square angular momentum of each state is
$$l_z_im^{}L^2\sqrt{E_T\delta }\{\begin{array}{cc}\sqrt{l/L},\hfill & \text{3D}\text{,}\hfill \\ \sqrt{k_Fl},\hfill & \text{2D}\text{.}\hfill \end{array}$$
(3)
Here $`l`$ is the transport mean free path of electrons, and $`k_F`$ is their Fermi wavevector.
One can also see from Eq. (3) that in a three-dimensional (3D) nanoparticle the orbital contribution to the $`g`$-factor is small $`g_i^{\mathrm{orb}}\sqrt{l/L}2`$. However, in the case of strong spin-orbit scattering, $`\delta \tau _{\mathrm{so}}1`$, the above estimated spin contribution is also small, $`g_i^{\mathrm{sp}}\sqrt{\delta \tau _{\mathrm{so}}}2`$. Therefore in this regime both contributions may have to be taken into account.
We will now show that at $`\delta \tau _{\mathrm{so}}1`$ the $`g`$-factor of a level is a random quantity with the distribution function
$$P(g)=3\sqrt{\frac{6}{\pi }}\frac{g^2}{g^2^{3/2}}\mathrm{exp}\left(\frac{3g^2}{2g^2}\right),$$
(4)
where the averaging $`\mathrm{}`$ is performed either over an ensemble of 3D nanoparticles or over different levels in a single nanoparticle. Furthermore, we express $`g^2`$ in terms of two quantities, $`\tau _{\mathrm{so}}`$ and $`l`$, which can be measured independently:
$$g^2=\frac{6}{\pi \mathrm{}}\delta \tau _{\mathrm{so}}+\alpha \frac{l}{L}.$$
(5)
Here the dimensionless constant $`\alpha `$ is determined by the geometry of the nanoparticle; its exact value will be discussed later. Eqs. (4) and (5) are the main result of this paper.
We begin the study of the $`g`$-factors of individual levels by finding relation between $`g_i`$ and the matrix elements of the operator of magnetic moment $`M`$. Due to the time-reversal symmetry, the levels in the nanoparticle are degenerate, with the wavefunctions $`|\psi _i`$ and $`|T\psi _i`$, where $`T`$ stands for the time reversal operator. In a small magnetic field $`H`$ the levels are split by the perturbation $`MH`$. Using the standard method of degenerate perturbation theory, one can easily find the splitting in the form (1) with the $`g`$-factor
$$g_i=2\frac{|\stackrel{}{\mu }|}{\mu _B},$$
(6)
where the real vector $`\stackrel{}{\mu }`$ is defined as
$$\mu _x+i\mu _y=T\psi _i|M|\psi _i,\mu _z=\psi _i|M|\psi _i.$$
(7)
The distribution function $`p(\stackrel{}{\mu })`$ is, by definition,
$`p(\stackrel{}{\mu })={\displaystyle }{\displaystyle \frac{d^3\lambda }{(2\pi )^3}}e^{i\stackrel{}{\lambda }\stackrel{}{\mu }}\mathrm{exp}(i\lambda _x\mathrm{Re}T\psi _i|M|\psi _i`$ (8)
$`i\lambda _y\mathrm{Im}T\psi _i|M|\psi _ii\lambda _z\psi _i|M|\psi _i).`$ (9)
To perform the averaging in Eq. (9), we use the Random Matrix Theory (RMT) approach . Instead of the ensemble of nanoparticles with strong spin-orbit scattering we will consider an ensemble of symplectic matrices of size $`2N\times 2N`$ with $`N1`$. Then the eigenfunctions $`\psi |`$ and $`T\psi |`$ of the Hamiltonian are $`N`$-component spinors:
$$\psi |=(\{\varphi _k^{},\chi _k^{}\}),T\psi |=(\{\chi _k,\varphi _k\}).$$
(10)
The operator of magnetic moment $`M`$ is a hermitian matrix, which due to its time-reversal properties can be diagonalized to the form $`\mathrm{diag}\{M_1,M_1,\mathrm{},M_k,M_k,\mathrm{},M_N,M_N\}`$. In the basis of the eigenfunctions of $`M`$ the matrix elements (7) take the form:
$`\psi |M|\psi `$ $`=`$ $`{\displaystyle \underset{k=1}{\overset{N}{}}}M_k(|\varphi _k|^2|\chi _k|^2),`$ (11)
$`T\psi |M|\psi `$ $`=`$ $`2{\displaystyle \underset{k=1}{\overset{N}{}}}M_k\varphi _k\chi _k.`$ (12)
The advantage of the RMT approach is that the ensemble averaging in Eq. (9) is easily performed using the Porter-Thomas distribution of the matrix elements:
$$\mathrm{}=\underset{k=1}{\overset{N}{}}\frac{d^2\varphi _kd^2\chi _k}{(\pi /2N)^2}e^{2N(|\varphi _k|^2+|\chi _k|^2)}\mathrm{}.$$
(13)
The averaging in Eq (9) with the help of (11)–(13) reduces to the calculation of $`N`$ identical quadruple Gaussian integrals. The result has the form
$$p(\stackrel{}{\mu })=\frac{d^3\lambda }{(2\pi )^3}e^{i\stackrel{}{\lambda }\stackrel{}{\mu }}\underset{k=1}{\overset{N}{}}\left(1+\frac{|\stackrel{}{\lambda }|^2M_k^2}{(2N)^2}\right)^1.$$
(14)
In the limit of large $`N1`$, this integral becomes Gaussian too, and we find
$$p(\stackrel{}{\mu })=\left(\frac{2N^2}{\pi \mathrm{Tr}M^2}\right)^{3/2}\mathrm{exp}\left(\frac{2N^2}{\mathrm{Tr}M^2}|\stackrel{}{\mu }|^2\right).$$
(15)
Taking into account the relation (6), we now immediately find the distribution function of the $`g`$-factor in the form (4) with the mean square $`g`$ defined as
$$g^2=\frac{3}{\mu _B^2}\frac{\mathrm{Tr}M^2}{N^2}.$$
(16)
As a phenomenological theory, RMT enabled us to find the functional form of the distribution function; however, the width of the distribution (16) is now expressed in terms of a phenomenological parameter $`\mathrm{Tr}M^2/N^2`$. The relation between this parameter and the microscopic properties of the system cannot be established within the RMT. To do this, one has to find an observable quantity $`Q`$ which can be evaluated within both the phenomenological RMT and a microscopic theory.
We choose $`Q`$ to be the energy absorbed in unit time from a weak external ac magnetic field $`H(t)=H_0\mathrm{cos}\omega t`$. Such a field induces transitions between quantum levels in the grain, leading to the absorption of the energy in the grain. Within the RMT, the absorption can be found with the help of the Fermi golden rule as
$$Q=2\pi \underset{kk_f<p}{}\left|\frac{1}{2}H_0\psi _k|M|\psi _p\right|^2\delta (\omega +ϵ_kϵ_p)\omega .$$
(17)
The expression inside $`\mathrm{}`$ is the rate of absorption of quanta of radiation interacting with the system. The absorption occurs due to the transitions from occupied states with $`kk_f`$ to the empty states, $`p>k_f`$, and are induced by the term $`\frac{1}{2}H_0Me^{i\omega t}`$ in the corresponding coupling Hamiltonian.
In RMT the energy levels and the eigenfunctions are uncorrelated, i.e., the averaging over the energy levels and matrix elements in Eq. (17) can be performed independently. Also, at $`\omega \delta `$, one can neglect the correlations of the densities of states at energies separated by $`\omega `$, and we find
$$Q=\frac{\pi \omega ^2H_0^2}{2\delta ^2}|M_{kp}|^2.$$
(18)
It is important to note that contrary to the above calculation of the $`g`$-factors of individual levels, Eqs. (6) and (7), here the matrix elements $`M_{kp}`$ are between eigenstates with different energies. Upon averaging over the ensemble $`|M_{kp}|^2`$ becomes independent of $`k`$ and $`p`$. Its magnitude can be found by presenting the invariant $`\mathrm{Tr}M^2`$ as a sum of $`|M_{kp}|^2`$ over all $`2N`$ values of $`k`$ and $`p`$. Since the number of diagonal matrix elements $`2N`$ is small compared to the number of the off-diagonal ones, $`4N^22N`$, we conclude that $`|M_{kp}|^2=\mathrm{Tr}M^2/4N^2`$ at $`N1`$. Combining this relation with Eqs. (18) and (16), we find
$$g^2=\frac{24\delta ^2}{\pi \mathrm{}\omega ^2(\mu _BH_0)^2}Q.$$
(19)
It is noteworthy that this result, obtained within the RMT, contains no phenomenological parameters. It establishes a relation between the property of a single level, the $`g`$-factor, and a macroscopic quantity $`Q`$ insensitive to the effects of discreteness of levels.
To evaluate the absorption rate $`Q`$ for a given nanoparticle, it is convenient to express it as $`Q=\frac{1}{2}\omega A^{\prime \prime }H_0^2`$ in terms of the imaginary part $`A^{\prime \prime }`$ of the $`zz`$-component of the tensor of magnetic polarizability of the sample $`A_{ik}`$, defined as $`M_i=A_{ik}H_k`$, Ref. . Then the result for the mean square $`g`$-factor takes the form
$$g^2=\frac{12\delta ^2}{\pi \mathrm{}\mu _B^2}\frac{A^{\prime \prime }(\omega )}{\omega }.$$
(20)
It is well known that at $`\omega 0`$ the imaginary part of the polarizability vanishes as $`A^{\prime \prime }(\omega )\omega `$.
Since $`M=\mu _B^{}l_z+2\mu _Bs_z`$, one can distinguish between the orbital and spin contributions to the magnetic polarizability of the nanoparticle. The spin contribution has the form
$$A_s(\omega )=\frac{\mu _B^2\delta ^1}{1i\omega \tau _{\mathrm{so}}/2}.$$
(21)
Here the numerator is the usual static Pauli susceptibility of the electron gas of the nanoparticle, and the denominator accounts for the fact that spin correlations decay exponentially with the time constant $`\tau _{\mathrm{so}}/2`$. Substituting the imaginary part of the polarizability (21) into (20), we reproduce the first term of our main result (5).
The orbital contribution to the magnetic polarizability is due to the magnetic moment of the eddy currents generated in the sample by the ac magnetic field. For a particle of a shape symmetrical with respect to the rotations around the $`z`$-axis one can easily find
$$A^{\prime \prime }=\frac{\omega \sigma }{4c^2}\overline{\rho _{}^2}V,\omega 0.$$
(22)
Here $`\sigma `$ is the conductivity of the metal, $`V`$ is the volume of the nanoparticle, and $`\overline{\rho _{}^2}`$ is the “moment of inertia” of the grain, assuming unit density. For a spherical nanoparticle of radius $`L`$ the combination of Eqs. (20) and (22) reproduces the second term in Eq. (5) with $`\alpha =(6/5)(m/m^{})^2`$.
For the thin ring geometry Eqs. (20) and (22) result in $`g^2=3m^2L^2D\delta /\pi ^3\mathrm{}^3`$, where $`L`$ stands for the circumference of the ring. In the context of the persistent current problem the orbital effect of magnetic field on the splitting of the energy levels was studied earlier by Kravtsov and Zirnbauer. They used the non-linear $`\sigma `$-model techniques to solve the general problem of crossover from the symplectic ensemble to the orthogonal one. A special limiting case of that solution gave rise to a distribution function of level splittings, which reassuringly coincides with our result for the thin ring geometry. Phenomenologically the crossover problem was solved within RMT by Mehta and Pandey. Our approach enables one to express their phenomenological crossover parameter to physical observables without resorting to $`\sigma `$-model calculations.
In the case of weak spin-orbit interaction, $`\delta \tau _{\mathrm{so}}1`$, the correction to the average $`g`$-factor is small and can be found by perturbation theory. In the lowest order perturbation theory in spin-orbit coupling one finds
$$g_i=2\frac{2\mathrm{}\delta }{\pi \tau _{\mathrm{so}}}\underset{ji}{}\frac{1}{(ϵ_iϵ_j)^2}.$$
(23)
Assuming that the energy levels $`ϵ_j`$ are equidistant, the sum (23) was evaluated by Kawabata. However, in a disordered system the Wigner-Dyson statistics of the energy levels is a more realistic assumption. In this case one of the levels $`j`$ can be close to level $`i`$, resulting in a particularly large correction to the $`g`$-factor. Taking into consideration the fact that the level repulsion in the orthogonal ensemble suppresses the probability $`p_o`$ of two levels being very close, $`p_o(ϵ_iϵ_j)=\pi ^2|ϵ_iϵ_j|/6\delta ^2`$,, we find that the average value of the sum in Eq. (23) is logarithmically large:
$$g=2\frac{\pi }{6}\frac{\mathrm{}}{\delta \tau _{\mathrm{so}}}\mathrm{ln}\frac{\delta \tau _{\mathrm{so}}}{\mathrm{}}.$$
(24)
The logarithmic divergence in Eq. (24) was cut-off at the energy scale $`|ϵ_iϵ_j|\sqrt{\delta /\tau _{\mathrm{so}}}`$, because of the additional level repulsion caused by the weak spin-orbit coupling.
Our results are summarized in Fig. 1. The lower curve shows the dependence of the average $`g`$-factor on the strength of the spin-orbit scattering in the case of a diffusive nanoparticle. In the regime of $`1/\delta \tau _{\mathrm{so}}1`$ the behavior is described by Eq. (24), and at $`1/\delta \tau _{\mathrm{so}}1`$ the average $`g`$-factor drops in agreement with Eq. (5). In the latter regime $`g=(8/3\pi )^{1/2}g^2^{1/2}`$, as one can easily see from the distribution function (4).
The values of the mean free path are usually not well known in the experiment, and it is possible that $`l`$ exceeds the grain size $`L`$. The middle curve in Fig. 1 shows schematically the behavior of a ballistic nanoparticle, where all the scattering of electrons is due to the reflection from the boundaries. Qualitatively, in this case one can use the results for the diffusive case with the transport mean free path replaced by the size of the grain. It is then obvious from Eq. (5) that $`g`$ levels off at a value of order unity in the limit of strong spin-orbit scattering, as illustrated in the figure. One can conduct a quantitative study in a simple model of a spherical ballistic nanoparticle of radius $`L`$ with totally diffusive scattering off the boundaries. The magnetic polarizability can be calculated with the help of the Kubo formula, which relates it to the correlation function of angular momenta at different times. This correlation function can be evaluated by considering the classical electron trajectories inside the nanoparticle, in the spirit of Ref. . The exact result amounts to replacing $`l5L/8`$ in Eq. (5) and keeping $`\alpha `$ the same as in the case of a diffusive sphere. Taking into consideration the rapid decrease of the distribution function (4) of the $`g`$-factor at $`g^2g^2`$, one should conclude that the nanoparticles showing the values of $`g^2`$ well below the ballistic value $`g^2=(3/4)(m/m^{})^2`$ are most likely in the diffusive regime.
The top curve in Fig. 1 shows the behavior of the $`g`$-factor in the case of a 2D quantum dot in a perpendicular magnetic field. In accordance with Eq. (3), in the strong spin-orbit scattering case, $`1/\delta \tau _{\mathrm{so}}1`$, the orbital moment $`l_z`$ reaches a very large value $`\sqrt{k_Fl}`$. In experiments with quantum dots in GaAs heterostructures the effect of the orbital field should be further enhanced due to a small effective mass $`m^{}0.067m`$ of electrons, so one can expect to find $`g(m/m^{})\sqrt{k_Fl}`$. On the other hand, at weaker spin-orbit scattering, $`1/\delta \tau _{\mathrm{so}}1`$, the orbital enhancement of the $`g`$-factor is reduced, as the orbital motion does not affect the $`g`$-factor at all in the absence of spin-orbit coupling. The orbital contribution at $`1/\delta \tau _{\mathrm{so}}1`$ can be found within the first-order perturbation theory in the spin-orbit coupling, resulting in $`g(m/m^{})(\mathrm{}k_Fl/\delta \tau _{\mathrm{so}})^{1/2}`$. Thus, the $`g`$-factor of a 2D quantum dot is dominated by the effects of the orbital motion of electrons at $`1/\delta \tau _{\mathrm{so}}(k_Fl)^1(m^{}/m)^2`$, whereas at $`1/\delta \tau _{\mathrm{so}}(k_Fl)^1(m^{}/m)^2`$ values close to $`g=2`$ should be observed.
This work was supported by NSF under Grants DMR-9974435, DMR-9731756, and DMR-9812340. KM also acknowledges support by A.P. Sloan Foundation and the kind hospitality of the TPI at the University of Minnesota. The authors acknowledge the hospitality of ICTP in Trieste, Italy and Centre for Advanced Study in Oslo, Norway, where part of this work was performed. We are grateful to B.L. Altshuler, K.B. Efetov, S. Guéron, V.E. Kravtsov, and D.C. Ralph for useful discussions. |
warning/0001/astro-ph0001169.html | ar5iv | text | # Formation of Primordial Galaxies under UV background Radiation
## 1 INTRODUCTION
It has been widely accepted that the formation of the first generation of objects, say Pop III objects, is regulated by the cooling by primordial hydrogen molecules. A number of authors have explored the formation of Pop III objects in so-called ’dark age’ ($`z>5`$) by concentrating on the role of hydrogen molecules (Matsuda, Sato, & Takeda 1965; Yoneyama 1972; Hutchins 1976; Carlberg 1981; Palla, Salpeter, & Stahler 1983; Susa, Uehara, & Nishi 1996; Uehara et al. 1996; Annonis & Norman 1996; Tegmark et al. 1997; Nakamura & Umemura 1999). Also, even if the gas contains metals, the metallic line cooling is overwhelmed by the hydrogen/helium cooling at $`T>10^4`$K when the metallicity is lower than $`10^2Z_{}`$ (Böhringer & Hensler 1989). In fact, recently the metal abundance in the intergalactic space is inferred to be at a level of $`10^3Z_{}`$ (Cowie et al. 1995; Songaila & Cowie 1996; Songaila 1997; Cowie & Songaila 1998). Moreover, at lower temperature of $`T<10^4`$K, the $`\mathrm{H}_2`$ cooling is estimated to be still predominant as long as the metallicity is lower than $`10^2Z_{}`$, from the comparison of the cooling of the solar abundance gas (Spitzer 1978) with the maximal $`\mathrm{H}_2`$ cooling (Kang & Shapiro 1992). Hence, hydrogen molecules are likely to play a fundamental role on the formation of galaxies under the metal poor environment.
In addition, there seem to be situations where UV background radiation significantly influences the dynamical as well as thermal evolution of pregalactic clouds. At the epochs of $`z<5`$, the existence of UV background of $`10^{21\pm 0.5}\mathrm{erg}\mathrm{s}^1\mathrm{cm}^2\mathrm{str}^1\mathrm{Hz}^1`$ at the Lyman limit is inferred from so-called proximity effect of Ly$`\alpha `$ forest (Bajtlik, Duncan, & Ostriker 1988; Giallongo et al. 1996). Also, the Gunn-Peterson optical depths show that the intergalactic space was in fact highly ionized (Gunn & Peterson 1965; Schneider, Schmidt, & Gunn 1989, 1991). The UV background is attributed to the radiation from quasars and partially from young galaxies (e.g. Giallongo et al. 1996). Even in the dark age of the universe, say $`z>5`$, the first generation of objects as well as unseen proto-quasars might be external photoionization sources for subsequently collapsing objects. So far, many authors have considered the dynamical effects produced by the UV background radiation (Umemura & Ikeuchi 1984; Dekel & Rees 1987; Babul & Rees 1992; Efstathiou 1992; Chiba & Nath 1994; Thoul & Weinberg 1996; Quinn, Katz, & Efstathiou 1996; Navarro & Steinmetz 1997), to account for Ly$`\alpha `$ clouds or to reconcile the paradox that in the hierarchical clustering paradigm for galaxy formation, low mass galaxies are overproduced compared with observations (White & Frenk 1991; Kauffman, White, & Guiderdoni 1993; Cole et al. 1994). The simulations have been hitherto based upon the assumption that the medium is optically thin against UV photons. However, the gas clouds become optically thick as the gravitational collapse proceeds. The significance of the optical depth has been stressed by the accurate treatment on radiative transfer (Tajiri & Umemura 1998).
The primary effects of UV radiation are photoionization, UV heating, and photo-dissociation of $`\mathrm{H}_2`$ \[Haiman, Rees, & Loeb 1996 (hereafter HRL), 1997; Corbelli, Galli, & Palla 1997; Kepner, Babul, & Spergel 1997\]. HRL, with including radiation transfer effects in one-zone clouds, estimated the UV heating rate and the cooling rate under the assumption of the chemical equilibrium. They found that the cooling rate can exceed the heating rate at $`T\begin{array}{c}<\hfill \\ \hfill \end{array}10^4\mathrm{K}`$ for high density. However, basically the chemical reactions regarding $`\mathrm{H}_2`$ formation are not in equilibrium (e.g. Susa et al. 1998; Bertoldi & Draine 1996; Diaz-Miller, Franco & Shore 1998; Hollenbach & Tielens 1999). In particular, if ionization processes such as shock heating or photoionization take place, the chemical processes become out of equilibrium (e.g. Shapiro & Kang 1987). Thus, the non-equilibrium process for the formation of $`\mathrm{H}_2`$ should be involved, coupled with hydrodynamical process.
In this paper, we re-examine the thermal and dynamical evolution of pregalactic clouds under UV background radiation. In the present analysis, we solve properly the radiative transfer of diffuse UV photons and include non-equilibrium chemical reactions regarding $`\mathrm{H}_2`$ formation due to shock-ionization and photoionization. The goal is to elucidate the key physics on the effects of the UV background radiation upon the thermal evolution of pregalactic clouds. For the purpose, we make the dynamical model as simple as possible. A cosmological density perturbation far beyond the Jeans scale forms a flat pancake-like disk. Although the pancake formation is originally studied by Zel’dovich (1970) in the context of the adiabatic fluctuations in baryon or hot dark matter-dominated universes, recent numerical simulations have shown that such pancake structures also emerge in a CDM cosmology (e.g., Cen et al. 1994). Thus, the pancakes are thought to be a ubiquitous feature in gravitational instability scenarios. Here, we consider pancakes in the plane-parallel symmetry. Also, previous 1D/2D/3D simulations on the pancake collapse in a dark matter-dominated universe show that the pancake is finally dominated by baryons after the caustics (Shapiro, Struck-Marcell, & Melott 1983; Bond et al.1984; Shapiro & Struck-Marcell 1985; Yuan, Centrella, & Norman 1991; Umemura 1993; Cen et al. 1994; Miralda-Escude & Rees 1994; Hernquist et al. 1995; Zhang, Anninos, & Norman 1995; Anninos & Norman 1996). The presence of dark matter increases the shock velocity of the falling matter into the caustics. According to previous analyses (e.g. Shapiro & Kang 1987), such shocks change the thermal evolution in the absence of a UV background when the temperature exceeds $`10^4`$ K since the temperature controls H<sub>2</sub> fraction in the postshock regions. However, under UV background radiation, the difference of shock velocity will not alter the results significantly, because the sheets are initially ionized by UV radiation and therefore quickly heated up to $`10^4`$K. Thus, the thermal evolution of shocked region will be rather similar for any infall velocity, regardless of the shock temperature. Thus, in order to highlight the relevant physics of thermal processes, we deliberately exclude the dark matter contribution. Nonetheless, some dynamical effects are anticipated from dark matter, e.g. Jeans instability. They are discussed later in the light of the present results.
In section 2, we describe the basic equations and initial conditions. In section 3, the numerical results are presented. In section 4, we make the physical interpretation of numerical results and present the condition on the cooling of a collapsing pancake. Section 5 is devoted to the implications for galaxy formation under UV background radiation. In the last section, we summarize the results.
## 2 Formulation
### 2.1 Basic Equations
In this section, we give the basic equations for hydrodynamical calculations. We assume the plane-parallel symmetry throughout this paper. A set of the ordinary hydrodynamical equations is given as below.
$`{\displaystyle \frac{d\left(1/\rho \right)}{dt}}{\displaystyle \frac{u}{m}}`$ $`=`$ $`0,`$ (1)
$`{\displaystyle \frac{du}{dt}}+{\displaystyle \frac{p}{m}}`$ $`=`$ $`g,`$ (2)
$`{\displaystyle \frac{dE}{dt}}+{\displaystyle \frac{\left(up\right)}{m}}`$ $`=`$ $`ug\mathrm{\Lambda }(\rho ,T,y_i),`$ (3)
where $`\rho ,p,u`$, and $`z`$ denote the density, the thermal pressure, the velocity, and the length measured from the mid-plane, respectively, and $`E`$ is the total energy per unit mass,
$`E`$ $`=`$ $`{\displaystyle \frac{p}{\rho \left(\gamma 1\right)}}+{\displaystyle \frac{u^2}{2}},`$ (4)
$`m(z)`$ is the column density,
$`m(z)`$ $``$ $`{\displaystyle _0^z}\rho (z)𝑑z,`$ (5)
and $`g`$ is the gravitational acceleration,
$`g`$ $``$ $`2\pi Gm(z).`$ (6)
The symbol $`\mathrm{\Lambda }(\rho ,T,y_i)`$ denotes the cooling rate minus heating rate per unit mass due to the radiative and chemical processes of i-th species. We take H<sub>2</sub> cooling rate from Hollenbach & Mckee (1979), and the other atomic cooling rates are the same as those in Shapiro & Kang (1987), except the cooling rate concerning on helium. Photoheating rate is calculated by taking into account the radiation transfer effects. The formulation is given in Appendix A. The equation of state is assumed as $`p=\rho kT/\mu _M`$, where $`k`$ denote the Boltzman constant, and $`\mu _M`$ is the mean molecular weight of the five species discussed below in this section. The hydrodynamical equations are solved by the Piecewise Parabolic Method (PPM) described in Colella & Woodward (1984). PPM is one of the accurate methods to resolve the strong shock front. We take typically 800 spatial Lagrange grids. The hydrodynamical routine is tested by the Sod’s problem. This code resolves the shock front typically by a few mesh.
In the equations (1), (2), and (3), the cosmic expansion is neglected. It is because we have considered the massive clouds far beyond the Jeans scale ($`\lambda _J10[(1+z)/10]^{1.5}\mathrm{kpc}`$) and also postulated that the initial stage is close to the maximum expansion for a density fluctuation. The effects of the cosmic expansion could be important for a density perturbation near the Jeans scale, since the temperature is raised up to $`10^4`$ K by UV background radiation in spite of the expansion and consequently the dynamics could be significantly influenced by the UV heating (e.g. Umemura & Ikeuchi 1984). However, far beyond the Jeans scale, the thermal pressure of gas around $`10^4`$ K anyway does not affect the dynamical evolution until the maximum expansion as shown by previous calculations (e.g. Thoul & Weinberg 1996). Therefore, we attempt to pursue the history after the maximum expansion.
The hydrodynamical equations are coupled with the non-equilibrium rate equations for chemical reactions:
$`{\displaystyle \frac{dy_i}{dt}}={\displaystyle \underset{j}{}}k_jy_j+n{\displaystyle \underset{k,l}{}}k_{kl}y_ky_l+n^2{\displaystyle \underset{m,n,s}{}}k_{mns}y_my_ny_s,`$ (7)
where $`y_i`$ is the fraction of i-th species, $`y_in_i/n`$, with $`n`$ being the number density of hydrogen nuclei, and $`k`$’s are the coefficients of reaction rates (see Table 1). In the present calculations, we take into account six species e, H, H<sup>+</sup>, H<sup>-</sup>, H<sub>2</sub>, and H$`{}_{}{}^{+}{}_{2}{}^{}`$. Here, we neglect helium. This is because the helium lines are not an effective coolant for lower temperature ($`T\begin{array}{c}<\hfill \\ \hfill \end{array}10^4`$ K) in which we are here especially interested. Nonetheless, some of helium lines including recombination lines may affect the ionization structure, because they are energetic enough to ionize HI. However, according to Osterborck (1989), the difference between the ionization structure of pure hydrogen and hydrogen + helium gas for 40000 K black body radiation is small. In the case of the power-law spectrum of UV background radiation which we consider, the difference could be larger, because the higher energy photons contribute more to the ionization of helium compared to the case of black body radiation. In fact, in the recent calculation by Abel & Haehnelt (1999), in which the radiation transfer effects of helium recombination photons are taken into account, the temperature of the clouds could differ by a factor of 2. Although we have dismissed helium to focus upon elucidating the key physics, we should keep in mind that helium could be important in highly quantitative arguments.
The photoionization and heating processes due to the external UV radiation are pursued by solving the frequency-dependent radiative transfer equation for hydrogen,
$$\mu \frac{dI_\nu }{dz}=\kappa _\nu I_\nu +\eta _\nu ,$$
(8)
where $`\mu `$, $`\kappa _\nu `$, and $`\eta _\nu `$ denote the direction cosine, the absorption coefficient, and the emissivity, respectively. The frequency-integration in the transfer equation can be done analytically for the continuum, while the recombination line near the Lyman limit is separately treated by solving transfer equation with a source term which comes from the recombination. The details are described in Appendix A. Equation (8) assumes steadiness of the radiation fields. When a neutral cloud is exposed to background ionizing radiation, ionized regions spread inward. In the ionized regions, the radiation propagates basically at the light speed, which is typically $`10^3`$ times larger than the hydrodynamical velocity of the system. Thus, as for the radiation fields, it is sufficient to solve the steady radiative transfer equation. But, the ionization front propagates with a different speed, which is determined by the balance between the number of neutral atoms flowing through the front per second and the corresponding number of ionizing photons reaching the front. Resultantly, the speed of ionization front is much lower than the light speed, typically 500 km s<sup>-1</sup> $`(n/10^2\mathrm{cm}^3)^1`$ for UV background of $`10^{21}\mathrm{erg}\mathrm{s}^1\mathrm{cm}^2\mathrm{str}^1\mathrm{Hz}^1`$. In order to capture the ionization front propagation, we take the following numerical procedures. First, the steady radiative transfer equation (8) is solved, where typically 800 spatial grids are used, which number is equivalent to the number of Lagrange meshes assigned for hydrodynamics, and 40 angular grids are adopted. We have checked that the resultant accuracy is not significantly changed if we use the larger number of angular grids. The integration of equation (8) is performed with the summation of the exact formal solution in every grid (Stone, Miharas, & Norman 1992). This method is accurate, even if the optical depth of each grid can exceeds unity. Also, the validity of this method is enhanced by the fact that the ionization front is not sharply edged, but fairly dull for a power law of UV radiation (Tajiri & Umemura 1998). For the photons which cause H<sub>2</sub> dissociation through the Solomon process (e.g. HRL1996), we employ the self-shielding function given by Draine & Bertoldi (1996), instead of directly solving the radiation transfer equation. Next, the rate equations including photoionization are solved implicitly for a time step which is determined as bellow. Finally, the hydrodynamical equations with the thermal equation is integrated explicitly.
The time step of the calculation $`\mathrm{\Delta }t`$ is taken as follows:
$`\mathrm{\Delta }t`$ $`=`$ $`\mathrm{min}(0.99t^{\mathrm{hydro}},0.8t^{\mathrm{grav}},0.3t^{\mathrm{cool}},0.3t^{\mathrm{heat}}).`$ (9)
Here $`t^{\mathrm{hydro}}`$ represents the hydrodynamical time, and $`0.99t^{\mathrm{hydro}}`$ is employed so that the Courant condition is satisfied. $`t^{\mathrm{grav}}`$ denotes the time scale of gravity, which is defined as $`t^{\mathrm{grav}}\sqrt{p/\gamma \rho }/|g|`$, where $`p`$, $`\gamma `$, $`\rho `$, and $`g`$ represents the pressure, the adiabatic index, the density, and the gravitational acceleration, respectively. $`t^{\mathrm{cool}}`$ and $`t^{\mathrm{heat}}`$ represent local cooling time, and heating time scale, respectively. With these procedures, we have checked the propagation speed of the ionization front defined by the point $`y_{\mathrm{HI}}=0.1`$. In all of our simulations, the I-front propagation speeds are less than 1 percent of the speed of light. Thus, the assumption of the steady radiation transfer is valid.
### 2.2 Initial Conditions
We assume the initial density distributions to have a cosine profile;
$$n_{\mathrm{ini}}(z)=\overline{n}_{\mathrm{ini}}\left[1+\epsilon \mathrm{cos}\left(\pi z/\lambda \right)\right],$$
(10)
where $`\overline{n}_{\mathrm{ini}}`$ and $`\lambda `$ are the initial mean number density and the thickness of the sheet (See Fig.1), and $`\epsilon `$ denotes the density contrast, which is set to be 0.5 throughout this paper. The ranges of $`\overline{n}_{\mathrm{ini}}`$ and $`\lambda `$ which we consider are respectively $`10^6\mathrm{cm}^3\overline{n}_{\mathrm{ini}}5\times 10^2\mathrm{cm}^3`$ and $`0.3\mathrm{kpc}\lambda 3\mathrm{M}\mathrm{p}\mathrm{c}`$. The initial velocity is null for every mass layer, because the initial stage is implicitly assumed to be close to the maximum expansion stage of a density fluctuation. In general, an overdense region at the maximum expansion is surrounded by an underdense region which is still expanding. Hence, if we introduce the initial velocity distributions more realistically, the further mass may accrete from the underdense region after the pancake collapse. However, the envelope does not seem to contribute to the self-shielding because of its low density, and thus the self-shielding is determined by the column density of the first collapsed sheet. Hence, we neglect the surrounding expanding underdense region, and model only the overdense region. Moreover, the velocity gradient may exist in a collapsing overdense region, and then leads to the delay of mass accretion from outer regions. The degree of self-shielding is determined by the total recombination number per unit time ($`N_{\mathrm{rec}}`$) in the volume against the UV photon number per unit time ($`N_{\mathrm{UV}}`$) from the boundary. If the envelope has a density distribution in proportion to $`r^x`$, $`N_{\mathrm{rec}}`$ is proportional to $`r^{2x+1}`$, whereas $`N_{\mathrm{UV}}`$ is constant, as far as the envelope undergoes nearly sheet-like collapse. Thus, an envelope with a distribution steeper than $`r^{1/2}`$ would not contribute to the self-shielding. In other words, the inner regions which collapse nearly simultaneously contribute mainly to the self-shielding. Although there are such qualitative expectations, we cannot assess accurately the effects by the initial velocity gradient and the cosmological expansion, because we have not included them. They should be quantitatively investigated in the future analyses.
The initial temperature and chemical composition are given respectively by the thermal equilibrium and by the ionization equilibrium. Also, the fractions of H<sub>2</sub>, H<sup>-</sup> and H$`{}_{}{}^{+}{}_{2}{}^{}`$ are initially determined by the chemical equilibrium. When determining the initial equilibrium state, H<sub>2</sub> cooling rate is ignored, because we set the initial condition so that the UV heating should overwhelm the H<sub>2</sub> cooling. It will be shown below that this condition can be satisfied even if the cloud is almost neutral.
The incident radiation intensity, $`I_\nu ^{\mathrm{in}}`$, is assumed to have a power low spectrum as
$$I_\nu ^{\mathrm{in}}=I_{21}(\nu /\nu _L)^\alpha 10^{21}\mathrm{erg}\mathrm{s}^1\mathrm{cm}^2\mathrm{str}^1\mathrm{Hz}^1,$$
(11)
where $`\nu _L`$ is the frequency at the Lyman limit. The observations require the diffuse UV radiation to be at a level of $`I_{21}=10^{\pm 0.5}`$ at $`z=1.74.1`$ (Bajtlik, Duncan, & Ostriker 1988; Giallongo et al. 1996). At higher redshifts, we have no indication for $`I_{21}`$. Here, we assume $`I_{21}=0.1`$, and the scaling to $`I_\nu ^{\mathrm{in}}`$ is argued. As for $`\alpha `$, two typical cases are considered, i.e. $`\alpha =1`$ or 5, which are the same choice as Thoul & Weinberg (1996). The former represents a quasar-like spectrum, and the latter resembles a spectrum of star-forming galaxies around the Lyman limit frequency.
## 3 Thermal and Dynamical Evolution
In this section, we show the characteristic behaviors of the thermal/dynamical evolution resulting from numerical runs, and elucidate the dependence of the final states upon the initial states.
### 3.1 Hard Spectrum Case: $`\alpha =1`$
Figures 2 and 3 show the time evolution of spatial distributions of various physical quantities in the case of $`\alpha =1`$. Figure 2 represents the case with $`\lambda =100`$ kpc and $`\overline{n}_{\mathrm{ini}}=10^2`$ cm<sup>-3</sup>. In this figure, each horizontal axis represents the Lagrange coordinate, normalized by the half total surface density, $`m_0`$. The thick line shows the distribution at $`t=1.28/\sqrt{4\pi G\overline{\rho }_{\mathrm{ini}}}`$, where $`\overline{\rho }_{\mathrm{ini}}=m_p\overline{n}_{\mathrm{ini}}`$ with $`m_p`$ being the proton mass. The calculations are terminated there. Figure 2 shows that in the course of the collapse, a thin cold sheet quickly forms due to the efficient molecular cooling, which is confined by a shock-heated high temperature layer. The lower panels in Figure 2 show the corresponding HI and H<sub>2</sub> distributions. Only an outer envelope ($`m/m_0\begin{array}{c}>\hfill \\ \hfill \end{array}0.9`$) is photoionized and the rest parts of the sheet are self-shielded against UV radiation. As a result, the sheet is mostly neutral throughout the dynamical evolution, excepting the collisionally ionized postshock regions. In the early phase, UV heating exceeds the H<sub>2</sub> line cooling even in the almost neutral regions. The H<sub>2</sub> cooling becomes efficient as the collapse proceeds to a further degree, and finally the temperature drops down to $`100`$ K. Then, the collisional dissociation of hydrogen molecules becomes ineffective, and thus the fraction of hydrogen molecules at $`m=0`$ continues to increase.
Figure 3 represents the case with lower density and larger size, i.e., $`\overline{n}_{\mathrm{ini}}=5\times 10^5\mathrm{cm}^3`$ and $`\lambda =1`$ Mpc. The sheet also collapses because the initial thickness is larger than the initial Jeans length. However, the central temperature does not drop below $`10^4`$ K, and the resultant H<sub>2</sub> fraction is too low to overwhelm the UV heating. Why the evolution results in the different final states from the case shown in Figure 2 is basically understood by the difference of the degree of self-shielding against the external UV radiation. In the previous case, the most regions are so strongly shielded from the external UV that abundant H<sub>2</sub> form and the cooling rate due to H<sub>2</sub> becomes predominant. Contrastively, in the present case, the external UV permeates the inner regions to a considerable degree, so that UV heating overwhelms H<sub>2</sub> cooling everywhere until the collapse stops due to the thermal pressure.
### 3.2 Soft Spectrum Case: $`\alpha =5`$
Figure 4 shows the case of a softer spectrum, $`\alpha =5`$. The other initial parameters are the same as Figure 3. Compared to the previous hard spectrum cases, the inner parts of the sheet are quickly self-shielded as the collapse proceeds. This is due to the fact that there are fewer high energy photons which penetrate the deep inside the sheet owing to the smaller ionization cross-section (see e.g. Tajiri & Umemura 1998). As a result, the photoionization rate as well as the UV heating rate damps strongly in the inner regions, so that a cold layer ($`T<10^4`$ K) due to H<sub>2</sub> cooling emerges eventually. The self-shielded regions grow thicker and the the fraction of H<sub>2</sub> goes up to a level of $`10^3`$ similar to the case of Figure 2.
### 3.3 Summary of the Dependence on Initial States
In Figure 5, the final physical states of the sheets for $`\alpha =1`$ at $`t=1.28/\sqrt{4\pi G\rho _{\mathrm{ini}}}`$ are summarized on the initial density-to-size diagram. Depending upon the initial thickness $`\lambda `$ and mean density $`\overline{n}_{\mathrm{ini}}`$, the sheets evolve in different ways. Filled circles in Figure 5 denote the initial conditions on which the clouds eventually cool below $`5\times 10^3`$ K during the collapse. Open circles denote the ones on which the clouds cannot cool ($`T>5\times 10^3`$ K). The boundary of the evolutionary bifurcation is well fitted by a simple power low relation as
$`\lambda _{\mathrm{cool}}=1.1\mathrm{Mpc}\left({\displaystyle \frac{\overline{n}_{\mathrm{ini}}}{1.0\times 10^4\mathrm{cm}^3}}\right)^{0.8}.`$ (12)
The small open squares, open triangles, and filled triangles in Figure 5 trace the evolutionary paths for three typical parameters. In the paths, the time-dependent thickness ($`H`$) is defined by the thickness within which 80% of the total mass of the sheet is contained and the time-dependent mean density ($`\overline{n}`$) is defined by the spatially averaged density over $`H`$;
$`H`$ $``$ $`2{\displaystyle _0^{0.9m_0}}{\displaystyle \frac{1}{\rho }}𝑑m,`$ (13)
$`\overline{n}`$ $``$ $`{\displaystyle \frac{2}{H}}{\displaystyle _0^H}n𝑑z.`$ (14)
The open squares and open triangles are respectively the results for the initial parameters of $`(\lambda ,\overline{n}_{\mathrm{ini}})=(1\mathrm{M}\mathrm{p}\mathrm{c},5\times 10^5\mathrm{cm}^3)`$ and $`(100\mathrm{k}\mathrm{p}\mathrm{c},1\times 10^3\mathrm{cm}^3)`$. For such initial conditions, the sheet shrinks due to Jeans instability \[$`\lambda >\lambda _J(T=10^4\mathrm{K})`$\]. However, the collapse ceases by the thermal pressure before it intersects the critical line marked by $`\lambda _{\mathrm{shield}}`$, above which the sheet can be self-shielded and cool down due to formed H<sub>2</sub>. (The further details on $`\lambda _{\mathrm{shield}}`$ are discussed in the next section.) On the other hand, the filled triangles, corresponding to the run with $`(\lambda ,\overline{n}_{\mathrm{ini}})=(100\mathrm{k}\mathrm{p}\mathrm{c},1\times 10^2\mathrm{cm}^3)`$, come into the regions $`\lambda >\lambda _{\mathrm{shield}}`$ before the thermal pressure prevents the sheet from collapsing. Consequently, this sheet is shielded and cools below $`T5\times 10^3K`$.
Figure 6 shows the dependence on the initial states for $`\alpha =5`$. Because of the smaller number of runs, it is somewhat hard to divide clearly the parameter space. If we dare to draw the boundary, it is proportional to $`\overline{n}_{\mathrm{ini}}^{0.66}`$, which is numerically
$`\lambda _{\mathrm{cool}}=6.0\times 10^2\mathrm{kpc}\left({\displaystyle \frac{\overline{n}_{\mathrm{ini}}}{1.0\times 10^4\mathrm{cm}^3}}\right)^{0.66}.`$ (15)
This criterion is slightly different from the previous case, in the sense that $`\lambda _{\mathrm{cool}}`$ for $`\alpha =5`$ does not cross the intersection point of $`\lambda _J(T=10^4\mathrm{K})=\lambda _{\mathrm{shield}}`$, although $`\lambda _{\mathrm{cool}}`$ for $`\alpha =1`$ does.
## 4 Physical Conditions for H<sub>2</sub> Cooling under UV background
In this section, we attempt to understand the underlying physics of the present numerical results. To begin with, we introduce a scale, $`\lambda _{\mathrm{shield}}`$, which characterizes the relative efficiency of H<sub>2</sub> cooling to UV heating. Then, coupled with some physical arguments, we try to estimate the boundary of the bifurcation, $`\lambda _{\mathrm{cool}}`$, which has been derived by the present numerical calculations.
As a measure of the penetration of UV photons into the cloud, a shielding length, $`\lambda _{\mathrm{shield}}`$, is defined by the balance between the UV heating rate and the H<sub>2</sub> cooling rate, which are respectively
$`\mathrm{Heating}\mathrm{Rate}`$ $`=`$ $`n(z)y_{\mathrm{HI}}\mathrm{\Gamma }_{\mathrm{HI}}\left(\tau _{\nu _L,v}\left(z\right)\right),`$ (16)
$`\mathrm{Cooling}\mathrm{Rate}`$ $`=`$ $`n(z)^2\mathrm{\Lambda }_{\mathrm{H}_2}(y_{H_2}(z),T(z)),`$ (17)
where $`\mathrm{\Gamma }_{\mathrm{HI}}`$ is the UV heating rate per particle and $`\mathrm{\Lambda }_{\mathrm{H}_2}`$ denotes the H<sub>2</sub> cooling function. As shown in Appendix A \[eq.(A12)\], in an optically thick regime, the UV heating rate is expressed by a power law of the optical depth at the Lyman limit, not by an exponential law. This is again because higher energy photons have the smaller ionization cross-section. The optical depth at the Lyman limit measured from the outer boundary ($`z=z_{\mathrm{out}}`$) is
$`\tau _{\nu _L,v}(z)`$ $``$ $`{\displaystyle _z^{z_{\mathrm{out}}}}𝑑zny_{\mathrm{HI}}\sigma _{\nu _L},`$ (18)
where $`\sigma _\nu `$ is the ionization cross-section. If the slab is almost neutral,
$`\tau _{\nu _L,v}(0)`$ $``$ $`{\displaystyle _0^{z_{\mathrm{out}}}}𝑑zn\sigma _{\nu _L},`$ (19)
$`=`$ $`\overline{n}_{\mathrm{ini}}\sigma _{\nu _L}\lambda /2.`$ (20)
Equating (16) with (17) at the midplane of the slab ($`z=0`$), and also using equations (20) and (A12), we can assess $`\lambda _{\mathrm{shield}}`$ as
$$\lambda _{\mathrm{shield}}\frac{2}{\overline{n}_{\mathrm{ini}}\sigma _{\nu _L}}\left(\frac{2\pi I_{\nu _L}^{\mathrm{in}}\sigma _{\nu _L}\nu _L}{3\overline{n}_{\mathrm{ini}}\mathrm{\Lambda }_{\mathrm{H}_2}}\frac{\mathrm{\Gamma }\left(\beta \right)}{1+\beta }\right)^{\frac{1}{\beta }}\overline{n}_{\mathrm{ini}}^{11/\beta }$$
(21)
where $`\beta 1+\left(\alpha 1\right)/3`$. The $`\lambda _{\mathrm{shield}}`$ is plotted in Figures 5 and 6 by long-dashed lines. The cooling rate has been assumed to be $`\mathrm{\Lambda }_{\mathrm{H}_2}=10^{26}\mathrm{cm}^3\mathrm{s}^1`$, which is a maximal value of H<sub>2</sub> cooling for the primordial gas that is once heated up to $`T\begin{array}{c}>\hfill \\ \hfill \end{array}10^4\mathrm{K}`$ (Shapiro & Kang 1987; Susa et al. 1998).
The obtained $`\lambda _{\mathrm{shield}}`$ represents the degree of self-shielding against the external UV at the initial stages: (1) For $`\lambda >\lambda _{\mathrm{shield}}(\overline{n}_{\mathrm{ini}})`$, the sheet is promptly self-shielded enough to be cooled by the H<sub>2</sub>. (2) For $`\lambda <\lambda _{\mathrm{shield}}(\overline{n}_{\mathrm{ini}})`$, the sheet is not self-shielded initially, so that the gas is heated up to $`T\begin{array}{c}>\hfill \\ \hfill \end{array}10^4`$ K due to predominant UV heating. A similar argument with $`\lambda _{\mathrm{shield}}`$ is also applicable to each dynamical stage. In the case, we should just replace the density $`\overline{n}_{\mathrm{ini}}`$ by $`\overline{n}`$, and $`\lambda `$ by $`H`$, both of which are defined in section 3. Then, we interpret $`\lambda _{\mathrm{shield}}`$ as the boundary beyond which the sheet is quickly self-shielded and consequently cools down owing to the efficient formation of H<sub>2</sub>. In fact, the clouds denoted by filled circles in Figure 5 satisfy the condition $`H>\lambda _{\mathrm{shield}}`$ before the collapse is abruptly decelerated by approaching $`\lambda _J(T=10^4\mathrm{K})`$. Eventually, the clouds cool down below $`5\times 10^3`$ K. On the other hand, the clouds denoted by open circles above $`\lambda _J`$ in Figure 5 result in warm sheets of $`10^4`$ K, because the clouds first meet $`\lambda _J(T=10^4\mathrm{K})`$ before intersect $`\lambda _{\mathrm{shield}}`$. If we assume that the sheet clouds evolve keeping the relation $`\overline{n}H=\mathrm{const}.`$ as anticipated in the ideal sheet collapse, the boundary of the bifurcation in the space $`(\overline{n}_{\mathrm{ini}},\lambda )`$ is expected to be a line which satisfies $`\overline{n}_{\mathrm{ini}}\lambda =\mathrm{const}.`$ and meets the intersection point of $`\lambda _{\mathrm{shield}}=\lambda _J`$. This prediction, $`\lambda _{\mathrm{cool}}\overline{n}_{\mathrm{ini}}^1`$, is close to the numerically obtained boundary, $`\lambda _{\mathrm{cool}}\overline{n}_{\mathrm{ini}}^{0.8}`$. The small difference of the dependence may come from the fact that the simulations contain the spatial structure of density and temperature, and therefore partial self-shielding.
We remark that the Jeans length $`\lambda _J`$ is maximally scaled by $`\sqrt{\mathrm{\Omega }_\mathrm{B}/\mathrm{\Omega }_0}`$ in the presence of dark matter, where $`\mathrm{\Omega }_\mathrm{B}`$ is the baryon density parameter and $`\mathrm{\Omega }_0`$ is the total density parameter. However, the baryonic component dominates the gravity of the sheets in the final phase of the collapse, although dark matter component does initially. Thus, what actually happens is the intermediate of these two extreme cases. Although calculations that include dark matter are necessary to evaluate the effect quantitatively, we leave them elsewhere.
## 5 Implications for Galaxy Formation
Based upon the present numerical results, we consider the context of the galaxy formation under UV background radiation. First of all, the cooling by atomic processes of primordial gas is essential for the formation of H<sub>2</sub> molecules which control the star formation in primordial objects. The elaborate analyses by Rees & Ostriker (1977) and also by Blumenthal et al. (1984) show that the cooling mass is roughly constant almost regardless of the virial temperature and density, which is $`10^{12}M_{}`$. If there is UV background radiation, the cooling mass could alter because the line cooling by H and He at $`\begin{array}{c}<\hfill \\ \hfill \end{array}10^5`$K is seriously reduced (e.g. Thoul & Weinberg 1996). The mass, however, is basically determined by the cooling mechanisms at $`\begin{array}{c}>\hfill \\ \hfill \end{array}10^5`$K. At such temperature, the cooling in photoionized gas is dominated by collisional ionization, radiative recombination, and thermal bremsstrahlung, and possibly the Compton cooling at high redshifts. These mechanisms may potentially cool the clouds with $`10^{12}M_{}`$ down to several $`10^4`$K. In fact, it is shown by numerical calculations that the clouds with $`10^{12}M_{}`$ can cool and collapse under UV background (Umemura & Ikeuchi 1984). Thus, we set here the upper mass of primordial galaxies to be $`10^{12}M_{}`$. Then, the evolution of pregalactic clouds under UV background radiation is discriminated into four categories; (a) promptly self-shielded clouds, (b) starburst pancakes, (c) retarded star-forming galaxies, and (d) expanding clouds. In Figure 7, the parameter regions corresponding to the four categories are shown for $`\alpha =1`$, with lines of equal mass, say, $`10^8M_{}`$ and $`10^{12}M_{}`$. The upper abscissa is the collapse redshifts, $`z_c`$, if the initial stage is assumed to be at the maximum expansion of a spherical top-hat density fluctuation. $`\mathrm{\Omega }_0=1`$, $`h=0.5`$, and $`\mathrm{\Omega }_\mathrm{B}h^2=0.02`$ are assumed in order to interpret the maximal expansion density into the collapse redshift $`z_c`$, where $`h`$ is the present Hubble constant in units of $`100\mathrm{k}\mathrm{m}\mathrm{s}^1\mathrm{Mpc}^1`$. In this figure, 1$`\sigma `$, 2$`\sigma `$, and 3$`\sigma `$ density fluctuations expected in a standard CDM are also shown. Some further details for each category are discussed in the following, restricting ourselves to the case $`\alpha =1`$.
### 5.1 Promptly Shielded Clouds
The evolution of promptly self-shielded clouds above $`\lambda _{\mathrm{shield}}`$ is virtually equivalent to the evolution under no UV background radiation. The collapse redshifts of such clouds are expected to be $`z_c\begin{array}{c}>\hfill \\ \hfill \end{array}10`$. The formation of primordial objects under no UV background has been hitherto extensively studied by numerous authors (Matsuda, Sato, & Takeda 1965; Yoneyama 1972; Hutchins 1976; Palla, Salpeter, & Stahler 1983; Susa, Uehara, & Nishi 1996; Annonis & Norman 1996; Tegmark et al. 1997). As a result, we can expect the first generation of objects in the mass range of $`10^6M_{}`$ at $`z_c=15`$ down toward $`10^3M_{}`$ at $`z_c200`$ (Tegmark et al. 1997).
### 5.2 Initial Starbursts in Pancakes
The region (b) in Figure 7 results in cold pancakes due to H<sub>2</sub> cooling, in which initial starbursts may take place. The instability of a shock-compressed layer has been discussed by several authors (Elmegreen & Elmegreen 1978; Vishniac 1983; Lubow & Pringle 1993; Whitworth et al. 1994; Yamada & Nishi 1998). According to Elmegreen & Elmegreen (1978), the fastest growing modes have the size of the sheet thickness and the fragmentation timescale is $`t_{\mathrm{frag}}=(4\pi \rho (0)w^2)^{1/2},`$ where $`w^20.14`$ in the high external pressure limit. We estimate the line mass $`l_0`$ at fragmentation by a condition $`t_{\mathrm{frag}}=t_{dyn}`$, where $`t_{dyn}=\rho (0)/\dot{\rho }(0)`$. Under an assumption that cold ($`T\begin{array}{c}<\hfill \\ \hfill \end{array}10^3`$ K) sheets fragment into filaments, the resultant line mass is tabulated in Table 2 with other properties, where $`f`$ is the ratio of $`l_0`$ to the critical line density $`l_c2kT_0/m_pG`$, and $`x_s`$ and $`x_c`$ denote the mass fraction of the shocked matter and cooled matter, respectively. It is noted that $`f`$ is smaller than unity for the fastest growing modes. Hence, for the fragmentation, it is necessary for other growing modes to accumulate further mass so that the filaments would be super-critical. The super-critical filaments would eventually bear massive stars (Nakamura & Umemura 1999).
We can estimate the minimum mass of the starburst pancakes by $`M_{\mathrm{SB}}\overline{\rho }_{\mathrm{ini}}\lambda _{\mathrm{cool}}^3`$. Again, if the initial stage is at the maximum expansion, then
$$M_{\mathrm{SB}}=2.2\times 10^{11}M_{}\left(\frac{1+z_c}{5}\right)^{4.2}\left(\frac{\mathrm{\Omega }_\mathrm{B}h^2}{0.02}\right)^{1.4}.$$
(22)
The collapse epochs of the clouds in region (b) range from $`z_c3`$ to $`z_c10`$. The clouds may undergo further shrinking after the sheet-like collapse, because the rotation barrier is smaller by $`\mathrm{\Lambda }_{\mathrm{spin}}^2`$ than the maximum expansion size, where $`\mathrm{\Lambda }_{\mathrm{spin}}`$ is the dimensionless spin parameter which is peaked around 0.05 (Heavens & Peacock 1988). Consequently, the violent relaxation would take place in a collisionless fashion. Hence, the starburst pancakes highly possibly lead to the dissipationless galaxy formation. It is worthy noting that $`M_{\mathrm{SB}}`$ lies between 1$`\sigma `$ and 2$`\sigma `$ fluctuations in a standard CDM scenario. Thus, they could form a relatively clustered population of galaxies.
### 5.3 Retarded Galaxy Formation
The clouds with the smaller mass than $`M_{\mathrm{SB}}`$ cannot cool below $`10^4`$K at the first pancake collapse, but they might pass $`\lambda _{\mathrm{shield}}`$ in the course of shrinking down to the rotation barrier, so that the clouds are self-shielded against external UV radiation to cool down by H<sub>2</sub> molecules. Assuming $`\mathrm{\Lambda }_{\mathrm{spin}}=0.05`$, the marginal size, $`\lambda _{\mathrm{rot}}`$, above which the clouds can pass $`\lambda _{\mathrm{shield}}`$ is much smaller than size of the Jeans unstable clouds for $`z_c0`$. Thus, no cloud can be stopped by the rotation barrier before the cloud is self-shielded. In the clouds in the region (c), the star formation is likely to be retarded until the pancake disk shrinks considerably. Therefore, the region (c) tends to result in the so-called dissipational galaxy formation (e.g. Larson 1976; Carlberg 1985; Katz & Gunn 1991). The collapse epochs are relatively later ($`0\begin{array}{c}<\hfill \\ \hfill \end{array}z_c\begin{array}{c}<\hfill \\ \hfill \end{array}4`$), compared to the region (b), and the region (c) is corresponding to CDM fluctuations lower than 1$`\sigma `$. Thus, it is likely that the region (c) leads to a less clustered population of galaxies. Futhermore, for low redshifts ($`z\begin{array}{c}<\hfill \\ \hfill \end{array}1`$), the intensity of UV background is no longer at a level of $`I_{21}=0.1`$, but could be smaller by two orders of magnitude (Maloney 1993; Henry & Murthy 1993; Dove & Shull 1994). Since $`\lambda _{\mathrm{shield}}I_{21}`$ for $`\alpha =1`$ as seen in (21), the clouds with $`z\begin{array}{c}<\hfill \\ \hfill \end{array}1`$ in the region (c) might cool quickly in the course of shrinking down to the rotation barrier.
### 5.4 Expanding Clouds
In the region (d), the clouds are no longer gravitationally bound because of the enhancement of thermal pressure due to UV heating. They are relatively low mass systems. They could be Lyman alpha absorption systems which are seen in QSO spectra (Umemura & Ikeuchi 1984, 1985; Bond, Szalay, & Silk 1988).
## 6 CONCLUSIONS
We have numerically explored the thermal and dynamical evolution of pregalactic clouds under UV background radiation. The plane-parallel collapse of primordial gas clouds has been pursued, including chemical reactions with respect to hydrogen molecules as well as atomic hydrogen. Also, the radiation transfer for the ionizing photons has been properly treated. As a result, it is found that the cloud evolution under UV background branches off into four categories in the initial parameter space of density and size. They are (a) promptly self-shielded clouds, which evolve into dense objects with $`\begin{array}{c}<\hfill \\ \hfill \end{array}10^6M_{}`$ at collapse redshifts ($`z_c`$) greater than 10, (b) starburst pancake clouds with the mass higher than $`M_{\mathrm{SB}}=2.2\times 10^{11}M_{}\left[(1+z_c)/5\right]^{4.2}`$, which lead to the dissipationless galaxy formation at $`3\begin{array}{c}<\hfill \\ \hfill \end{array}z_c\begin{array}{c}<\hfill \\ \hfill \end{array}10`$, (c) retarded star-forming galaxies with the mass lower than $`M_{\mathrm{SB}}`$, which undergo star formation in the course of shrinking down to the rotation barrier at $`0\begin{array}{c}<\hfill \\ \hfill \end{array}z_c\begin{array}{c}<\hfill \\ \hfill \end{array}4`$, consequently leading to the dissipational galaxy formation, and (d) less massive expanding clouds, which could be detected as Lyman alpha absorbers in QSO spectra. If we assume a standard CDM cosmology, density fluctuations of 1$`\sigma `$-2$`\sigma `$ coincides with $`M_{\mathrm{SB}}`$. That is, fluctuations higher than 2$`\sigma `$ result in the dissipationless formation of massive galaxies at $`3\begin{array}{c}<\hfill \\ \hfill \end{array}z_c\begin{array}{c}<\hfill \\ \hfill \end{array}10`$, and eventually constitute a clustered population. From a further realistic point of view, there must be local subgalactic clumps growing in a pregalactic cloud in a bottom-up scenario like a CDM cosmology. They are likely to be self-shielded earlier than uniform components. Thus, $`M_{\mathrm{SB}}`$ can be considered as a measure of determining the bulge-disk ratios (B/D) of formed galaxies. Above $`M_{\mathrm{SB}}`$, the B/D ratio should be larger, while the ratio becomes smaller below $`M_{\mathrm{SB}}`$.
###### Acknowledgements.
We are very grateful to Steven N. Shore for reviewing the article and making valuable comments and criticism, and also for enormous assistance as a editor. We also thank T. Nakamoto for continuous encouragement. We thank also R. Nishi and H. Uehara for useful discussions. This work is supported in part by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists, No.2370 (HS), and the Grants-in Aid of the Ministry of Education, Science, Culture, and Sports, 09874055 (MU).
## Appendix A Frequency Integration for Photoionization and UV Heating Rate
In this appendix, a method to integrate the frequency dependence in radiation transfer equation is presented (see also Tajiri & Umemura 1998). This method is applied to plane-parallel calculations in this paper. However, it is also potent for the 3D calculations on cosmic ionization problem (Nakamoto, Umemura & Susa 1999). The frequency-dependent radiation transfer equation is given by a general form as,
$$𝒏I_\nu =\kappa _\nu I_\nu +\eta _\nu $$
(A1)
where $`𝒏`$ is the unit directional vector, $`\kappa _\nu `$ and $`\eta _\nu `$ denote the opacity and the emissivity, respectively. For the gas composed of pure hydrogen, the emissivity is almost null for $`\nu \begin{array}{c}>\hfill \\ \hfill \end{array}\nu _L+\mathrm{\Delta }\nu _T`$, where $`\mathrm{\Delta }\nu _T`$ denotes the width of the Lyman limit emission resulting from radiative recombination;
$$\mathrm{\Delta }\nu _TkT/h_\mathrm{P},$$
(A2)
where $`T`$, $`k`$ and $`h_\mathrm{P}`$ denote the gas temperature, and Planck constant, respectively. Typically, $`\mathrm{\Delta }\nu _T/\nu _L0.1`$ for $`T=10^4`$K. As far as higher energy photons of $`\nu \begin{array}{c}>\hfill \\ \hfill \end{array}\nu _L+\mathrm{\Delta }\nu _T`$ are concerned, the solution of equation (A1) is simply
$`I_\nu =I_\nu ^{\mathrm{in}}\mathrm{exp}\left(\tau _\nu \left(\omega \right)\right),`$ (A3)
where $`I_\nu ^{\mathrm{in}}`$ denotes the incident intensity at the boundary. The optical depth $`\tau _\nu (\omega )`$ in the solid angle $`\omega `$ is rewritten in terms of the optical depth at Lyman limit, $`\tau _{\nu _L}(\omega )`$,
$`\tau _\nu (\omega )=\tau _{\nu _L}(\omega )\left(\nu _L/\nu \right)^3,`$ (A4)
because the photoionization cross-section is $`\sigma _\nu \sigma _{\nu _L}(\nu _L/\nu )^3`$. Using the equations (A3) and (A4), we obtain the photoionization rate coefficient for this energy range of photons, $`k_{\mathrm{HI}}^{abs}`$, in terms of an integration as
$`k_{\mathrm{HI}}^{abs}`$ $`=`$ $`{\displaystyle _{\nu _L+\mathrm{\Delta }\nu _T}^{\mathrm{}}}{\displaystyle \frac{I_\nu }{h\nu }\sigma _\nu 𝑑\omega 𝑑\nu },`$ (A5)
$``$ $`{\displaystyle \frac{I_{\nu _L}^{\mathrm{in}}\sigma _{\nu _L}}{h}}{\displaystyle \frac{1}{3\tau _{\nu _L}(\omega )^{1+\alpha /3}}\gamma (1+\alpha /3,\tau _{\nu _L}\left(\omega \right))𝑑\omega },`$
$``$ $`\tau _{\nu _L,v}^{1\alpha /3}(\tau _{\nu _L,v}1),`$ (A6)
where we have assumed $`I_\nu ^{\mathrm{in}}=I_{\nu _L}^{\mathrm{in}}(\nu /\nu _L)^\alpha `$ and $`\gamma (a,b)`$ denotes the incomplete gamma function. $`\tau _{\nu _L,v}`$ denotes the optical depth measured in the vertical directions of a slab, which is already introduced in section 4. It is noted that the integration in frequency space has been carried out just analytically with the incomplete gamma function. In other words, we do not have to solve the radiation transfer equation for $`\nu \begin{array}{c}>\hfill \\ \hfill \end{array}\nu _L+\mathrm{\Delta }\nu _T`$.
On the other hand, photons with $`\nu _L\nu \nu _L+\mathrm{\Delta }\nu `$ are scattered to produce diffuse radiation. For hydrogen, the scattering albedo is given by $`[\alpha _A(T)\alpha _B(T)]/\alpha _A(T)`$, where $`\alpha _A(T)`$ is the total recombination coefficient to all bound levels and $`\alpha _B(T)`$ is the recombination coefficient to all excited levels. Hence, we have to integrate the radiation transfer equation including the emissivity by scatterings. In this frequency range, the frequency-dependence of the opacity and the emissivity are thought to be small, because $`\mathrm{\Delta }\nu _T`$ is ten times smaller than $`\nu _L`$ for $`T=10^4\mathrm{K}`$. So, we approximate the opacity and emissivity to be constant in this frequency range. After the transfer equation is solved, we calculate the angle averaged intensity $`J_{\nu _L}`$ in terms of the numerical solution $`I_{\nu _L}`$. Using $`J_{\nu _L}`$, we obtain the photoionization rate coefficient for this frequency range, $`k_{\mathrm{HI}}^{sca}`$;
$`k_{\mathrm{HI}}^{sca}4\pi {\displaystyle \frac{J_{\nu _L}\sigma _{\nu _L}}{h\nu _L}}\mathrm{\Delta }\nu _T`$ (A7)
Finally, by adding (A6) to (A7), we obtain the total ionization rate $`k_{\mathrm{HI}}`$;
$$k_{\mathrm{HI}}=k_{\mathrm{HI}}^{abs}+k_{\mathrm{HI}}^{sca}.$$
(A8)
The UV heating rate $`\mathrm{\Gamma }_{\mathrm{HI}}`$ is also obtained in a similar way. The result is,
$`\mathrm{\Gamma }_{\mathrm{HI}}`$ $`=`$ $`\mathrm{\Gamma }_{\mathrm{HI}}^{abs}+\mathrm{\Gamma }_{\mathrm{HI}}^{sca},`$ (A9)
$`\mathrm{\Gamma }_{\mathrm{HI}}^{abs}`$ $`=`$ $`{\displaystyle _{\nu _L+\mathrm{\Delta }\nu _T}^{\mathrm{}}}{\displaystyle \frac{I_\nu }{h\nu }\sigma _\nu (h\nu h\nu _L)𝑑\omega 𝑑\nu },`$
$``$ $`{\displaystyle \frac{I_{\nu _L}^{\mathrm{in}}\sigma _{\nu _L}}{h}}{\displaystyle \frac{1}{3\tau _{\nu _L}(\omega )^{1+\left(\alpha 1\right)/3}}\gamma (1+\left(\alpha 1\right)/3,\tau _{\nu _L}\left(\omega \right))h\nu _L𝑑\omega }`$
$`h\nu _Lk_{\mathrm{HI}}^{abs},`$ (A10)
$`\mathrm{\Gamma }_{\mathrm{HI}}^{sca}`$ $``$ $`4\pi J_{\nu _L}\sigma _{\nu _L}\mathrm{\Delta }\nu _T`$ (A11)
In particular, for an infinite sheet, we have an asymptotic expression for the absorption part ($`\mathrm{\Gamma }_{\mathrm{HI}}^{abs}`$) in an optically thick limit as
$$\mathrm{\Gamma }_{\mathrm{HI}}^{abs}\frac{2\pi I_{\nu _L}^{\mathrm{in}}\sigma _{\nu _L}\nu _L}{3}\frac{\mathrm{\Gamma }\left(1+\left(\alpha 1\right)/3\right)}{2+\left(\alpha 1\right)/3}\tau _{\nu _L,v}^{1\left(\alpha 1\right)/3},$$
(A12)
where $`\mathrm{\Gamma }(a)`$ represents the gamma function. It should be noted that the UV heating rate is not proportional to $`\mathrm{exp}(\tau _{\nu _L,v})`$, but is proportional to $`\tau _{\nu _L,v}^{1\left(\alpha 1\right)/3}`$ for large $`\tau _{\nu _L,v}`$. If we choose $`\alpha =1`$ as in the present paper, the UV heating rate is proportional to $`\tau _{\nu _L,v}^1`$. This shallow dependence upon the optical depth comes from a steep dependence of ionization cross section upon frequencies. |
warning/0001/hep-th0001044.html | ar5iv | text | # 1 Introduction
## 1 Introduction
After some intensive study carried out last year, (for example see ), it becomes now evident that there are several motivations to study tachyon condensation scenario. For example, (i) To seek correct vacuum of modular invariant but tachyonic string theories, for example in bosonic strings, type 0 (OA, OB), and also some heterotic strings. (ii) To find stable non-SUSY soliton. Particularly successful examples are the construction of the type I spinor particle , and also non-SUSY string junction . (iii) To find unstable anomaly free string theories . Some examples are $`U(N)\times U(N)`$ type IIB superstring theory, $`U(N)`$ type IIA, and type I analogue. This is closely related to Witten’s K-theory argument since in his approach arbitrary D-brane is constructed out of several pairs of $`D`$-9 and $`\overline{D}`$-9 branes. Such a $`D`$-9 configuration gives rise to the extra gauge symmetry in those theories.
Tachyon condensation is a dynamical process like Higgs mechanism. At this moment, although there are many important references , there are still many misteries in our understanding in terms of conformal field theory. In this paper, we choose a particulary simple example (bosonic string) and demonstrate the deformation in terms of boundary conformal field theory explicitly. We point out that the duality in conformal field theories play essential rôle to understand the deformation.
## 2 A brief review of tachyon condensation
Consider $`D`$-$`p`$$`\overline{D}`$-$`p`$ system. Open string sectors are labeled by $`U(2)`$ CP factor. Fermion number operator for CP factor becomes nontrivial.
$$(1)_{CP}^F=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)(=\sigma _3)$$
For the total open string wave function $`\mathrm{\Psi }^{total}=\mathrm{\Psi }^{osc}\mathrm{\Psi }^{CP}`$, the fermion number parity operator acts as,
$$(1)^F\mathrm{\Psi }^{total}=((1)_{osc}^F\mathrm{\Psi }^{osc})(\sigma _3\mathrm{\Psi }^{CP}\sigma _3).$$
It implies that:
(i) In $`\mathrm{\Psi }_{CP}=\sigma _0(=1),\sigma _3`$ sector, we have the usual GSO projection. There are $`U(1)_{\sigma _0}\times U(1)_{\sigma _3}`$ gauge fields at the massless level.
(ii) In $`\mathrm{\Psi }_{CP}=\sigma _1,\sigma _2`$ sector, we have the opposite GSO projection which gives a complex tachyon field ($`T,\overline{T}`$).
We remark that $`T`$ (resp. $`\overline{T}`$) have charge 2 (-2) under $`A_{\sigma _3}`$ but neutral under $`A_{\sigma _0}`$. We assume that the tachyon field has the following potential which is analogous to that of Higgs fields. Based on this potential we may discuss the process of the pair annihilation of the D-branes in the similar fashion with more familiar symmetry breaking mechanism<sup>2</sup><sup>2</sup>2 Recently there are some development to prove it by using the string field theory ..
There are two equivalent scenarios to realize the tachyon condensation. Let us start from $`D`$-$`(p+2)`$ and $`\overline{D}`$-$`(p+2)`$ branes.
* Step by step method proposed by A. Sen :
1. First consider kink configuration for complex tachyon ($`T(x)\pm T_0`$ as $`x\pm \mathrm{}`$) which define D-$`(p+1)`$ brane.
2. Since moduli of tachyon is defined by $`S^1`$, such configuration is topologically unstable.
3. On this D-$`(p+1)`$ brane, there is a real-valued tachyon field. One may again consider a kink configuration, which defines $`D`$-$`p`$ brane.
4. Since the moduli is $`Z_2`$, this tachyon configuration becomes topologically stable which defines $`D`$-$`p`$ brane.
* Topological method proposed by E. Witten :
We consider a topologically nontrivial configuration (vortex) for gauge fields and tachyons in two dimensions,
$`{\displaystyle \frac{1}{2\pi }}{\displaystyle F_{\sigma _3}}`$ $`=`$ $`n{\displaystyle \frac{1}{2\pi }}{\displaystyle F_{\sigma _0}}=0`$
$`T(x)`$ $``$ $`e^{in\theta }|T_0|\text{as}|x|\mathrm{}`$
$``$ $`x_1\sigma _1+x_2\sigma _2+O(|x|^2).`$
D-$`p`$ and D-$`\overline{p}`$ can be pair annihilated if tachyon field takes $`|T|=|T_0|`$ at spacial infinity. Then n D-$`(p2)`$ branes appear at the core of vorticity. $`U(1)_{\sigma _0}\times U(1)_{\sigma _3}`$ is broken to $`U(1)_{\sigma _0}`$.<sup>3</sup><sup>3</sup>3 Fate of $`U(1)_{\sigma _0}`$ is known to be Witten’s U(1) problem. See for example for the attempts to understand the issue.
BCFT method is easier to be applied to the first scenario. On the other hand, its topological nature is illuminated in the second. We will use both of them.
## 3 A Scenario for Bosonic string
In the following sections, we will mainly consider the tachyon condensation of the bosonic string by following . Some of the key motivations are (i) it is the simplest string models, and (ii) it is nonetheless the most generic string theory. It is supposed to contain superstring theories, type 0 theories, heterotic strings as its particular vacuum. It is therefore important to seek the fate of the bosonic string if any. It would be also desirable if one may find a dynamical scenario for the compactification of the 26 dimensional space-time.
A possible candidate of the destination of the open bosonic string would be,
1. $`SO(32)`$ type I string
2. Type 0 string ($`SO(32)\times SO(32)`$ theory) which is more plausible since it has bosonic spectrum and still contains a closed string tachyon.
Open unoriented bosonic string was studied by some groups , and it has been known that a consistent model exists with $`D=26`$ bosonic string and space filling $`D`$-25-branes. Tadpole free condition restricts the number of 25-branes to be $`2^{13}=8192`$. This condition comes from the cancellation of massless excitation arising from the boundary states of the open boundary and the crosscap. If this condition is satisfied, the model becomes a finite theory if one applies the zeta function regularization to the infinity from the tachyon mode.
Although it is very interesting that such a consistent model exists in purely bosonic theory, it is obviously discouraging that
1. it has a huge gauge symmetry $`SO(8192)`$ to be realistic,
2. it is defined only in 26 dimensions,
3. it has many open string tachyons since there is no GSO projection,
4. there is no D-brane charges and every D-branes becomes unstable.
In the following discussions, we would like to point out that these weak points may be eliminated by using a topologically nontrivial configuration for the tachyon fields.
### 3.1 Sen’s argument
Sen applied applied his step by step method to the system with two D-$`(p+1)`$-branes. There are four tachyons arising from four CP sectors of the open string. One may introduce $`Z_2`$ Wilson line to one of the D-branes to make the intertwining open string anti-periodic in one space direction. One D-$`p`$-brane will appear at the position of the kink of the tachyon field,<sup>4</sup><sup>4</sup>4 Exact description in terms of boundary state will be discussed in the next section. while original two D-branes disappear. In the exact treatment of such process, one need to compactify one of the space direction on a circle. By such compactification, the mass squared of tachyon field increases while the radius of $`S^1`$ become smaller. At a specific radius, the tachyon mode becomes massless and the exact treatment of the tachyon condensation becomes possible. In this process, one is forced to reduce the number of the uncompactified dimensions. Successive application of this idea to $`SO(2^{13})`$ theory would imply that one $`D`$-12 brane will be produced at the end of tachyon condensation while we have 13 space-time dimensions. Since such a theory can not be tadpole free in any respect, this scenario seems rather unnatural.
Sen have also argued that there are other options.
* Starting from a single D-brane, there exists an open string tachyon. If one consider the kink for this tachyon mode, one D-$`(p+1)`$-brane may produce several D-$`p`$-branes.
* If tachyon develops several kinks, there may appear several D-$`p`$-branes. Therefore, the number of D-branes can be arbitrary.
The first possibility seems not plausible since Cardy’s constraint on the boundary state implies that it is not possible to have an anti-periodic tachyon field in case of a single D-brane. The second possibility seems rather artificial. Such an arbitrariness always arises if we seek the tachyon condensation in topologically unstable configurations. However, as wee see, one may achieve some kind of topological stability even in the case of bosonic string. If we restrict our scope to the topologically stable sectors, there is no such arbitrariness.
### 3.2 Our proposal
It seems to us that the only topologically stable tachyon condensation scenario is a process where two D-$`(p+2)`$-branes with tachyon condensation produce a single D-$`p`$ brane while compactifying two space dimensions toroidally. In this scenario, we have $`2^{13d/2}`$ D-branes while compactifying $`d`$ dimensions. It matches with the tadpole free conditions for generic open string. If we take $`d=16`$, we have $`SO(32)`$ theory at dimension 10 which is a candidate model which can be dual to type I or type 0 string theories.
To explain the topological nature, let us consider two D-branes. The gauge group associated with it is $`SU(2)`$ except for the overall $`U(1)`$. The vacuum moduli of the tachyon potential is $`S^3`$. In such a situation, one may not expect topologically stable configuration since $`\pi _1(S^3)=0`$.
However, at the special infinity, the gauge transformation can be nontrivial. We recall that the open string transforms in adjoint representation. The actual gauge group which acts nontrivially on the open string is not $`SU(2)`$ but $`SO(3)`$. In this context one may have a topologically nontrivial loop in the gauge configuration space since,
$$\pi _1(SO(3))=𝐙_2.$$
It also means that the topologically nontrivial sector is in a sense unique. One may interpret it that the number of the D-$`p`$ brane which will appear at the core is restricted to be one.
To illustrate this idea more explicitly, we compactify two dimensions toroidally and require the behavior of the tachyon field at the core of vorticity,
$$T(x)x^1\sigma ^1+x^2\sigma ^2(\text{as }|x|0).$$
(1)
Tachyon must be anti-periodic in $`x^i`$ directions in $`\sigma ^i`$ sector. This configuration appeared in to describe nontrivial vortex configuration.
Roughly speaking, such a configuration is realized if we introduce a kind of non-abelian Wilson lines which produces curvature,
$$A_1\sigma _2A_2\sigma _1$$
It costs some energy but since they turn out to be in the topologically nontrivial, such a configuration is stable. A simple realization is possible if we impose the following boundary condition for the open string,
$`\mathrm{\Psi }(x^1+2\pi R,x^2)`$ $`=`$ $`\sigma _2\mathrm{\Psi }(x^1,x^2)\sigma _2,`$
$`\mathrm{\Psi }(x^1,x^2+2\pi R)`$ $`=`$ $`\sigma _1\mathrm{\Psi }(x^1,x^2)\sigma _1.`$
It belongs to the topologically nontrivial sector since the gauge transformation (=Wilson line) around the loop is $`\sigma ^1\sigma ^2(\sigma ^1)^1(\sigma ^2)^1=1`$ <sup>5</sup><sup>5</sup>5This is a projective representation of the lattice translation group. Similar nontrivial representation in CP appeared in the description of the discrete torsion in orbifold conformal field theory.. We summarize the periodicity along each direction,
| | $`\sigma ^0`$ | $`\sigma ^1`$ | $`\sigma ^2`$ | $`\sigma ^3`$ |
| --- | --- | --- | --- | --- |
| $`x^1`$ | + | $``$ | + | $``$ |
| $`x^2`$ | + | + | $``$ | $``$ |
Gauge symmetry is broken from $`U(2)`$ to $`U(1)`$ which is quite unusual but necessary to describe pair annihilation of two D-branes.
Since we simply impose the antiperiodic boundary condition to some components of the open string, the spectrum of the topologically nontrivial sector is very easy to calculate. If the original open string lives in radius $`R`$, the momentum distribution of the twisted theory is the equally separated momenta with the separation $`1/2R`$. Therefore, the spectrum of the topologically nontrivial sector becomes identical the either of the following two systems,
1. one D-$`(p+2)`$ brane in radius $`2R`$,
2. one D-$`p`$ brane in radius $`1/2R`$.
It supports our expectation that only one D-$`p`$ brane appear at the core of vorticity.
To define the unoriented string theory, we need to impose the orientation projection $`\mathrm{\Omega }`$ to the open string wave function. At each mass level, we have project out either symmetric (anti-symmetric) part of the wave function. Such the projection is compatible with our boundary condition since Pauli matrices satisfies well-defined parity under adjoint action of $`\mathrm{\Omega }`$. The symmetry breaking pattern $`U(2)U(1))`$ remains the same but the precise correspondence in the momentum distribution is corrupted.
To extend our analysis to the compactification of the higher dimensional tori is straightforward. To have nontrivial configuration in $`2d`$ dimensions, general prescription will be,
$`\mathrm{\Psi }(x^{262d},\mathrm{},x^{252d+i}+2\pi R,\mathrm{},x^{25})`$
$`=\stackrel{~}{\mathrm{\Gamma }}_i\mathrm{\Psi }(x^{262d},\mathrm{},x^{252d+i},\mathrm{},x^{25})\stackrel{~}{\mathrm{\Gamma }}_i`$
where $`\stackrel{~}{\mathrm{\Gamma }}`$ satisfies,
$`\{\stackrel{~}{\mathrm{\Gamma }}_i,\stackrel{~}{\mathrm{\Gamma }}_j\}=2\delta _{ij}`$
$`\{\stackrel{~}{\mathrm{\Gamma }}_i,\mathrm{\Gamma }_i\}=0[\stackrel{~}{\mathrm{\Gamma }}_i,\mathrm{\Gamma }_j]=0(ij)`$
With this prescription, $`\mathrm{\Psi }`$ behaves at the core as,
$$\mathrm{\Psi }(x)=\underset{i=1}{\overset{2d}{}}x^{252d+i}\mathrm{\Gamma }_i+O(|x|^2)$$
which is identical to the tachyon configuration of the superstring as discussed by Witten . With this boundary condition, Chan-Paton gauge symmetry is broken from $`U(2^d)`$ to $`U(1)`$.
### 3.3 Symmetry enhancement from closed strings?
At $`R=1/\sqrt{2}`$, we get extra gauge symmetry from vertex operator $`e^{iX(z)/R}`$ in the closed string sector. In general, if the compactified direction is given by the root lattice of $`G`$, gauge group is enhanced to $`G`$. If we compactify $`16`$ dimensions, the maximal enhancement is given by $`SO(32)`$ or $`E_8\times E_8`$. This well-known mechanism was used to construct the heterotic string theory.
Tachyon fields which interpolates the different D-branes becomes anti-periodic in the compactified direction and it has the mode expansion,
$$T(x)=\underset{n𝐙}{}T_{n1/2}e^{i(n1/2)X}.$$
Mass of each mode $`T_{n1/2}`$ is given by $`m^2=1+(n1/2)^2/2R^2`$. The lowest mode $`T_{\pm 1/2}`$ becomes massless when $`R=1/2\sqrt{2}`$. We expect that the existence of massless mode signifies the enhanced gauge symmetry. However the radius for the massless tachyon is half of that of the closed string gauge enhancement point.
## 4 Tachyon condensation and Boundary CFT
As we saw in the last section, if we fix the radius of the target space to $`R=1/2\sqrt{2}`$, some of the tachyon mode becomes massless. In the case of the closed string excitation, such a massless mode is used to deform the background (metric, antisymmetric tensors etc.) of the target space. On the other hand, the deformation by the massless open string mode usually triggers the deformation of the D-brane to which open string is attached. In our situation, the deformation by massless mode induces the tachyon condensation and we want to follow the change of the D-brane state in such a process as exactly as possible. In this section, we give an explicit description of boundary conformal field theory in Sen’s scenario.
Before we discuss the detail, it is useful to to indicate that the radius $`R=1/2\sqrt{2}`$ is exactly the point where the moduli space of toroidal compactification and orbifold compactification meets. This point is quite essential in the following discussion.
### 4.1 Analysis in open string Hilbert space
In this section, we review Sen’s discussion of the analysis which uses the open string Hilbert space. To treat the condensation of the tachyon field, Sen introduced new variables to describe the compactified direction $`X`$,
$`e^{i\sqrt{2}X}`$ $`=`$ $`Y+iZ`$
$`e^{i\sqrt{2}Y}`$ $`=`$ $`XiZ`$
$`e^{i\sqrt{2}Z}`$ $`=`$ $`X+iY`$
By using the new variables, tachyon condensation is equivalent to the deformation induced by $`Y\sigma _1dz`$.
As we emphasized before, the radius for the massless tachyon is not equal to the radius for the gauge enhancement in the closed string sector. Although we introduced the variables of $`SU(2)`$ current algebra, we need to project out some part of the spectrum. We should not forget that $`e^{\pm i\sqrt{2}X}`$ is antiperiodic in direction $`X`$.
Such a projection can be carried out by introducing two parity operators,
$`h`$ $``$ $`XX+\pi /\sqrt{2}`$
$`g`$ $``$ $`XX`$
The first one is the translation in $`X`$ direction. Open string modes in $`\sigma ^1,\sigma ^2`$ sectors becomes odd under this operator.
On the other hand, we do not originally have any projection of the second parity operator. However, to define a similar projection in $`Y`$ variable, these two operators change their rôle,
$$h_X=g_Yg_X=h_Y.$$
This relation motivates us to introduce the second operator.
Let us now proceed to define necessary projections in terms of $`Y`$ variable. Originally in $`Y`$ direction, we have both boundary condition
$$\mathrm{\Psi }(Y+2\pi )g\mathrm{\Psi }(Y)=\pm \mathrm{\Psi }(Y).$$
Since we have both periodic and anti-periodic boundary conditions, it is more natural to extend periodicity in $`4\pi `$.
$$\mathrm{\Psi }(Y+4\pi )=\mathrm{\Psi }(Y)$$
By the tachyon condensation deformation in $`Y`$ direction, we deform this boundary condition to
$`\mathrm{\Psi }(Y+4\pi )`$ $`=`$ $`\sigma _1\mathrm{\Psi }(Y)\sigma _1`$
The eigenvalues of $`h`$ and $`g`$ in each sectors are now deformed. We summarize it in the next table,
| | $`h_Y=g_X`$ | $`g_Y=h_X`$ |
| --- | --- | --- |
| $`I`$ | $`\pm 1`$ | $`+`$ |
| $`\sigma _1`$ | $`\pm 1`$ | $``$ |
| $`\sigma _2`$ | $`\pm i`$ | $``$ |
| $`\sigma _3`$ | $`\pm i`$ | $`+`$ |
If we combine the four sectors, we have both signs in $`g_Y`$ parity. For each sign, we have four twists $`\pm 1,\pm i`$ in the boundary condition. This indicates that momentum is now quantized in the unit $`1/4`$ i.e. the compactification radius is apparently changed to $`\sqrt{2}`$. If we take the T-dual in $`Y`$ direction, Radius becomes $`1/2\sqrt{2}`$ — original radius and two D $`(p+1)`$-branes become single $`p`$-brane.
### 4.2 Tachyon condensation in Boundary state
We now turn back to describe this tachyon condensation process in terms of the boundary state. The deformation is induced by insertions of
$$\mathrm{exp}\left(i\phi 𝑑zY\sigma _1\right)$$
at the boundary.
There are two difficult points which we need some care to discuss the deformation.
The first one is that the D-brane at the boundary must be switched by CP factor $`\sigma _1`$ at each insertion point as illustrated in the above figure. At each end of the inserted open string, we have different D-branes and we need to toggle CP factor in two ends.
The insertion of off-diagonal CP factor is usually treated by using the trace. In our case, we need some care since we have already deformed one of the D-brane by using $`Z_2`$ Wilson line.
The second point is that the deformation operator is the vertex operator in terms of original field, $`Y=\mathrm{cos}(X)`$. We have to change the dynamical variable to $`Y`$ as in the open string approach. The $`Z_2`$ transformation $`h`$ act on $`Y`$ as $`YY`$. It suggests that the natural framework is to use $`Z_2`$ orbifold variable.
To describe orbifold CFT, it is useful to prepare the boundary states for the theory with $`S^1`$ compactification. We have two types of boundary states,
* Dirichlet: $`|D(\phi )`$ ($`\phi \phi +2\pi r`$)
$`\frac{1}{\sqrt{2r}}_{k=0}^{\mathrm{}}e^{ik\phi /r}\mathrm{exp}\left(_{n=1}^{\mathrm{}}a_n^{}\stackrel{~}{a}_n^{}\right)|0,k`$
* Neumann:$`|N(\stackrel{~}{\phi })`$ ($`\stackrel{~}{\phi }\stackrel{~}{\phi }+\pi /r`$)
$`\sqrt{r}_{w=0}^{\mathrm{}}e^{2irw\stackrel{~}{\phi }}\mathrm{exp}\left(_{n=1}^{\mathrm{}}a_n^{}\stackrel{~}{a}_n^{}\right)|w,0`$
Physical interpretation of the deformation parameters are well-known to be location of D-brane for $`\phi `$ and the Wilson line for $`\stackrel{~}{\phi }`$. In this language, the boundary state before tachyon condensation is
$$|N(0)+|N(\pi /2r).$$
(2)
The boundary states for $`S^1/Z_2`$ orbifold was discussed in .
In this case, the boundary state which describes the location of D-brane in the interval and those located at the edges are essentially different. We write them by using the boundary states for $`S^1`$ compactification.
1. Boundary state for interval
($`0<\phi <\pi r`$, $`0<\stackrel{~}{\phi }<\pi /2r`$ )
$`|D_O(\phi )`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}(|D(\phi )+|D(\phi ))`$
$`|N_O(\phi )`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}(|N(\phi )+|N(\phi ))`$
2. Boundary state for edges (half D-branes) ($`\phi =0,\pi r`$, $`\stackrel{~}{\phi }=0,\pi /2r`$):
$`|D_O(\phi )\pm `$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}|D(\phi )\pm {\displaystyle \frac{1}{2^{1/4}}}|D(\phi )_T`$
$`|N_O(\stackrel{~}{\phi })\pm `$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}|N(\stackrel{~}{\phi })\pm {\displaystyle \frac{1}{2^{1/4}}}|N(\stackrel{~}{\phi })_T`$
Open string partition function with specific $`Z_2\times Z_2`$ charge (generated by $`h`$ and $`g`$) can be written in terms of edge boundary states:
$`Z_{++}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{_nq^{4n^2}}{_n(1q^n)}}+{\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{_n(1+q^n)}}=N_O(\stackrel{~}{\phi }_\pm )\pm |\mathrm{\Delta }|N_O(\stackrel{~}{\phi }_\pm )\pm `$
$`Z_+`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{_nq^{4n^2}}{_n(1q^n)}}{\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{_n(1+q^n)}}=N_O(\stackrel{~}{\phi }_\pm )\pm |\mathrm{\Delta }|N_O(\stackrel{~}{\phi }_\pm )`$
$`Z_+`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{_nq^{4(n+1/2)^2}}{_n(1q^n)}}=N_O(\stackrel{~}{\phi }_\pm )\pm |\mathrm{\Delta }|N_O(\stackrel{~}{\phi }_{})\pm `$
$`Z_{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{_nq^{4(n+1/2)^2}}{_n(1q^n)}}=N_O(\stackrel{~}{\phi }_\pm )\pm |\mathrm{\Delta }|N_O(\stackrel{~}{\phi }_{})`$
where $`\stackrel{~}{\phi }_\pm =0,\pi /2r`$.
By choosing specific radius $`r=1/2\sqrt{2}`$, two $`Z_2`$ symmetry becomes interchangeable,
$$Z_+=Z_+.$$
This accidental symmetry is a consequence of the fact that this model is an orbifold CFT and at the same time a CFT on $`S^1`$. In terms of edge boundary states, interchange of $`h`$ and $`g`$, (or $`J_3`$ and $`J_1`$) is described by the change of the rôle of two edges and the sign of the twisted sector, $`|N_O(\stackrel{~}{\phi }_{ϵ_1})ϵ_2|N_O(\stackrel{~}{\phi }_{ϵ_2})ϵ_1.`$
To summarize, the deformation associated with the tachyon condensation can be achieved by the following steps. (i) Prepare the boundary state which describe two bosonic D-branes with $`Z_2`$ Wilson line on one of them (2). (ii) Reinterpret the boundary state as those of $`S_1/Z_2`$ orbifold model. (iii) Change the rôle of $`J_1`$ and $`J_3`$. In terms of the boundary state, it amounts to the interchange of two parameters $`\stackrel{~}{\phi }`$ and $`ϵ`$. (iv) Deform the parameter $`\stackrel{~}{\phi }`$ of one of the D-brane boundary states. More explicitly, it can be carried out as follows.
$`|N(0)+|N(\pi /2r)`$ $`=`$ $`\sqrt{2}{\displaystyle \underset{ϵ_1,ϵ_2}{}}|N_O(\stackrel{~}{\phi }_{ϵ_1})ϵ_2`$
$`=`$ $`\sqrt{2}{\displaystyle \underset{ϵ_1,ϵ_2}{}}|N_O(\stackrel{~}{\phi }_{ϵ_2})ϵ_1`$
$``$ $`{\displaystyle \frac{1}{\sqrt{2}}}{\displaystyle \underset{ϵ_1,ϵ_2}{}}|N_O(\stackrel{~}{\phi }_{ϵ_2})ϵ_1+\sqrt{2}|N_O({\displaystyle \frac{\pi }{\sqrt{2}}})`$
We note that the boundary states in the final line produces the same partition function in the open string sector as that of $`|D(0)`$ for $`S^1`$ compactified model. This calculation explicitly demonstrates the production of one D-$`p`$ brane from two D-$`(p+1)`$ branes by the tachyon condensation. However, we have to mention one subtlety. In passing from the second to the third line, the intermediate state in the course of deformation can not be properly interpreted as D-brane boundary state since it would produce partition function with fractional coefficient in the open string sector. Namely, it does not satisfy Cardy’s constraint. In this sense, one can not describe the tachyon condensation as the continuous deformation of the boundary state. This phenomena was also discussed by Sen and he explained it by showing the appearance tadpole. In this sense, it is more appropriate that two vacua (a) two D-$`p+1`$ brane system and (b) one D-$`p`$ brane system are not connected continuously but belong to the different topological sectors.
## 5 Discussion
In the second half of this paper, I discussed duality with orbifold theory is essential to describe tachyon condensation. This seems to be a generic feature since massless tachyon is always described by vertex operator with anti-periodic boundary condition. In our analysis, the twisted sectors actually do not play any rôle since we always sum over various sectors. I hope that these states may have some meanings to understand still mysterious Witten’s $`U(1)`$ problem In superstring case, $`Z_2`$ operator $`g`$ is replaced by $`(1)^F`$. Orbifold CFT is described by KT point. |
warning/0001/cond-mat0001391.html | ar5iv | text | # Theory of Ferromagnetism in Diluted Magnetic Semiconductor Quantum Wells
\[
## Abstract
We present a mean field theory of ferromagnetism in diluted magnetic semiconductor quantum wells. When subband mixing due to exchange interactions between quantum well free carriers and magnetic impurities is neglected, analytic result can be obtained for the dependence of the critical temperature and the spontaneous magnetization on the distribution of magnetic impurities and the quantum well width. The validity of this approximate theory has been tested by comparing its predictions with those from numerical self-consistent field calculations. Interactions among free carriers, accounted for using the local-spin-density approximation, substantially enhance the critical temperature. We demonstrate that an external bias potential can tune the critical temperature through a wide range.
\]
The discovery of carrier-mediated ferromagnetism in diluted magnetic semiconductors (DMS) has opened a broad and relatively unexplored frontier for both basic and applied research. The interplay between collective magnetic properties and semiconductor transport properties in these systems presents a rich phenomenology and offers the prospect of new functionality in electronic devices. The possibilities for manipulating this interplay are especially varied when the free carrier system is quasi-two-dimensional because of the dependence of system properties on the subband wavefunction. In this article we address the dependence of the ferromagnetic critical temperature $`T_c`$ of two-dimensional DMS ferromagnets on the spatial distribution of magnetic ions, the subband wavefunction of the free carrier system, and on their interplay. Our analysis is based on a formulation of the mean-field theory for carrier-induced ferromagnetism in a DMS which was developed in earlier work and is intended to be useful for any spatially inhomogeneous system. We find that $`T_c`$ is quite sensitive to the relative distributions of electrons and magnetic ions and can be altered in situ by applying a gate voltage.
For definiteness we address the case of a $`001`$ growth direction 2D hole gas in the Ga<sub>1-x</sub>Mn<sub>x</sub>As DMS system. Much of our analysis, however, applies equally well to other interfaces and to other DMS systems with cubic host semiconductors and, with one important caveat which we mention below, also to the case of a 2D electron gas mediated ferromagnetism. Our theory is based on an envelope function description of the valence band electrons and a spin representation of their kinetic-exchange interaction with d-electrons on the $`S=5/2`$ Mn<sup>++</sup> ions:
$$=_m+_f+J_{pd}\underset{i,I}{}\stackrel{}{S}_I\stackrel{}{s}_i\delta (\stackrel{}{r}_i\stackrel{}{R}_I),$$
(1)
where $`i`$ labels a free carrier, $`I`$ labels a magnetic ion and the exchange interaction energy $`J_{pd}0.15`$ eV nm<sup>3</sup>. In Eq.( 1) $`_m`$ is the Hamiltonian of the magnetic ions, $`_f`$ is the four-band Luttinger Hamiltonian for free carriers in the valence band, $`\stackrel{}{S}`$ is the magnetic ion spin and $`\stackrel{}{s}`$ is the electron-spin operator projected onto the $`j=3/2`$ valence band manifold of the Luttinger Hamiltonian. We assume here that the 2D free carrier density is sufficiently low that only the lowest energy heavy-hole subband is occupied, and that the quantum well of interest is sufficiently narrow that mixing between light-hole and heavy-hole bands can be neglected. These simplifications lead to a single parabolic band with the in-plane effective mass $`m_{}^{}0.11m_0`$ and the out-of-plane mass $`m_z^{}0.38m_0`$. The two spin-states in this band have definite $`j_z=\pm 3/2`$ and definite $`s_z=\pm 1/2`$; the projection of the transverse spin components onto this band is zero so that the kinetic exchange interaction has an Ising character.
Our mean-field theory is derived in a spin-density-functional theory framework and leads to a set of physically transparent coupled equations. In the absence of external fields, the mean polarization of a magnetic ion is given by
$$m_I=SB_S\left(J_{pd}S\left[n_{}(\stackrel{}{R}_I)n_{}(\stackrel{}{R}_I)\right]/2k_BT\right);$$
(2)
where $`B_S(x)`$ is the Brillouin function,
$`B_S(x)`$ $`=`$ $`{\displaystyle \frac{2S+1}{2S}}\mathrm{coth}({\displaystyle \frac{2S+1}{2S}}x){\displaystyle \frac{1}{2S}}\mathrm{coth}({\displaystyle \frac{1}{2S}}x)`$ (3)
$``$ $`{\displaystyle \frac{S+1}{3S}}x{\displaystyle \frac{(S+1)(2S^2+2S+1)}{90S^3}}x^3,x1.`$ (4)
The electron spin-densities $`n_\sigma (\stackrel{}{r})`$ are determined by solving the Schrödinger equation for electrons which experience a spin-dependent potential due to kinetic exchange with the polarized magnetic ions:
$`[{\displaystyle \frac{1}{2m_{}^{}}}({\displaystyle \frac{^2}{x^2}}+{\displaystyle \frac{^2}{y^2}}){\displaystyle \frac{1}{2m_z^{}}}{\displaystyle \frac{^2}{z^2}}+v_{es}(\stackrel{}{r})`$ (5)
$`+v_{xc,\sigma }(\stackrel{}{r}){\displaystyle \frac{\sigma }{2}}h_{pd}(\stackrel{}{r})]\psi _{k,\sigma }(\stackrel{}{r})=ϵ_{k,\sigma }\psi _{k,\sigma }(\stackrel{}{r});`$ (6)
$$n_\sigma (\stackrel{}{r})=\underset{k}{}f(ϵ_{k,\sigma })|\psi _{k,\sigma }(\stackrel{}{r})|^2.$$
(7)
In these equations $`v_{es}(\stackrel{}{r})`$ is the electrostatic potential, including band offset and ionized impurity contributions, $`v_{xc,\sigma }(\stackrel{}{r})`$ is an exchange-correlation potential on which we comment further below, and $`f(ϵ)`$ is the Fermi distribution function. The spin-dependent kinetic-exchange potential,
$$h_{pd}(\stackrel{}{r})=J_{pd}\underset{I}{}\delta (\stackrel{}{r}\stackrel{}{R}_I)m_I$$
(8)
is non-zero only in the ferromagnetic state. In the following we assume that the magnetic ions are randomly distributed and dense on a scale set by the free carrier Fermi wavevector in the $`\widehat{x}\widehat{y}`$ plane, and that their density in the growth direction, $`c(z)`$, can be precisely controlled. This allows us to take a continuum limit where the magnetic ion polarization and the kinetic-exchange potential depends only on the growth direction coordinate and
$$h_{pd}(z)=J_{pd}c(z)m(z).$$
(9)
In the following we discuss how the ferromagnetic transition temperature $`T_c`$ depends on $`c(z)`$.
The $`T_c`$ calculation is greatly simplified when the spin-dependence of the exchange correlation potential is neglected and the quantum well width $`w`$ is small enough to make subband mixing due to kinetic exchange interactions negligible, i.e., when $`Jc(z)\mathrm{}^2/m_z^{}w^2`$. We see later that the theory which results from these approximations is normally quite accurate. With subband mixing neglected, the only effect of the mean-field kinetic-exchange interaction on the 2D carriers is to produce a rigid spin-splitting of the 2D bands, given by the first-order perturbation theory expression:
$$ϵ_{}ϵ_{}=J_{pd}𝑑z^{}m(z^{})c(z^{})|\psi (z^{})|^2$$
(10)
where $`\psi (z)`$ is the growth direction envelope function of the lowest subband. For the case of a quantum well with infinite barriers, $`\psi (z)=\sqrt{2/w}\mathrm{cos}(\pi z/w)`$. The electron spin-polarization can then be obtained by summing over occupied free-carrier states. It follows that, as long as both spin-$``$ and spin-$``$ bands are partly occupied,
$$n_{}(z)n_{}(z)=\frac{m_{}^{}}{2\pi \mathrm{}^2}|\psi (z)|^2J_{pd}𝑑z^{}m(z^{})c(z^{})|\psi (z^{})|^2.$$
(11)
Eq. (11) can be used to eliminate the free-carrier spin-polarization from Eq. (2) and to obtain a self-consistent equation for the function $`m(z)`$. The equation depends only on the kinetic-exchange coupling constant $`J_{pd}`$, the free-carrier in-plane effective mass $`m_{}^{}`$, the subband wavefunction $`\psi (z)`$, and the magnetic ion distribution function $`c(z)`$. It is important to realize that the entire system behaves collectively. It has a single critical temperature which depends on the function $`c(z)`$, rather than a z-dependent $`T_c`$ like that which appears in a naive adaptations of the bulk RKKY theory for $`T_c`$ to quasi-2D systems.
An explicit expression for the critical temperature can be obtained by linearizing the self-consistent equation (2). Expanding the Brillouin function to the first order in its argument gives the homogeneous linear integral equation
$$m(z)=\frac{1}{k_BT}𝑑z^{}K(z,z^{})m(z^{})$$
(12)
where the kernel
$$K(z,z^{})=\frac{S(S+1)J_{pd}^2}{12}\frac{m_{}^{}}{\pi \mathrm{}^2}|\psi (z)|^2c(z^{})|\psi (z^{})|^2.$$
(13)
Since the linearized equation is satisfied only at the critical temperature, $`k_BT_c`$ is equal to the largest eigenvalue of $`K(z,z^{})`$, the eigenvalue corresponding to eigenfunction $`f(z)|\psi (z)|^2`$. It follows that
$$T_c=\frac{S(S+1)}{12}\frac{J_{pd}^2}{k_B}\frac{m_{}^{}}{\pi \mathrm{}^2}𝑑z|\psi (z)|^4c(z).$$
(14)
The critical temperature is proportional to $`J_{pd}^2`$ and $`m_{}^{}`$, is independent of 2D carrier density, and for a system with uniformly distributed magnetic impurities is inversely proportional to the quantum well width, as illustrated by the solid line in Fig. 1. The accuracy of Eq. (14) has been tested by numerically solving the set of self-consistent Eqs. (2)-(9) together with the Poisson equation for $`v_{es}(z)`$ and the local-spin-density-approximation (LSDA) formula for $`v_{xc,\sigma }(z)`$. If exchange and correlation effects are neglected ($`v_{xc,\sigma }(z)=0`$), the self-consistent numerical results represented by circles in Fig. 1 are identical to those obtained from Eq. (14) for quantum well widths $`w=10`$, 15, and 20 nm. Interactions among the 2D carriers substantially enhance the critical temperature and also cause a strong dependence of $`T_c`$ on the 2D carrier density $`N`$, as illustrated by squares ($`N=1\times 10^{11}`$ cm<sup>2</sup>) and triangles ($`N=0.5\times 10^{11}`$ cm<sup>2</sup>) in Fig. 1. Note that unlike the 3D case, where $`T_c`$ is proportional to Fermi wavevector in the non-interacting limit and is an increasing function of density even if the interactions are included, the critical temperature for all three quantum wells of Fig. 1 is larger at lower density, $`N=0.5\times 10^{11}`$ cm<sup>2</sup>, where the interactions are stronger.
Before discussing the implications of the expression (14) for $`T_c`$ we comment on its relation to the RKKY theory of carrier induced ferromagnetism in DMS’s. As applied to bulk 3D electron systems the two approaches are fundamentally equivalent, although there are differences in detail which can be important. The principle difference is that the RKKY theory treats the free-carrier magnetic-ion interaction perturbatively. As a result both approaches give the same value for the critical temperature where the magnetization vanishes; differences appear at lower temperatures where the electron system is strongly or completely spin-polarized and a perturbative treatment of its interactions with the magnetic ions fails. Both approaches fail to account for the retarded character of the free-carrier mediated interaction between magnetic ions in these low density systems, and for correlated quantum fluctuations in magnetic ion and free carrier subsystems which are important for some properties. Both approaches are able to account for interaction effects in the free carrier system, which increase the tendency toward ferromagnetism as illustrated in Fig. 1, and in the density-functional approach appear in the spin-dependence of the exchange correlation potential. It is in inhomogeneous cases, such as the reduced dimension situation considered here, that the density functional approach has an advantage. Our approach provides a simple description of the collective behavior of the quasi-2D ferromagnet and is able to handle geometric features which can be engineered to produce desired properties. For example we show in Fig. 2 a plot of the dependence of the critical temperature on the fraction $`d`$ of the quantum well occupied by magnetic ions. When the integrated 2D density of magnetic ions ($`cd`$) is fixed, the critical temperature increases as magnetic impurities are transferred toward the center of the quantum well where the free-carrier wavefunction has its maximum. For fixed 3D density $`c`$ the number of magnetic impurities decreases with decreasing $`d`$. This effect is stronger than the increase of the overlap between the carrier wavefunction and magnetic ion distribution resulting in the decrease of $`T_c`$, as shown in the inset of Fig. 2. In both cases the analytic results of Eq. (14) are in qualitative agreement with the full numerical LSDA calculations.
One remarkable feature of quasi-2D ferromagnetic DMS’s is the possibility of tuning $`T_c`$ through a wide range in situ, by the application of a gate voltage. In Fig. 3 we illustrate this for the case of a $`w=10`$ nm quantum well with magnetic ions covering only a $`d=3`$ nm portion near one edge. Even for such a narrow quantum well, the critical temperature can be varied over an order of magnitude by applying a bias voltage which draws electrons into the magnetic ion region.
In closing we remark that for several reasons DMS ferromagnetism mediated by 2D electron systems is likely to occur only at inaccessibly low temperatures. One important factor is the smaller value of the exchange interaction parameter in the conduction band case. For example for the typical n-type II-VI DMS quantum well<sup>3</sup>, (Zn<sub>1-x</sub>Cd<sub>x</sub>Se)<sub>m-f</sub>(MnSe)<sub>f</sub>, the $`sd`$ exchange $`J_{sd}4\times 10^3`$ eV nm<sup>3</sup>. Our theory then gives $`T_c10`$ mK for a 10 nm wide quantum well. Equally important, however, is the weak magnetic anisotropy expected in the conduction band case, which will lead to soft spin-wave collective modes and substantial $`T_c`$ suppression.
This work was supported by the National Science Foundation under grants DMR-9714055 and DGE-9902579, and by the Grant Agency of the Czech Republic under grant 202/98/0085. |
warning/0001/hep-th0001147.html | ar5iv | text | # 1 INTRODUCTION
## 1 INTRODUCTION
The Maxwell equations and the Dirac equation are among the most celebrated equations of physics. Campolattaro (1980a, 1980b) started with the analysis of the Maxwell equations by writing the electromagnetic field tensor $`F_{\mu \nu }`$ in the equivalent bilinear form
$$F^{\mu \nu }=\overline{\mathrm{\Psi }}S^{\mu \nu }\mathrm{\Psi }$$
(1)
where $`\mu ,\nu =0,1,2,3.`$ $`\mathrm{\Psi }`$ is a Dirac spinor, $`\overline{\mathrm{\Psi }}=\mathrm{\Psi }^{}\gamma ^0`$ is the Dirac conjugation of $`\mathrm{\Psi }`$. $`S^{\mu \nu }`$ is the spin operater defined by
$$S^{\mu \nu }=\frac{i}{4}[\gamma ^\mu ,\gamma ^\nu ]$$
(2)
and the $`\gamma `$’s are the Dirac matrices satisfying
$$\gamma ^\mu \gamma ^\nu +\gamma ^\nu \gamma ^\mu =2\eta ^{\mu \nu }$$
(3)
with $`\eta ^{\mu \nu }`$ the Minkowski metric tensor given by $`\eta ^{\mu \nu }=diag(1,1,1,1).`$
In this representation the dual tensor
$$F^{\mu \nu }=\overline{\mathrm{\Psi }}\gamma ^5S^{\mu \nu }\mathrm{\Psi }$$
(4)
From now on, the Einstein sum convention is adopted throughout. The Maxwell equations read (a comma followed by an index represents the partial derivative with respect to the variable with that index)
$`(\overline{\mathrm{\Psi }}S^{\mu \nu }\mathrm{\Psi })_{,\mu }`$ $`=`$ $`j^\nu `$ (5)
$`(\overline{\mathrm{\Psi }}\gamma ^5S^{\mu \nu }\mathrm{\Psi })_{,\mu }`$ $`=`$ $`0`$ (6)
Moreover, the duality (Rainich, 1925; Misner and Wheeler, 1957) by the complexion $`\alpha `$, namely
$$\overline{F}^{\mu \nu }=F^{\mu \nu }\mathrm{cos}\alpha +F^{\mu \nu }\mathrm{sin}\alpha $$
(7)
is equivalent to a Touschek-Nishijima (Touschek, 1957; Nishijima, 1957) transformation for the spinor $`\mathrm{\Psi }`$ to the spinor $`\mathrm{\Psi }^{}`$ given by
$$\mathrm{\Psi }^{}=e^{\gamma ^5\alpha /2}\mathrm{\Psi }$$
(8)
with
$$e^{\gamma ^5\alpha }=\mathrm{cos}\alpha +\gamma ^5\mathrm{sin}\alpha $$
(9)
and
$`\mathrm{cos}\alpha `$ $`=`$ $`{\displaystyle \frac{\overline{\mathrm{\Psi }}\mathrm{\Psi }}{\rho }}`$ (10)
$`\mathrm{sin}\alpha `$ $`=`$ $`{\displaystyle \frac{\overline{\mathrm{\Psi }}\gamma ^5\mathrm{\Psi }}{\rho }}`$ (11)
$`\rho `$ being the positive square root of
$$\rho ^2=(\overline{\mathrm{\Psi }}\mathrm{\Psi })^2+(\overline{\mathrm{\Psi }}\gamma ^5\mathrm{\Psi })^2$$
(12)
Campolattaro showed that the two spinor Maxwell equations (5) and (6) is equivalent to a single nonlinear first-order equation for the spinor, namely
$$\gamma ^\mu \mathrm{\Psi }_{,\mu }=i\gamma ^\mu \frac{e^{\gamma ^5\alpha }}{\rho }\{Im(\overline{\mathrm{\Psi }}_{,\mu }\mathrm{\Psi })j_\mu \gamma ^5Im(\overline{\mathrm{\Psi }}_{,\mu }\gamma ^5\mathrm{\Psi })\}\mathrm{\Psi }$$
(13)
The relation between Dirac and Maxwell equations was also considered by Vaz and Rodrigues (1993).
On the other hand, Witten (1994) introduced the Seiberg-Witten equations. For more details, see, e. g., (Moore, 1996; Morgan, 1996). By counting the solutions of the equations with an Abelian gauge group, new invariants of 4-manifolds can be obtained. These invariants are closely related to Donaldson’s polynomial invariants, but in many respects much simpler to work with. One of the Seiberg-Witten equation is the Dirac equation on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$,
$$\stackrel{~}{D}M=\gamma _i\stackrel{~}{_i}M=0$$
(14)
where the Riemannian indices $`i=1,2,3,4`$. $`\stackrel{~}{D}`$ is the Dirac operator on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$. $`MS^+`$, $`S^+`$ is the positive chirality spinor bundle over $`X`$. $`\stackrel{~}{_i}`$ is a covariant derivative acting on $`S^+`$. $`\gamma _i`$, the Clifford matrices satisfy
$$\gamma _i\gamma _j+\gamma _j\gamma _i=2\delta _{ij}$$
(15)
Locally,
$`S^+`$ $`=`$ $`S_0^+L^{\frac{1}{2}}`$ (16)
$`\stackrel{~}{_i}`$ $`=`$ $`_i1+1_i^{^{}}`$ (17)
Here $`S_0^+`$ is the local positive chirality spinor bundle over $`X`$, $`L^{\frac{1}{2}}`$ is the square root of the line bundle $`L`$ over $`X`$, $`_i=_i+\mathrm{\Gamma }_i`$, $`\mathrm{\Gamma }_i`$ is the induced Levi-Civita connection on $`S_0^+`$. $`_i^{^{}}=_i+ia_i,`$ $`a_i`$ is the connection on $`L^{\frac{1}{2}}`$ over $`X`$. (For the later use, notice that when the covariant derivative $`_i=_i+\mathrm{\Gamma }_i`$ acts on the $`\gamma ^,`$s or tensors on the Riemannian manifold $`X`$, $`\mathrm{\Gamma }_i`$ is the Levi-Civita connection.)
The second Seiberg-Witten equation is
$$F_{ij}^+=M,S_{ij}M$$
(18)
Here $`,`$ represents Hermite inner product, it is the pointwise inner product. $`S_{ij}=\frac{i}{4}[\gamma _i,\gamma _j]`$ is the spin operater on $`X`$, and $`F_{ij}^+`$ is the self dual part of the curvature on the line bundle over $`X`$. However, Witten did not tell us where the second equation comes from. Obviously, a profound understanding of the Seiberg-Witten equations must help to promote both the applications to physics and further generalizations.
In this paper, we shall generalize Campolattaro’s viewpoint and set up the correspondence between the spinor and the curvature on the line bundle over the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold. We also set up the correspondence between the spinor and the $`2`$–differential form on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold. Seiberg-Witten equations and the perturbed Seiberg-Witten equations can then be well understood from this “direct” physical point of view.
We further suggest that the second Seiberg-Witten equation (18) (also the second perturbed Seiberg-Witten equation) can be considered as a nonlinear Dirac-like equation.
Since the connection is more basic than the curvature, we shall derive the spinor representation of the connection on the associated unitary line bundle on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold.
## 2 SPINOR REPRESENTATION OF THE CURVATURE ON THE LINE BUNDLE OVER $`Spin_4^c`$ MANIFOLD
### 2.1 THE CASE OF CURVED SPACE–TIME
In order to obtain the spinor representation of the curvature of a unitary connection on the line bundle over Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold, we first consider the case when the space–time $`X`$ is a manifold with a pseudo-Riemannian metric of Lorentz signature. As described in more detail in (Misner, Thorne and Wheeler,1973), the Maxwell equations in the curved space–time are
$`F_{\mu \nu ,\gamma }+F_{\nu \gamma ,\mu }+F_{\gamma \mu ,\nu }`$ $`=`$ $`0`$ (19)
$`F_{;\nu }^{\mu \nu }`$ $`=`$ $`j^\mu `$ (20)
Where $`\mu ,\nu ,\gamma =0,1,2,3.`$ These equations have the same form as in the case of Minkowski space–time, but with a comma replaced by a semicolon in Eq.(20). Here the semicolon stands for the covariant derivative.
One can naturally assume that there exists a spinor $`\mathrm{\Psi }`$ on the curved space–time $`X`$, such that
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}S^{\mu \nu }\mathrm{\Psi }`$ (21)
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}\gamma ^5S^{\mu \nu }\mathrm{\Psi }`$ (22)
where $`\mu ,\nu =0,1,2,3.`$ Notice that at this time $`\mathrm{\Psi }`$ and $`\gamma ^\mu `$ is defined on the curved space-time.
We can now ask the question: When is $`iF`$ the curvature of a unitary connction on some line bundle over the curved space-time $`X`$ with the Hermitian metric?
One necessary condition is that $`F`$ satisfy the Bianchi identity $`dF=0`$, This is just one of the Maxwell equations (19). The other is the first Chern class $`c_1(L)`$ of a line bundle $`L`$ over $`X`$ is “quantized” – $`c_1(L)`$ integrates to an integer over any two dimensional cycle in $`X`$. This fact can be interpreted as requiring quantization of electric charge. Maintaining a perfect duality between electric and magnetic fields would then require quantization of magnetic charge as well.
From now on, we suppose that both electric and magnetic charges are quantized. The second condition is then automatically satisfied. We claim that these two conditions are also sufficient.
Now Eq.(21) are automatically the spinor representation of the curvature on the line bundle $`L`$ over the curved space–time $`X`$.
### 2.2 THE CASE OF $`Spin_4^c`$ MANIFOLD
Let $`X`$ be an oriented, closed four-dimensional Riemannian manifold. A Spin<sup>c</sup> structure exists on any oriented four-manifold (Hirzebruch and Hopf, 1958; Lawson and Michelson, 1989).
The curvature $`F`$ of a unitary connection on the line bundle $`L`$ over Spin$`{}_{4}{}^{}{}_{}{}^{c}`$-manifold $`X`$ satisfies the Bianchi identity,
$$dF=0$$
(23)
We also denote
$$_jF_{ij}=j_i$$
(24)
Here the covariant derivative $`_j=_j+\mathrm{\Gamma }_j`$, $`\mathrm{\Gamma }_j`$ is the Levi-Civita connection on $`X`$.
From the previous discussions, we have naturally the spinor representation of the curvature $`F`$,
$`F_{ij}`$ $`=`$ $`W^{}S_{ij}W`$ (25)
$`=`$ $`W,S_{ij}W`$
$`F_{ij}`$ $`=`$ $`W,\gamma _5S_{ij}W`$ (26)
Here $`W`$ is a spinor on $`X`$, $`i,j=1,2,3,4`$.
## 3 SPINOR REPRESENTATION OF $`2`$–FORMS ON THE $`Spin_4^c`$ MANIFOLD
### 3.1 THE CASE OF MINKOWSKI SPACE–TIME
Campolattaro (1990a, 1990b) assumed that together with an electric current $`j_\mu `$, there exists also a magnetic monopole current $`g_\mu .`$ Maxwell equations read
$`(F^{\mu \nu })_{,\nu }`$ $`=`$ $`g^\mu `$ (27)
$`F_{,\nu }^{\mu \nu }`$ $`=`$ $`j^\mu `$ (28)
There exists a spinor, such that
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}S^{\mu \nu }\mathrm{\Psi }`$ (29)
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}\gamma ^5S^{\mu \nu }\mathrm{\Psi }`$ (30)
It was shown that the spinor equation Eq.(13) in the presence of magnetic monopoles, reads
$$\gamma ^\mu \mathrm{\Psi }_{,\mu }=i\gamma ^\mu \frac{e^{\gamma ^5\alpha }}{\rho }\{Im(\overline{\mathrm{\Psi }}_{,\mu }\mathrm{\Psi })j_\mu \gamma ^5[Im(\overline{\mathrm{\Psi }}_{,\mu }\gamma ^5\mathrm{\Psi })g_\mu ]\}\mathrm{\Psi }$$
(31)
### 3.2 THE CASE OF CURVED SPACE–TIME
Given $`X`$, a four-dimensional manifold with a pseudo-Rimaniann metric of Lorentz signature. The generalized Maxwell equations read
$`(F^{\mu \nu })_{;\nu }`$ $`=`$ $`g^\mu `$ (32)
$`F_{;\nu }^{\mu \nu }`$ $`=`$ $`j^\mu `$ (33)
One has the same expressions as in the previous discussions,
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}S^{\mu \nu }\mathrm{\Psi }`$ (34)
$`F^{\mu \nu }`$ $`=`$ $`\overline{\mathrm{\Psi }}\gamma ^5S^{\mu \nu }\mathrm{\Psi }`$ (35)
Notice that at this time $`\gamma ^\mu `$ and $`\mathrm{\Psi }`$ are defined in the curved space–time.
From Eq.(32), we state that the Bianchi identity is no longer satisfied: $`F^{}`$ is only a $`2`$–differential form on $`X`$. Given a $`2`$-form $`F^{}`$ one has
$$F^{}=F+\omega $$
(36)
Here $`F`$ is the curvature of a unitary connection on some line bundle over $`X`$, $`\omega `$ is a $`2`$-form over $`X`$. Later we shall adopt this notation.
### 3.3 THE CASE OF $`Spin_4^c`$ MANIFOLD
Denote $`X`$ a Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold. Given a $`2`$-form $`F+\omega `$ on $`X`$. Denote
$`_j(F_{ij}+\omega _{ij})`$ $`=`$ $`g_i`$ (37)
$`_j(F_{ij}+\omega _{ij})`$ $`=`$ $`j_i`$ (38)
We have natually the spinor representation
$`F_{ij}+\omega _{ij}`$ $`=`$ $`W,S_{ij}W`$ (39)
$`F_{ij}+\omega _{ij}`$ $`=`$ $`W,\gamma _5S_{ij}W`$ (40)
Here $`W`$ is a spinor on $`X`$, $`i,j=1,2,3,4`$.
## 4 NONLINEAR DIRAC-LIKE EQUATION AND THE SECOND SEIBERG-WITTEN EQUATION
From Eqs.(25) and (26), The self-dual part of the curvature on $`L`$ over the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$ is
$`F_{ij}^+`$ $`=`$ $`{\displaystyle \frac{1}{2}}(F_{ij}+F_{ij})`$ (41)
$`=`$ $`W,{\displaystyle \frac{1+\gamma _5}{2}}S_{ij}W`$
Denote $`\frac{1+\gamma _5}{2}W=M`$, Then
$$F_{ij}^+=M,S_{ij}M$$
(42)
Eq.(42) is just the second Seiberg-Witten equation (18). From Eqs.(23) and (24), one has
$$_iF_{ji}^+=_iM,S_{ji}M=\frac{1}{2}j_j$$
(43)
Notice that $`\stackrel{~}{_i}S_{ji}=_iS_{ji}`$. Eq.(42) has the equivalent form
$$ImM,\gamma _j\stackrel{~}{D}M=M,_iS_{ji}MImM,\stackrel{~}{_j}M\frac{1}{2}j_j$$
(44)
Just as Campolattaro’s considerations, given $`F_{ij}^+`$ thus $`j_j`$, one can verify that the positive chirality spinor $`M`$ satisfy the nonlinear Dirac-like equation on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$
$$\stackrel{~}{D}M=\frac{i}{M,M}\{M,_kS_{ik}MImM,\stackrel{~}{_i}M\frac{1}{2}j_i\}\gamma _iM$$
(45)
Since the term $`M,\gamma _j\gamma _iM`$ is pure imaginary, if $`ij`$. One can show that Eq.(45) is the sufficient condition of Eq.(44), thus is the sufficient condition of the second Seiberg-Witten equation (42).
Notice that Seiberg-Witten equations
$`\stackrel{~}{D}M`$ $`=`$ $`0`$
$`F_{ij}^+`$ $`=`$ $`M,S_{ij}M`$
are equivalent to
$`\stackrel{~}{D}M`$ $`=`$ $`0`$
$`\stackrel{~}{D}M`$ $`=`$ $`{\displaystyle \frac{i}{M,M}}\{M,_kS_{ik}MImM,\stackrel{~}{_i}M{\displaystyle \frac{1}{2}}j_i\}\gamma _iM`$
From this point of view, the second Seiberg-Witten equation can be replaced by the nonlinear Dirac-like equation (45).
## 5 NONLINEAR DIRAC-LIKE EQUATION AND THE PERTURBED SECOND SEIBERG-WITTEN EQUATION
From Eqs.(39) and (40), write $`\varphi =\omega ^+=\frac{1}{2}(\omega +\omega )`$, one has
$$F_{ij}^++\varphi _{ij}=M,S_{ij}M$$
(46)
Eq.(46) is just the second perturbed Seiberg-Witten equation. From Eqs.(37) and (38), we have
$$_iM,S_{ji}M=\frac{j_j+g_j}{2}$$
(47)
Just as the previous discussions, given $`j_j+g_j`$, we have a nonlinear Dirac-like equation on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$,
$$\stackrel{~}{D}M=\frac{i}{M,M}\{M,_kS_{ik}MImM,\stackrel{~}{_i}M\frac{j_i+g_i}{2}\}\gamma _iM$$
(48)
This equation is the sufficient condition of the second perturbed Seiberg-Witten equation (46). Notice that the perturbed Seiberg-Witten equations
$`\stackrel{~}{D}M`$ $`=`$ $`0`$
$`F_{ij}^++\varphi _{ij}`$ $`=`$ $`M,S_{ij}M`$
are equivalent to
$`\stackrel{~}{D}M`$ $`=`$ $`0`$
$`\stackrel{~}{D}M`$ $`=`$ $`{\displaystyle \frac{i}{M,M}}\{M,_kS_{ik}MImM,\stackrel{~}{_i}M{\displaystyle \frac{j_i+g_i}{2}}\}\gamma _iM`$
So one can say that it is important to study the nonlinear Dirac-like equations (45) and (48).
## 6 SPINOR REPRESENTATION OF $`U(1)`$ CONNECTION ON $`Spin_4^c`$ MANIFOLD
Now we derive the spinor representation of the connection $`A`$ on the line bundle $`L`$ over Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold $`X`$. Choose a positive chirality spinor $`M`$, which is the solution of the Dirac equation
$$\gamma _i\stackrel{~}{_i}M=0$$
(49)
One has
$$M,\gamma _j\gamma _i\stackrel{~}{_i}M=0$$
(50)
which is equal to
$$M,S_{ji}\stackrel{~}{_i}M\frac{i}{2}M,\stackrel{~}{_j}M=0$$
(51)
By taking the Hermitian conjugation of Eq.(51), one has
$$\stackrel{~}{_i}M,S_{ji}M+\frac{i}{2}\stackrel{~}{_j}M,M=0$$
(52)
Notice that $`\stackrel{~}{_i}S_{ji}=_iS_{ji}`$. By adding Eqs.(51) and (52), one obtains,
$$_iM,S_{ji}M=M,(_iS_{ji})MImM,\stackrel{~}{_j}M$$
(53)
Eq.(53) is completely equivalent to the Dirac Eq.(49).
Since the computation is local, we can write
$$M=\mathrm{\Psi }\lambda $$
(54)
where $`\mathrm{\Psi }S_0^+`$ is a local positive chirality spinor on $`X`$, and $`\lambda L^{\frac{1}{2}}`$. Notice that
$$\mathrm{\Psi }_1\lambda _1,\mathrm{\Psi }_2\lambda _2=\mathrm{\Psi }_1,\mathrm{\Psi }_2\lambda _1,\lambda _2$$
(55)
Eq.(53) reads
$`_i[\mathrm{\Psi },S_{ji}\mathrm{\Psi }\lambda ,\lambda ]`$ $`=`$ $`\mathrm{\Psi },(_iS_{ji})\mathrm{\Psi }\lambda ,\lambda `$ (56)
$`Im\mathrm{\Psi }\lambda ,_j\mathrm{\Psi }\lambda +\mathrm{\Psi }_j^{^{}}\lambda `$
Here $`_j=_j+\mathrm{\Gamma }_j`$. $`_j^{^{}}=_j+ia_j`$, $`a_j`$ is the connection on $`L^{\frac{1}{2}}`$. We have
$`a_j`$ $`=`$ $`{\displaystyle \frac{Im\mathrm{\Psi },_j\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}+{\displaystyle \frac{\mathrm{\Psi },(_iS_{ji})\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}`$ (57)
$`{\displaystyle \frac{_i[\mathrm{\Psi },S_{ji}\mathrm{\Psi }\lambda ,\lambda ]}{\mathrm{\Psi },\mathrm{\Psi }\lambda ,\lambda }}{\displaystyle \frac{Im\lambda ,_j\lambda }{\lambda ,\lambda }}`$
Since $`L`$ is unitary, i.e., $`\lambda ,\lambda =1.`$ One has
$$Im\lambda ,_j\lambda =i\lambda ^1_j\lambda $$
The transformation rule of $`a_j`$ is
$$ia_j^{}=ia_j+\lambda ^1_j\lambda $$
(58)
We also denote $`a^{}`$ by $`a`$. It leads to
$`a_j`$ $`=`$ $`{\displaystyle \frac{Im\mathrm{\Psi },_j\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}+{\displaystyle \frac{\mathrm{\Psi },(_iS_{ji})\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}{\displaystyle \frac{_i\mathrm{\Psi },S_{ji}\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}`$ (59)
$`=`$ $`{\displaystyle \frac{Im\mathrm{\Psi },\gamma _jD\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}}.`$
Here $`D=\gamma _i_i`$ is the Dirac operator acting on the local positive chirality spinor $`\mathrm{\Psi }`$. This means that the Dirac equation (49) can be locally reduced to
$$\gamma _j(_j+ia_j)\mathrm{\Psi }=0$$
(60)
Notice that $`da=\frac{1}{2}F`$. For any unitary connection $`A`$ on $`L`$, $`F=dA`$. we can choose
$$A=2a$$
(61)
Finally, one has the spinor representation of $`A`$,
$$A_j=2\frac{Im\mathrm{\Psi },\gamma _jD\mathrm{\Psi }}{\mathrm{\Psi },\mathrm{\Psi }}.$$
(62)
Where $`j=1,2,3,4.`$
## 7 CONCLUSIONS
We have analysed the relation between the spinor and the curvature on a unitary line bundle over the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold. The relation between the spinor and the $`2`$–form on the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold has also been considered.
We have proposed two nonlinear Dirac-like equations which are equivalent to the second Seiberg-Witten equation and the second perturbed Seiberg-Witten equation respectively. This means one can study (the self dual part of) the curvature (or the $`2`$-form ) in the quantum mechanics level.
We have also derived the local spinor representation of the unitary connection on the associated line bundle over the Spin$`{}_{4}{}^{}{}_{}{}^{c}`$ manifold.
These discussions can be generalized to the non-Abelian cases.
The relation between the properties of the nonlinear Dirac-like equation and the moduli space of the (perturbed) Seiberg-Witten equations need to be further studied.
ACKNOWLEDGMENTS
One of us (L. Z. Hu) woud like to thank Profs. M.-L. Ge and Y. S. Wu for valuable discussions. This work is supported in part by the Research Center of Mathematics, the National Education Committee of China. |
warning/0001/hep-ph0001063.html | ar5iv | text | # A proposal on the search for the hybrid with 𝐼^𝐺(𝐽^{𝑃𝐶})=1⁻(1⁻⁺) in the process 𝐽/𝜓→𝜌𝜔𝜋𝜋 at upgraded BEPC/BES 11footnote 1The project is supported by the National Natural Science Foundation of China under Grant No. 19991487, and Grant No. LWTZ-1298 of the Chinese Academy of Sciences
## I. Introduction
Apart from the ordinary $`q\overline{q}`$ mesons, the new hadronic states such as glueballs ($`gg/ggg`$), hybrids ($`q\overline{q}g)`$ and four-quark states ($`qq\overline{q}\overline{q}`$) also exist according to the predictions of QCD. Discovery and confirmation of any one of these new hadronic states would be the strong support to the QCD theory. Therefore, the search for and identifying these new hadronic states is a very excited and attractive research subject both theoretically and experimentally.
These new hadronic states can have the same quantum number $`J^{PC}`$ as the ordinary $`q\overline{q}`$ mesons, what’s more, they can also have the exotic quantum number $`J^{PC}`$ which are not allowed in the quark model such as $`1^+`$, and thus they can not mix with the ordinary mesons. Experimentally, GAMS collaboration, E179 collaboration at KEK, VES group, E852 collaboration at BNL and Crystal Barrel all claimed that the evidence for the exotic state with $`J^{PC}=1^+`$ was observed. The observed $`\rho \pi `$, $`\eta \pi `$ and $`\eta ^{}\pi `$ couplings of this state qualitatively support the hypothesis that it is a hybrid meson, although other interpretations cannot be eliminated.
In terms of the predictions of the lattice QCD, the lowest lying glueball with $`J^{PC}=1^+`$ has a higher mass than $`J/\psi `$. Bag model calculations predict that the lowest lying $`qq\overline{q}\overline{q}`$ states do not carry exotic quantum numbers and form nonets carrying the same quantum numbers as $`q\overline{q}`$ nonets, and that most $`qq\overline{q}\overline{q}`$ states can fall apart into two mesons and thus have a decay width in the order of their mass, which leads to that most $`qq\overline{q}\overline{q}`$ states are expected to be essentially unconfined and will not be observed as resonance peaks with reasonably narrow widths. Therefore, the search for the gluballs and four-quark states with $`J^{PC}=1^+`$ at BEPC/BES could be disappointing. However, lattice QCD predicts that the mass of the hybrid with $`J^{PC}=1^+`$ is $`1.22.5`$ GeV. In addition, the naive estimate of pQCD predicts that the $`J/\psi `$ hadronic decay processes are favorable to the production of hybrids. So, if the hybrids exist, the search for hybrid with $`J^{PC}=1^+`$ at BEPC/BES should be fairly hopeful.
H. Yu and Q.X. Shen have already discussed the possibility of the search for the hybrid with $`J^{PC}=1^+`$ in the process $`J/\psi \rho X`$, $`X\eta \pi (\eta ^{}\pi ,\rho \pi )`$. For the decay modes of the hybrid with $`I^G(J^{PC})=1^{}(1^+)`$, according to the symmetrization selection rule, the $`\eta \pi `$, $`\eta ^{}\pi `$ modes are strongly suppressed (A possible mechanism to explain why the $`1^+`$ state was observed in the above suppressed decay channels is planned for separated publication). The $`\rho \pi `$ mode is allowed, but this is a P-wave mode and thus the $`\rho \pi `$ mode should not be a dominant decay mode. The dominant decay mode should be the $`b_1(1235)\pi `$. Therefore, the probability of discovering the hybrid with $`J^{PC}=1^+`$ in the process $`J/\psi \rho X`$, $`Xb_1\pi `$, in principle, should be higher than that in the process $`J/\psi \rho X`$, $`X\eta \pi (\eta ^{}\pi ,\rho \pi )`$. Also, since the dominant decay mode of $`b_1`$ is $`\omega \pi `$, compared to the study on the two-step two-body decay process of $`J/\psi `$ in Ref. , the study on the three-step two-body decay process $`J/\psi \rho X`$, $`Xb_1\pi `$, $`b_1\omega \pi `$ perhaps could present more information to the experimentists. In this work, we shall consider the process $`J/\psi \rho X`$, $`Xb_1\pi `$, $`b_1\omega \pi `$.
This work is organized as follows. In Sect. II, we give the moment expressions for the resonances $`X`$ with the above spin-parity in the process $`J/\psi \rho X`$, $`Xb_1\pi `$, $`b_1\omega \pi `$ in terms of the generalized moment analysis method. , and in Sect. III, we discuss how to identify the resonances $`X`$ with different spin-parity. Our conclusion is reached in Sect. IV.
## II. Moment analysis
We consider the process
$$e^++e^{}J/\psi \rho +X,Xb_1+\pi ,b_1\omega +\pi .$$
(1)
The S matrix element of the process (1) can be written as
$$\rho _{\lambda _\rho }\omega _{\lambda _\omega }\pi \pi |S1|e_r^+e_r^{}^{}\psi _{\lambda _J}|T|e_r^+e_r^{}^{}\rho _{\lambda _\rho }X_{\lambda _X}|T_1|\psi _{\lambda _J}b_{1\lambda _{b_1}}\pi |T_2|X_{\lambda _X}\omega _{\lambda _\omega }\pi |T_3|b_{1\lambda _{b_1}},$$
(2)
where
$$\psi _{\lambda _J}|T|e_r^+e_r^{}^{}e_\mu ^{\lambda _J}(\stackrel{}{p}_J)\overline{v}_r(\stackrel{}{p}_+)\gamma ^\mu u_r^{}(\stackrel{}{p}_{});$$
(3)
$$\rho _{\lambda _\rho }X_{\lambda _X}|T_1|\psi _{\lambda _J}A_{\lambda _\rho ,\lambda _X}^{J_X}D_{\lambda _J,\lambda _\rho \lambda _X}^1(0,\theta _\rho ,0);$$
(4)
$$b_{1\lambda _{b_1}}\pi |T_2|X_{\lambda _X}B_{\lambda _{b_1}}^{J_X}D_{\lambda _X,\lambda _{b_1}}^{J_X}(\varphi _1,\theta _1,\varphi _1);$$
(5)
$$\omega _{\lambda _\omega }\pi |T_3|b_{1\lambda _{b_1}}C_{\lambda _\omega }D_{\lambda _{b_1},\lambda _\omega }^1(\varphi _2,\theta _2,\varphi _2);$$
(6)
And $`\lambda _J`$, $`\lambda _\rho `$, $`\lambda _X`$, $`\lambda _{b_1}`$ and $`\lambda _\omega `$ are the helicities of $`J/\psi `$, $`\rho `$, $`X`$, $`b_1`$ and $`\omega `$, respectively; $`r`$ and $`r^{}`$ are the polarization indexes of the positron and electron, respectively; $`\stackrel{}{p}_J`$, $`\stackrel{}{p}_+`$, $`\stackrel{}{p}_{}`$ are the momenta of $`J/\psi `$, positron and electron in the c.m. system of $`e^+e^{}`$, respectively; $`A_{\lambda _\rho ,\lambda _X}^{J_X}`$, $`B_{\lambda _{b_1}}^{J_X}`$ and $`C_{\lambda _\omega }`$ are the helicity amplitudes of the processed $`J/\psi \rho X`$, $`Xb_1\pi `$ and $`b_1\omega \pi `$, respectively; $`\theta _\rho `$ is the polar angle in the c.m. system of $`e^+e^{}`$ in which $`z`$ axis is chosen to be along the direction of the incident positron and the vector meson $`\rho `$ lies in $`xz`$ plane; $`(\theta _1,\varphi _1)`$ describes the direction of the momentum of $`b_1`$ in the rest frame of $`X`$ where the $`z_1`$ axis is chosen to be along the direction of the momentum of $`X`$ in the c.m. system of $`e^+e^{}`$; Similarly, $`(\theta _2,\varphi _2)`$ described the direction of the momentum of the vector mesons $`\omega `$ in the rest frame of $`b_1`$ where the $`z_2`$ axis is along the momentum of $`b_1`$ in the rest frame of $`X`$; The function $`D_{m,n}^J`$ is the $`(2J+1)`$-dimensional representation of the rotation group. Owing to the parity conservation for the process (1), these helicity amplitudes satisfy the following symmetry relations:
$`A_{\lambda _\rho ,\lambda _X}^{J_X}=P_X(1)^{J_X}A_{\lambda _\rho ,\lambda _X}^{J_X},`$
$`B_{\lambda _{b_1}}^{J_X}=P_X(1)^{J_X}B_{\lambda _{b_1}}^{J_X},`$
$`C_{\lambda _\omega }=C_{\lambda _\omega },`$ (7)
where $`P_X`$ is the parity of $`X`$.
The angular distribution for the process (1) is
$`W(\theta _\rho ,\theta _1,\varphi _1,\theta _2,\varphi _2)`$
$`{\displaystyle \underset{\lambda _J,\lambda _J^{}}{}}{\displaystyle \underset{\lambda _X,\lambda _X^{}}{}}{\displaystyle \underset{\lambda _{b_1},\lambda _{b_1}^{}}{}}{\displaystyle \underset{\lambda _\rho ,\lambda _\omega }{}}I_{\lambda _J,\lambda _J^{}}A_{\lambda _\rho ,\lambda _X}^{J_X}A_{\lambda _\rho ,\lambda _X^{}}^{J_X}B_{\lambda _{b_1}}^{J_X}B_{\lambda _{b_1}^{}}^{J_X}C_{\lambda _\omega }C_{\lambda _\omega }^{}`$
$`\times D_{\lambda _J,\lambda _\rho \lambda _X}^1(0,\theta _\rho ,0)D_{\lambda _J^{},\lambda _\rho \lambda _X^{}}^1(0,\theta _\rho .0)`$
$`\times D_{\lambda _X,\lambda _{b_1}}^{J_X}(\varphi _1,\theta _1,\varphi _1)D_{\lambda _X^{},\lambda _{b_1}^{}}^{J_X}(\varphi _1,\theta _1,\varphi _1)`$
$`\times D_{\lambda _{b_1},\lambda _\omega }^1(\varphi _2,\theta _2,\varphi _2)D_{\lambda _{b_1}^{},\lambda _\omega }^1(\varphi _2,\theta _2,\varphi _2),`$ (8)
where the density matrix elements $`I_{\lambda _J,\lambda _J^{}}`$ is
$`I_{\lambda _J,\lambda _J^{}}{\displaystyle \frac{1}{4}}{\displaystyle \underset{r,r^{}}{}}\psi _{\lambda _J}|T|e_r^+e_r^{}^{}\psi _{\lambda _J^{}}|T|e_r^+e_r^{}^{}^{}2|\stackrel{}{p}_+|^2\delta _{\lambda _J,\lambda _J^{}}\delta _{\lambda _J,\pm 1}.`$ (9)
The moments for the process (1) can be defined by
$`M(j,L,M,\mathrm{},m)=`$
$`{\displaystyle 𝑑\theta _\rho \mathrm{sin}\theta _\rho d\theta _1\mathrm{sin}\theta _1d\varphi _1d\theta _2\mathrm{sin}\theta _2d\varphi _2W(\theta _\rho ,\theta _1,\varphi _1,\theta _2,\varphi _2)}`$
$`\times D_{0,M}^j(0,\theta _\rho ,0)D_{M,m}^L(\varphi _1,\theta _1,\varphi _1)D_{m,0}^{\mathrm{}}(\varphi _2,\theta _2,\varphi _2).`$ (10)
Eq. (10) can be reduced to
$`M(j,L,M,\mathrm{},m)`$
$`{\displaystyle \underset{\lambda _J=\pm 1}{}}{\displaystyle \underset{\lambda _X,\lambda _X^{}}{}}{\displaystyle \underset{\lambda _{b_1},\lambda _{b_1}^{}}{}}{\displaystyle \underset{\lambda _\rho ,\lambda _\omega }{}}A_{\lambda _\rho ,\lambda _X}^{J_X}A_{\lambda _\rho ,\lambda _X^{}}^{J_X}B_{\lambda _{b_1}}^{J_X}B_{\lambda _{b_1}^{}}^{J_X}C_{\lambda _\omega }C_{\lambda _\omega }^{}`$
$`\times 1\lambda _Jj0|1\lambda _J1(\lambda _\rho \lambda _X^{})j(M)|1(\lambda _\rho \lambda _X)`$
$`\times J_X\lambda _X^{}LM|J_X\lambda _XJ_X\lambda _{b_1}^{}Lm|J_X\lambda _{b_1}`$
$`\times 1\lambda _{b_1}^{}\mathrm{}m|1\lambda _{b_1}1\lambda _\omega \mathrm{}0|1\lambda _\omega ,`$ (11)
where $`j_1m_1j_2m_2|j_3m_3`$ is Clebsch-Gordan coefficients.
In the process $`Xb_1\pi `$, if we restrict $`\mathrm{}_f1`$, where $`\mathrm{}_f`$ is the relative orbital angular momentum between $`b_1`$ and $`\pi `$, the quantum number $`I^G(J_X^{P_XC})`$ of $`X`$ allowed by the parity-isospin conservation law in the process (1) are $`1^{}(1^+)`$, $`1^{}(0^{++})`$, $`1^{}(1^{++})`$, and $`1^{}(2^{++})`$. For the resonances $`X`$ with $`J_X^{P_XC}=0^{++},1^+,1^{++}`$, and $`2^{++}`$, the nonzero moment expressions derived from Eq. (11) are shown in Appendix A, B, C, and D.
There are four, twenty-one, sixteen, and thirty-one nonzero moment expressions for $`J_X^{P_XC}=0^{++}`$, $`1^+`$, $`1^{++}`$, and $`2^{++}`$, respectively. In the following section, we shall discuss how to identify the $`X`$ with the above $`J_X^{P_XC}`$.
## III. Discussion
Since the helicity amplitudes $`|C_0|^2`$ and $`|C_1|^2`$ are independent of the spin-parity of the resonance $`X`$, we find that if $`|C_0|^2|C_1|^2`$ the moment expressions have the following characteristics: For $`J_X^{P_XC}=0^{++}`$, the moments is always equal to zero in the case $`L>0`$ or $`M>0`$ or $`m>0`$; For $`J_X^{P_XC}=1^{++}`$, the nonzero moments with $`L=0,1,2`$, $`M=0,1,2`$ and $`m=0,2`$ exist but the moments are zero in the case $`m=1`$; For $`J_X^{P_XC}=1^+`$, the nonzero moments with $`L=0,1,2`$, $`M=0,1,2`$ and $`m=0,2`$ exist, the nonzero moments with $`m=1`$ also exist; For $`J_X^{P_XC}=2^{++}`$, apart from the nonzero moments with $`L=0,1,2`$, $`M=0,1,2`$ and $`m=0,1,2`$, the nonzero moments with $`L=3,4`$ exist. Therefore, from these characteristics, we can easily identify the resonances $`X`$ with $`J_X^{P_XC}`$= $`0^{++}`$, $`1^+`$, $`1^{++}`$ and $`2^{++}`$ experimentally.
However, if $`|C_0|^2=|C_1|^2`$, some of the preceding characteristics disappear, which leads to that the situations in the case $`|C_0|^2=|C_1|^2`$ are more complex than those in the case $`|C_0|^2|C_1|^2`$. We will turn to the special case $`|C_0|^2=|C_1|^2`$ below.
(A) $`|C_0|^2=|C_1|^2`$, $`|B_0^1|^2|B_1^1|^2`$ and $`3|B_0^2|^24|B_1^2|^2`$
In this case, only for $`J_X^{P_XC}=2^{++}`$, there are four nonzero moments with $`L=4`$, so the resonance with $`J_X^{P_XC}=2^{++}`$ can be distinguished from other resonances. Then, for $`J_X^{P_XC}=0^{++}`$, there are only two nonzero moments with $`L=0`$, and for $`J_X=1`$, there are four nonzero moments with $`L=2`$, in addition to two nonzero moments with $`L=0`$, hence the resonance with $`J_X=0`$ can also be distinguished from that with $`J_X=1`$. Finally, to distinguish the resonance with $`J_X^{P_XC}=1^+`$ from that with $`J_X^{P_X}=1^{++}`$, we consider the following moment expression $`H\frac{1}{8}M(00000)\frac{5}{4}M(02000)\frac{5}{4}M(20000)+\frac{25}{2}M(22000)`$ and find that the $`H`$ satisfies
$$H\{\begin{array}{cc}0,& (J_X^{P_XC}=1^{++}),\\ \frac{27}{4}(|A_{00}^1|^2|B_0^1|^22|A_{11}^1|^2|B_0^1|^22|A_{00}^1|^2|B_1^1|^2)|C_1|^2,& (J_X^{P_XC}=1^+).\end{array}$$
(12)
Using Eq. (12), we can still distinguish the resonance $`X`$ with $`J_X^{P_XC}=1^+`$ from that with $`J_X^{P_XC}=1^{++}`$.
(B) $`|C_0|^2=|C_1|^2`$, $`|B_0^1|^2|B_1^1|^2`$ and $`3|B_0^2|^2=4|B_1^2|^2`$
In this case, compared to the case (A), the numbers of the nonzero moments for $`J_X^{P_XC}=0^{++}`$, $`1^+`$ and $`1^{++}`$ remain unchange, but for $`J_X^{P_XC}=2^{++}`$, the moments with $`L=4`$ disappear, There are still only two nonzero moments with $`L=0`$ for $`J_X^{P_XC}=0^{++}`$ and six nonzero moments with $`L=0,2`$ not only for $`J_X=1`$ but also for $`J_X=2`$. In this case, owing to Eq. (12) remains unchange and
$$H\frac{15}{2}|A_{00}^2|^2|B_1^2|^2|C_1|^2<0(J_X^{P_XC}=2^{++}),$$
(13)
the crucial point is to distinguish the resonance with $`J_X^{P_XC}=1^+`$ from that with $`J_X^{P_XC}=2^{++}`$. We also find $`H_1\frac{1}{4}M(02000)\frac{5}{2}M(22000)`$ satisfies
$$H_1\{\begin{array}{cc}3(|A_{00}^2|^2+|A_{11}^2|^2)|B_1^2|^2|C_1|^2>0,& (J_X^{P_XC}=2^{++}),\\ \frac{9}{5}|A_{11}^1|^2|B_1^1|^2|C_1|^2>0,& (J_X^{P_XC}=1^{++}),\\ \frac{9}{5}(|A_{00}^1|^2|A_{11}^1|^2)(|B_1^1|^2|B_0^1|^2)|C_1|^2,& (J_X^{P_XC}=1^+).\end{array}$$
(14)
So, if it is determined experimentally that $`H>0`$ or $`H_10`$, from Eq.(12)$``$(14), the $`J_X^{P_XC}`$ of $`X`$ must be $`1^+`$. However, if $`H<0`$ or $`H_1>0`$, we can not distinguish the resonance with $`J_X^{P_XC}=2^{++}`$ from that with $`J_X^{P_XC}=1^+`$.
(C) $`|C_0|^2=|C_1|^2`$ and $`|B_0^1|^2=|B_1^1|^2`$
In this case, there are only two nonzero moments with $`L=M=\mathrm{}=m=0`$ both for $`J_X^{P_XC}=0^{++}`$ and $`J_X^{P_XC}=1^+`$, there are two nonzero moments with $`L=0`$ and four nonzero moments with $`L=2`$ for $`J_X^{P_XC}=1^{++}`$, and there are two nonzero moments with $`L=0`$ and at least four nonzero moments with $`L=2`$ for $`J_X^{P_XC}=2^{++}`$. Therefore, the resonances with $`J_X^{P_XC}=1^{++}`$ and $`2^{++}`$ can be distinguished from the resonance with $`J_X^{P_XC}=0^{++}`$ ( or $`1^+`$ ). But it is almost impossible to distinguish the resonance with $`J_X^{P_XC}=1^+`$ from that with $`J_X^{P_XC}=0^{++}`$ except in the radiative $`J/\psi `$ decay process. Because for the radiative $`J/\psi `$ decay process $`e^++e^{}J/\psi \gamma X`$, $`Xb_1\pi `$, $`b_1\omega \pi `$, $`A_{00}^0=A_{01}^0=A_{00}^1=A_{01}^1`$, we find
$$M(00000)10M(20000)\{\begin{array}{cc}0,& (J_X^{P_XC}=0^{++}),\\ 108|A_{11}^1|^2|B_1^1|^2|C_1|^2,& (J_X^{P_XC}=1^+).\end{array}$$
(15)
Obviously, using Eq. (15) we can distinguish the $`0^{++}`$ state from the $`1^+`$ state in the radiative $`J/\psi `$ decay process.
From the above discussions, we get that if $`|C_0|^2|C_1|^2`$ we can easily identify the resonances $`X`$ with $`J_X^{P_XC}`$= $`0^{++}`$, $`1^+`$, $`1^{++}`$ and $`2^{++}`$, but if $`|C_0|^2=|C_1|^2`$ and $`|B_0^1|^2=|B_1^1|^2`$ ( or $`|C_0|^2=|C_1|^2`$ and $`3|B_0^2|^2=4|B_1^2|^2`$ ), the identification of the resonances $`X`$ with $`J_X^{P_XC}`$=$`1^+`$ and $`0^{++}`$ ( or $`2^{++}`$ ) is very difficult. However, we also want to note the following two points: 1) Since the ratio of the helicity amplitudes for the process $`b_1(1235)\omega \pi `$, $`|C_0|`$ and $`|C_1|`$, can be measured experimentally in other process such as $`J/\psi b_1\pi `$, $`b_1\omega \pi `$. The measurement of the ratio of $`|C_0|`$ and $`|C_1|`$ can be first performed in order to confirm whether $`|C_0|^2`$ is equal to $`|C_1|^2`$ or not; 2) Even though $`|C_0|^2=|C_1|^2`$, one could expect that the probability of the simultaneous appearance of $`|C_0|^2=|C_1|^2`$ and $`|B_0^1|^2=|B_1^1|^2`$ ( or $`|C_0|^2=|C_1|^2`$ and $`3|B_0^2|^2=4|B_1^2|^2`$ ) would be fairly small.
It is worth pointing out that the above moment expressions and the discussions are also valid for the process $`J/\psi \gamma X`$, $`Xb_1\pi `$, $`b_1\omega \pi `$ provided $`A_{00}^0=A_{01}^0=A_{00}^1=A_{01}^1=A_{00}^2=A_{01}^2=0`$.
## IV. Conclusion
The twenty-one nonzero moment expressions for $`J_X^{P_XC}=1^+`$ show the possibility of the resonance $`X`$ with $`J_X^{P_XC}=1^+`$ produced in the process (1) exists. At the same time, we can easily distinguish it from other resonances except for some rather special cases. Therefore, generally speaking, if the 50 million $`J/\psi `$ events in the upgraded BEPC/BES are obtained, the search for the hybrid with $`J^{PC}=1^+`$ in the process $`J/\psi \rho X`$, $`Xb_1\pi `$, $`b_1(1235)\omega \pi `$ is feasible.
## Appendix A: The nonzero moments for $`J_X^{P_XC}=0^{++}`$
$`M(00000)2(|A_{00}^0|^2+2|A_{10}^0|^2)|B_0^0|^2(|C_0|^2+2|C_1|^2),`$
$`M(00020){\displaystyle \frac{4}{5}}(|A_{00}^0|^2+2|A_{10}^0|^2)|B_0^0|^2(|C_0|^2|C_1|^2),`$
$`M(20000){\displaystyle \frac{2}{5}}(|A_{00}^0|^2|A_{10}^0|^2)|B_0^0|^2(|C_0|^2+2|C_1|^2),`$
$`M(20020){\displaystyle \frac{4}{25}}(|A_{00}^0|^2|A_{10}^0|^2)|B_0^0|^2(|C_0|^2|C_1|^2).`$
## Appendix B: The nonzero moments for $`J_X^{P_XC}=1^+`$
$`M(00000)2(|A_{00}^1|^2+2|A_{01}^1|^2+2|A_{10}^1|^2+2|A_{11}^1|^2)(|B_0^1|^2+2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(00020){\displaystyle \frac{4}{5}}(|A_{00}^1|^2+2|A_{01}^1|^2+2|A_{10}^1|^2+2|A_{11}^1|^2)(|B_0^1|^2|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(02000){\displaystyle \frac{4}{5}}(|A_{00}^1|^2|A_{01}^1|^2+2|A_{10}^1|^2|A_{11}^1|^2)(|B_0^1|^2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(02020){\displaystyle \frac{4}{25}}(|A_{00}^1|^2|A_{01}^1|^2+2|A_{10}^1|^2|A_{11}^1|^2)(2|B_0^1|^2+|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(02021){\displaystyle \frac{12}{25}}(|A_{00}^1|^2|A_{01}^1|^2+2|A_{10}^1|^2|A_{11}^1|^2)Re(B_1^1B_0^1)(|C_0|^2|C_1|^2),`$
$`M(02022){\displaystyle \frac{12}{25}}(|A_{00}^1|^2|A_{01}^1|^2+2|A_{10}^1|^2|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(20000){\displaystyle \frac{2}{5}}(|A_{00}^1|^2|A_{01}^1|^2|A_{10}^1|^2+2|A_{11}^1|^2)(|B_0^1|^2+2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(20020){\displaystyle \frac{4}{25}}(|A_{00}^1|^2|A_{01}^1|^2|A_{10}^1|^2+2|A_{11}^1|^2)(|B_0^1|^2|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(21121){\displaystyle \frac{6}{25}}Im(A_{01}^1A_{00}^1+A_{10}^1A_{11}^1)Im(B_1^1B_0^1)(|C_0|^2|C_1|^2),`$
$`M(22000){\displaystyle \frac{2}{25}}(2|A_{00}^1|^2+|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)(|B_0^1|^2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22020){\displaystyle \frac{2}{125}}(2|A_{00}^1|^2+|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)(2|B_0^1|^2+|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(22021){\displaystyle \frac{6}{125}}(2|A_{00}^1|^2+|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)Re(B_1^1B_0^1)(|C_0|^2|C_1|^2),`$
$`M(22022){\displaystyle \frac{6}{125}}(2|A_{00}^1|^2+|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22100){\displaystyle \frac{6}{25}}Re(A_{01}^1A_{00}^1A_{11}^1A_{10}^1)(|B_0^1|^2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22120){\displaystyle \frac{6}{125}}Re(A_{01}^1A_{00}^1A_{11}^1A_{10}^1)(2|B_0^1|^2+|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(22121){\displaystyle \frac{18}{125}}Re(A_{01}^1A_{00}^1A_{11}^1A_{10}^1)Re(B_1^1B_0^1)(|C_0|^2|C_1|^2),`$
$`M(22122){\displaystyle \frac{18}{125}}Re(A_{01}^1A_{00}^1A_{11}^1A_{10}^1)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22200){\displaystyle \frac{6}{25}}|A_{01}^1|^2(|B_0^1|^2|B_1^1|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22220){\displaystyle \frac{6}{125}}|A_{01}^1|^2(2|B_0^1|^2+|B_1^1|^2)(|C_0|^2|C_1|^2),`$
$`M(22221){\displaystyle \frac{18}{125}}|A_{01}^1|^2Re(B_1^1B_0^1)(|C_0|^2|C_1|^2),`$
$`M(22222){\displaystyle \frac{18}{125}}|A_{01}^1|^2|B_1^1|^2(|C_0|^2|C_1|^2).`$
## Appendix C: The nonzero moments for $`J_X^{P_XC}=1^{++}`$
$`M(00000)8(|A_{01}^1|^2+|A_{10}^1|^2+|A_{11}^1|^2)|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(00020){\displaystyle \frac{8}{5}}(|A_{01}^1|^2+|A_{10}^1|^2+|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(02000){\displaystyle \frac{4}{5}}(|A_{01}^1|^22|A_{10}^1|^2+|A_{11}^1|^2)|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(02020){\displaystyle \frac{4}{25}}(|A_{01}^1|^22|A_{10}^1|^2+|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(02022){\displaystyle \frac{12}{25}}(|A_{01}^1|^22|A_{10}^1|^2+|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(20000){\displaystyle \frac{4}{5}}(|A_{01}^1|^2+|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(20020){\displaystyle \frac{4}{25}}(|A_{01}^1|^2+|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22000){\displaystyle \frac{2}{25}}(|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(22020){\displaystyle \frac{2}{125}}(|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22022){\displaystyle \frac{6}{125}}(|A_{01}^1|^22|A_{10}^1|^22|A_{11}^1|^2)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22100){\displaystyle \frac{6}{25}}Re(A_{11}^1A_{10}^1)|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(22120){\displaystyle \frac{6}{125}}Re(A_{11}^1A_{10}^1)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22122){\displaystyle \frac{18}{125}}Re(A_{11}^1A_{10}^1)|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22200){\displaystyle \frac{6}{25}}|A_{01}^1|^2|B_1^1|^2(|C_0|^2+2|C_1|^2),`$
$`M(22220){\displaystyle \frac{6}{125}}|A_{01}^1|^2|B_1^1|^2(|C_0|^2|C_1|^2),`$
$`M(22222){\displaystyle \frac{18}{125}}|A_{01}^1|^2|B_1^1|^2(|C_0|^2|C_1|^2).`$
## Appendix D: The nonzero moments for $`J_X^{P_XC}=2^{++}`$
$`M(00000)2(|A_{00}^2|^2+2|A_{01}^2|^2+2|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)(|B_0^2|^2+2|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(00020){\displaystyle \frac{4}{5}}(|A_{00}^2|^2+2|A_{01}^2|^2+2|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)(|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(02000){\displaystyle \frac{4}{7}}(|A_{00}^2|^2+|A_{01}^2|^2+2|A_{10}^2|^2+|A_{11}^2|^22|A_{12}^2|^2)(|B_0^2|^2+|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(02020){\displaystyle \frac{4}{35}}(|A_{00}^2|^2+|A_{01}^2|^2+2|A_{10}^2|^2+|A_{11}^2|^22|A_{12}^2|^2)(2|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(02021){\displaystyle \frac{4\sqrt{3}}{35}}(|A_{00}^2|^2+|A_{01}^2|^2+2|A_{10}^2|^2+|A_{11}^2|^22|A_{12}^2|^2)Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(02022){\displaystyle \frac{12}{35}}(|A_{00}^2|^2+|A_{01}^2|^2+2|A_{10}^2|^2+|A_{11}^2|^22|A_{12}^2|^2)|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(04000){\displaystyle \frac{4}{63}}(3|A_{00}^2|^24|A_{01}^2|^2+6|A_{10}^2|^24|A_{11}^2|^2+|A_{12}^2|^2)(3|B_0^2|^24|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(04020){\displaystyle \frac{8}{315}}(3|A_{00}^2|^24|A_{01}^2|^2+6|A_{10}^2|^24|A_{11}^2|^2+|A_{12}^2|^2)(3|B_0^2|^2+2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(04021){\displaystyle \frac{4\sqrt{10}}{105}}(3|A_{00}^2|^24|A_{10}^2|^2+6|A_{10}^2|^24|A_{11}^2|^2+|A_{12}^2|^2)Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(04022){\displaystyle \frac{8\sqrt{15}}{315}}(3|A_{00}^2|^24|A_{10}^2|^2+6|A_{10}^2|^24|A_{11}^2|^2+|A_{12}^2|^2)|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(20000){\displaystyle \frac{2}{5}}(|A_{00}^2|^2|A_{01}^2|^2|A_{10}^2|^2+2|A_{11}^2|^2|A_{12}^2|^2)(|B_0^2|^2+2|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(20020){\displaystyle \frac{4}{25}}(|A_{00}^2|^2|A_{01}^2|^2|A_{10}^2|^2+2|A_{11}^2|^2|A_{12}^2|^2)(|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(21121){\displaystyle \frac{2}{25}}[3Im(A_{01}^2A_{00}^2+A_{10}^2A_{11}^2)+\sqrt{6}Im(A_{12}^2A_{11}^2)]Im(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(22000){\displaystyle \frac{2}{35}}(2|A_{00}^2|^2|A_{01}^2|^22|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)(|B_0^2|^2+|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22020){\displaystyle \frac{2}{175}}(2|A_{00}^2|^2|A_{01}^2|^22|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)(2|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(22021){\displaystyle \frac{2\sqrt{3}}{175}}(2|A_{00}^2|^2|A_{01}^2|^22|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(22022){\displaystyle \frac{6}{175}}(2|A_{00}^2|^2|A_{01}^2|^22|A_{10}^2|^2+2|A_{11}^2|^2+2|A_{12}^2|^2)|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(22100){\displaystyle \frac{\sqrt{2}}{35}}[\sqrt{6}Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)+6Re(A_{12}^2A_{11}^2)](|B_0^2|^2+|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22120){\displaystyle \frac{2}{175}}[\sqrt{3}Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)+3\sqrt{2}Re(A_{12}^2A_{11}^2)](2|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(22121){\displaystyle \frac{6}{175}}[Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)+\sqrt{6}Re(A_{12}^2A_{11}^2)]Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(22122){\displaystyle \frac{3\sqrt{2}}{175}}[\sqrt{6}Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)+6Re(A_{12}^2A_{11}^2)]|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(22200){\displaystyle \frac{2}{35}}[3|A_{01}^2|^2+2\sqrt{6}Re(A_{12}^2A_{10}^2)](|B_0^2|^2+|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(22220){\displaystyle \frac{2}{175}}[3|A_{01}^2|^2+2\sqrt{6}Re(A_{12}^2A_{10}^2)](2|B_0^2|^2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(22221){\displaystyle \frac{6}{175}}[\sqrt{3}|A_{01}^2|^2+2\sqrt{2}Re(A_{12}^2A_{10}^2)]Re(B_1^2B_0^2)](|C_0|^2|C_1|^2),`$
$`M(22222){\displaystyle \frac{6}{175}}[3|A_{01}^2|^2+2\sqrt{6}Re(A_{12}^2A_{10}^2)]|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(23121){\displaystyle \frac{6}{175}}[2Im(A_{01}^2A_{00}^2+A_{10}^2A_{11}^2)+\sqrt{6}Im(A_{11}^2A_{12}^2)]Im(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(23221){\displaystyle \frac{12\sqrt{5}}{175}}Im(A_{12}^2A_{10}^2)Im(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(24000){\displaystyle \frac{2}{315}}(6|A_{00}^2|^2+4|A_{01}^2|^26|A_{10}^2|^28|A_{11}^2|^2|A_{12}^2|^2)(3|B_0^2|^24|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(24020){\displaystyle \frac{4}{1575}}(6|A_{00}^2|^2+4|A_{01}^2|^26|A_{10}^2|^28|A_{11}^2|^2|A_{12}^2|^2)(3|B_0^2|^2+2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(24021){\displaystyle \frac{2\sqrt{10}}{525}}(6|A_{00}^2|^2+4|A_{01}^2|^26|A_{10}^2|^28|A_{11}^2|^2|A_{12}^2|^2)Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(24022){\displaystyle \frac{4\sqrt{15}}{1575}}(6|A_{00}^2|^2+4|A_{01}^2|^26|A_{10}^2|^28|A_{11}^2|^2|A_{12}^2|^2)|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(24100){\displaystyle \frac{\sqrt{10}}{315}}[6Re(A_{11}^2A_{10}^2A_{01}^2A_{00}^2)+\sqrt{6}Re(A_{12}^2A_{11}^2)](3|B_0^2|^24|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(24120){\displaystyle \frac{2\sqrt{10}}{1575}}[6Re(A_{11}^2A_{10}^2A_{01}^2A_{00}^2)+\sqrt{6}Re(A_{12}^2A_{11}^2)](3|B_0^2|^2+2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(24121){\displaystyle \frac{2}{105}}[6Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)\sqrt{6}Re(A_{12}^2A_{11}^2)]Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(24122){\displaystyle \frac{4}{105}}[\sqrt{6}Re(A_{01}^2A_{00}^2A_{11}^2A_{10}^2)Re(A_{12}^2A_{11}^2)]|B_1^2|^2(|C_0|^2|C_1|^2),`$
$`M(24200){\displaystyle \frac{2\sqrt{10}}{315}}[\sqrt{6}|A_{01}^2|^23Re(A_{12}^2A_{10}^2)](3|B_0^2|^24|B_1^2|^2)(|C_0|^2+2|C_1|^2),`$
$`M(24220){\displaystyle \frac{4\sqrt{10}}{1575}}[\sqrt{6}|A_{01}^2|^23Re(A_{12}^2A_{10}^2)](3|B_0^2|^2+2|B_1^2|^2)(|C_0|^2|C_1|^2),`$
$`M(24221){\displaystyle \frac{4}{105}}[\sqrt{6}|A_{01}^2|^2+3Re(A_{12}^2A_{10}^2)]Re(B_1^2B_0^2)(|C_0|^2|C_1|^2),`$
$`M(24222){\displaystyle \frac{4}{105}}[2|A_{01}^2|^2\sqrt{6}Re(A_{12}^2A_{10}^2)]|B_1^2|^2(|C_0|^2|C_1|^2).`$ |
warning/0001/nucl-th0001061.html | ar5iv | text | # Generator Coordinate Method Calculations for Ground and First Excited Collective States in 4He, 16O and 40Ca Nuclei
## I Introduction
The study of nucleon-nucleon correlation effects is important part of the contemporary nuclear physics . The basic idea of the independent particle models (IPM) consists in the assumption that nucleons move independently in a mean field created by the same nucleons. Consequently, each coherent motion of the nucleons creates changes in the mean field. The Hartree-Fock approximation accounts only partially for the dynamic nucleon correlations. There exist many experimental data (see e.g. the review in ) showing that the IPM are unable to describe basic nuclear characteristics. For instance, the nucleon momentum and density distributions in nuclei cannot be reproduced simultaneously . This is also the case with the occupation numbers, the hole-state spectral functions, with characteristics of nuclear reactions and others . This imposes the development of correlation methods going beyond the limits of the mean-field approximation (MFA) which account for nucleon-nucleon correlations (see e.g. ). This can be reached by extending the class of the trial functions which is used at the diagonalization of the nuclear Hamiltonian.
The generator coordinate method (GCM) is one of the methods beyond the MFA which have been applied successfully to studies of the collective nuclear motions. For instance, the GCM has been applied to investigate the breathing-mode giant monopole resonance within the framework of the relativistic mean-field theory in . In it the constrained incompressibility and the excitation energy of isoscalar giant monopole states were obtained for finite nuclei with various sets of Lagrangian parameters. An extension of the method of Ref. by using a more general ansatz for the generating functions of the GCM and by including the isovector giant monopole states has been done in . The use of the Skyrme effective forces has simplified the study of the monopole, dipole and quadrupole isoscalar and isovector vibrations in light double magic nuclei . An approach to the GCM using square-well and harmonic-oscillator construction potentials has been applied to calculate the energies of the ground and first monopole excited state, as well as the density and the nucleon momentum distributions in the ground state of <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca nuclei . In the occupation numbers, the depletion of the Fermi sea and the natural orbitals (NO) which diagonalize the ground state one-body density matrix (ODM) have been calculated. The NO related to the ground state and the single-particle potentials corresponding to them have been studied in detail in . In the two-nucleon center-of-mass and relative motion momentum distributions of $`np`$ pairs in the <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca nuclei have been calculated. The existence of high-momentum components in the one-nucleon and two-nucleon momentum distributions in the case of the square-well construction potential with infinite walls within the GCM has been obtained. The studies of the energies, the nucleon momentum and density distributions in <sup>4</sup>He and <sup>16</sup>O nuclei have been extended by means of a two generator coordinate scheme in .
In the last years giant resonances with various multipolarities different from the well-known dipole resonance have been discovered. Among these collective excitations the monopole excitations take a particular place. The isoscalar giant monopole resonances or the so-called breathing vibrational states (with $`I^\pi =0^+`$ at energies of approximately 13 to 20 MeV) have been well established experimentally (e.g. ) and have been considered to be compressional nuclear vibrations. Their study concerns the important problem of the compressibility of finite nuclei and nuclear matter (e.g. ). The description of such states is related mainly to general characteristics of the nuclei and weakly to the peculiarities of nuclear structure. Recently, in a series of articles Bishop et al.(see and references therein) applied the translationally invariant cluster (TIC) method to light nuclei, in particular to the <sup>4</sup>He nucleus. The basic properties, such as the energies and the density distributions of the ground and first excited breathing mode state in <sup>4</sup>He have been considered.
The <sup>4</sup>He nucleus has a well-established spectrum of excited states. Calculations for the monopole oscillations of helium which practically involve the whole nuclear volume have been performed on the base of the nonlinear time-dependent Hartree-Fock (TDHF) approximation . Its small amplitude limit, namely the random phase approximation (RPA), has been used extensively to describe nuclear collective motion . The TDHF method and the relativistic RPA have been recently used to extend the study of isoscalar monopole modes in finite nuclei up to <sup>208</sup>Pb. It has been found that some effective Lagrangians can describe ground states and giant resonances as well, and in particular, they can predict correctly breathing mode energies in medium and heavy nuclei. In general, it has been pointed out that when going from heavy to lighter systems the trend is that the collectivity becomes weaker. This fact poses the question about the role of different kinds of nucleon-nucleon correlations corresponding to single-particle and collective motion modes and, therefore, the investigations on the correlation effects became an important task in the nuclear theory.
The aim of the present work is to study the main characteristics of the ground and, in particular, of the first excited monopole state in the <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca nuclei within the GCM using Skyrme-type effective forces. It concerns the energies, the density distributions and radii, the nucleon momentum distributions, the natural orbitals and occupation numbers, as well as the pair center-of-mass and relative density and momentum distributions. As known, the natural orbital representation (NOR) gives a model-independent effective single-particle picture for the correlated states. We emphasize that in our work this is done also for the first excited monopole state in the nuclei considered on the basis of the ODM calculated for this state. It is known that the results of the GCM calculations depend on the type of the construction potential used to define the generating function in the method. In the present work we use three construction potentials, namely the harmonic oscillator ((HO) where the oscillator parameter is a generator coordinate), the square-well with infinite walls ((SW) where the radius of the well is a generator coordinate) and Woods-Saxon ((WS1) where the diffuseness of the Woods-Saxon well is a generator coordinate and (WS2) where the radial parameter of Woods-Saxon well is a generator coordinate). The values of the radial parameter in the WS1 case and the diffuseness parameter in the WS2 case are taken from . In our opinion, the calculations performed in this work can give an essential information about beyond MFA effects (accounted for in the considered approach to the GCM) on the mentioned characteristics. The results are compared with the available empirical data and with the calculations of other theoretical methods.
The basic relations of the GCM are given in Section 2. The results of the calculations and the discussion are presented in Section 3. The conclusions from the work are given in Section 4.
## II Basic Relations in GCM
In the case of one real generator coordinate $`x`$ the trial many-body wave function in the GCM is written as a linear combination:
$$\mathrm{\Psi }(𝐫_i)=_0^{\mathrm{}}f(x)\mathrm{\Phi }(\{𝐫_i\},x)dx,i=1..A,$$
(1)
where $`\mathrm{\Phi }(\{𝐫_i\},x)`$ is the generating function, $`f(x)`$ is the generator or weight function and $`A`$ is the mass number of the nucleus.
The application of the Ritz variational principle $`\delta E=0`$ leads to the Hill-Wheeler integral equation for the weight function:
$$_0^{\mathrm{}}[(x,x^{})EI(x,x^{})]f(x^{})𝑑x^{}=0,$$
(2)
where
$$(x,x^{})=<\mathrm{\Phi }(\{𝐫_i\}x)|\widehat{H}|\mathrm{\Phi }(\{𝐫_i\}x^{})>,$$
(3)
and
$$I(x,x^{})=<\mathrm{\Phi }(\{𝐫_i\}x)|\mathrm{\Phi }(\{𝐫_i\}x^{})>$$
(4)
are the energy and overlap kernels, respectively, and $`\widehat{H}`$ is the Hamiltonian of the system.
If the generating function $`\mathrm{\Phi }(\{𝐫_i\}x)`$ for a $`N=Z`$ nucleus is a Slater determinant built up from single-particle wave functions $`\phi _\lambda (𝐫,𝐱)`$ corresponding to a given construction potential then the energy kernel (3) has the form:
$$(x,x^{})=<\mathrm{\Phi }(\{𝐫_i\}x)|\mathrm{\Phi }(\{𝐫_i\}x^{})>H(x,x^{},𝐫)𝑑𝐫.$$
(5)
In the case of the Skyrme-like forces (for nucleus with $`Z=N`$ and without Coulomb and spin-orbital interaction) $`H(x,x^{},𝐫)`$ is given by:
$$H(x,x^{},𝐫)=\frac{\mathrm{}^2}{2m}T+\frac{3}{8}t_0\rho ^2+\frac{1}{16}(3t_1+5t_2)(\rho T+𝐣^2)+\frac{1}{64}(9t_15t_2)(\rho )^2+\frac{1}{16}t_3\rho ^{2+\sigma },$$
(6)
where $`t_0,t_1,t_2,t_3,\sigma `$ are the Skyrme-force parameters. The density $`\rho `$, the kinetic energy $`T`$ and the current density $`𝐣`$ are defined by:
$$\rho (x,x^{},𝐫)=4\underset{\lambda ,\mu =1}{\overset{A/4}{}}(N^1)_{\mu \lambda }\phi _\lambda ^{}(𝐫,x)\phi _\mu (𝐫,x^{}),$$
(7)
$$T(x,x^{},𝐫)=4\underset{\lambda ,\mu =1}{\overset{A/4}{}}(N^1)_{\mu \lambda }\phi _\lambda ^{}(𝐫,x)\phi _\mu (𝐫,x^{}),$$
(8)
$$𝐣(x,x^{},𝐫)=2\underset{\lambda ,\mu =1}{\overset{A/4}{}}(N^1)_{\mu \lambda }\{\phi _\lambda ^{}(𝐫,x)\phi _\mu (𝐫,x^{})[\phi _\lambda ^{}(𝐫,x)]\phi _\mu (𝐫,x^{})\},$$
(9)
where
$$N_{\lambda \mu }(x,x^{})=\phi _\lambda ^{}(𝐫,x)\phi _\mu (𝐫,x^{})𝑑𝐫.$$
(10)
The overlap kernel (4) is given by:
$$I(x,x^{})=[det(N_{\lambda \mu })]^4.$$
(11)
Solving the Hill-Wheeler equation(2) one can obtain the solutions $`f_0,f_1,\mathrm{}`$ for the weight functions which correspond to the eigenvalues of the energy $`E_0,E_1\mathrm{}`$.
The one-body density matrix $`\rho _i(𝐫,𝐫^{})`$ of the ground ($`i=0`$) and the first excited monopole ($`i=1`$) states in this GCM scheme has the form:
$$\rho _i(𝐫,𝐫^{})=f_i(x)f_i(x^{})I(x,x^{})\rho (x,x^{},𝐫,𝐫^{})𝑑x𝑑x^{},i=0,1,$$
(12)
where
$$\rho (x,x^{},𝐫,𝐫^{})=4\underset{\lambda ,\mu =1}{\overset{A/4}{}}(N^1)_{\mu \lambda }\phi _\lambda ^{}(𝐫,x)\phi _\mu (𝐫^{},x^{}).$$
(13)
It follows from (12) that the nucleon density distribution $`\rho _i(𝐫)`$ and the nucleon momentum distribution $`n_i(𝐤)`$ of the ground and the first excited monopole states can be expressed as:
$$\rho _i(𝐫)=f_i(x)f_i(x^{})I(x,x^{})\rho (x,x^{},𝐫)𝑑x𝑑x^{},i=0,1$$
(14)
and
$$n_i(𝐤)=f_i(x)f_i(x^{})I(x,x^{})\rho (x,x^{},𝐤)𝑑x𝑑x^{},i=0,1,$$
(15)
where
$$\rho (x,x^{},𝐫)=\rho (x,x^{},𝐫,𝐫^{}=𝐫),$$
(16)
$$\rho (x,x^{},𝐤)=4\underset{\lambda ,\mu =1}{\overset{A/4}{}}(N^1)_{\mu \lambda }\stackrel{~}{\phi }_\lambda ^{}(𝐤,x)\stackrel{~}{\phi }_\mu (𝐤,x^{})$$
(17)
and $`\stackrel{~}{\phi }(𝐤,x)`$ is the Fourier transform of $`\phi (𝐫,x)`$.
The rms radii can be calculated using the expression:
$$r_{rms}^{(i)}=\sqrt{\frac{r^4\rho _i(r)𝑑r}{r^2\rho _i(r)𝑑r}},i=0,1.$$
(18)
As shown e.g. in the attractive properties of the single-particle description can be preserved in the correlation methods using the natural orbital representation . In it the one-body density matrix has the simple form:
$$\rho _i(𝐫,𝐫^{})=\underset{\alpha }{}n_\alpha ^{(i)}\psi _\alpha ^{(i)}(𝐫)\psi _\alpha ^{(i)}(𝐫^{}),i=0,1,$$
(19)
where the natural orbitals (NO) $`\psi _\alpha ^{(i)}(𝐫)`$ form a complete orthonormal set of single-particle wave functions which diagonalize the density matrix $`\rho _i(𝐫,𝐫^{})`$ for the ground and the first excited monopole states. The natural occupation numbers (ON) $`n_\alpha ^{(i)}`$ for the state $`\alpha `$ satisfy the conditions:
$$0n_\alpha ^{(i)}1,\underset{\alpha }{}n_\alpha ^{(i)}=A.$$
(20)
It is seen from (19) that the natural orbitals and the occupation numbers can be found solving the equation:
$$\rho _i(𝐫,𝐫^{})\psi _\alpha ^{(i)}(𝐫^{})𝑑𝐫^{}=n_\alpha ^{(i)}\psi _\alpha ^{(i)}(𝐫),i=0,1.$$
(21)
For nuclei with total spin $`J=0`$ the one-body density matrix can be diagonalized in the $`\{ljm\}`$ subspace ($`l,j,m`$ being the quantum numbers corresponding to the angular momentum, total momentum and its projection). In the case of nuclei with spherical symmetry the natural orbitals have the form:
$$\psi _{nlm}^{(i)}(𝐫)=R_{nl}^{(i)}(r)Y_{lm}(\theta ,\phi )=\frac{u_{nl}^{(i)}(r)}{r}Y_{lm}(\theta ,\phi ),i=0,1,$$
(22)
where $`R_{nl}^{(i)}(r)`$ is the radial part of the NO’s and $`Y_{lm}(\theta ,\phi )`$ is the spherical function. The substitution of $`\psi _{nlm}^{(i)}(𝐫)`$ from (22) and $`\rho _i(𝐫,𝐫^{})`$ from (12) in equation (21) and the integration over the angular variables give the following equation for the radial part $`u_{nl}^{(i)}(r)`$ of the NO’s and the occupation numbers $`n_{nl}^{(i)}`$:
$$_0^{\mathrm{}}K_l^{(i)}(𝐫,𝐫^{})u_{nl}^{(i)}(r^{})𝑑r^{}=n_{nl}^{(i)}u_{nl}^{(i)}(r),i=0,1,$$
(23)
where
$$K_l^{(i)}(𝐫,𝐫^{})=rr^{}f_i(x)f_i(x^{})I(x,x^{})\underset{n,n^{}}{}(N^1)_{nl,n^{}l}_{n^{}l}^{}(r,x)_{nl}(r^{},x^{})dxdx^{},i=0,1.$$
(24)
The summation in (24) is performed over all single-particle wave functions with given $`l`$ forming the Slater determinant of the generating function and $`_{nl}(r,x)`$ are the radial parts of these functions.
The nucleon density distribution $`\rho _i(r)`$ and the nucleon momentum distribution $`n_i(k)`$ for nuclei with $`J=0`$ can be written in the following form in the NOR:
$$\rho _i(r)=\frac{1}{4\pi }\underset{l}{}2(2l+1)\underset{n}{}n_{nl}^{(i)}|R_{nl}^{(i)}(r)|^2,i=0,1,$$
(25)
$$n_i(k)=\frac{1}{4\pi }\underset{l}{}2(2l+1)\underset{n}{}n_{nl}^{(i)}|\stackrel{~}{R}_{nl}^{(i)}(k)|^2,i=0,1,$$
(26)
where
$$\stackrel{~}{R}_{nl}^{(i)}(k)=(\frac{2}{\pi })^{1/2}(i)^l_0^{\mathrm{}}r^2j_l(kr)R_{nl}^{(i)}(r)𝑑r,i=0,1$$
(27)
is the radial part of the NO in the momentum space and $`j_l(kr)`$ are spherical Bessel functions of order $`l`$.
The depletion of the Fermi sea can be defined by the expression:
$$𝒟^{(i)}=\frac{4}{A}\underset{j,\alpha \{F\}}{}(2j+1)(1n_\alpha ^{(i)}),i=0,1,$$
(28)
where $`\{F\}`$ refers to the Fermi sea.
The rms radius of the NO $`\psi _\alpha ^{(i)}(𝐫)`$ is given by the expression:
$$<r_\alpha ^{(i)}>=\sqrt{\frac{r^2|\psi _\alpha ^{(i)}(𝐫)|^2𝑑𝐫}{|\psi _\alpha ^{(i)}(𝐫)|^2𝑑𝐫}},i=0,1.$$
(29)
The two-body density matrix $`\rho ^{(2)}(\xi _1,\xi _2;\xi _1^{},\xi _2^{})`$ in the GCM has the form:
$$\rho ^{(2)}(\xi _1,\xi _2;\xi _1^{},\xi _2^{})=𝑑xf^{}(x)𝑑x^{}f(x^{})\rho ^{(2)}(x,x^{};\xi _1,\xi _2;\xi _1^{},\xi _2^{}),$$
(30)
where the coordinate $`\xi `$ includes the spatial coordinate $`𝐫`$, as well as the spin ($`s`$) and isospin ($`\tau `$) variables. If the generating wave function is a Slater determinant constructed from a complete set of one-particle wave functions $`\phi _i(x,\xi )`$ the matrix $`\rho ^{(2)}(x,x^{};\xi _1,\xi _2;\xi _1^{},\xi _2^{})`$ can be expressed as :
$$\rho ^{(2)}(x,x^{};\xi _1,\xi _2;\xi _1^{},\xi _2^{})=\frac{I(x,x^{})}{2}det\left(\begin{array}{cc}\rho ^{(2)}(x,x^{};\xi _1,\xi _1^{})& \rho ^{(2)}(x,x^{};\xi _1,\xi _2^{})\\ \rho ^{(2)}(x,x^{};\xi _2,\xi _1^{})& \rho ^{(2)}(x,x^{};\xi _2,\xi _2^{})\end{array}\right),$$
(31)
where
$$\rho ^{(2)}(x,x^{};\xi ,\xi ^{})=\underset{k,l=1}{\stackrel{A}{}}\left(N^1\right)_{lk}\phi _k^{}(x,\xi )\phi _l(x^{},\xi ^{}),$$
(32)
$`\left(N^1\right)_{lk}`$ is the inverse matrix of:
$$N_{kl}(x,x^{})=\underset{s,\tau }{}d𝐫\phi _k^{}(x,\xi )\phi _l(x^{},\xi ),$$
(33)
and
$$I(x,x^{})=det\left(N_{kl}\right).$$
(34)
The two-particle emission experiments require some knowledge of physical quantities associated with the TDM. For example, using the two-body density matrix $`\rho ^{(2)}`$ in coordinate space, Eq.(30), one can define the pair center-of-mass local density distribution:
$$\rho ^{(2)}(𝐑)=\rho ^{(2)}(𝐑+𝐬/2,𝐑𝐬/2)𝑑𝐬$$
(35)
and the pair relative local density distribution:
$$\rho ^{(2)}(𝐬)=\rho ^{(2)}(𝐑+𝐬/2,𝐑𝐬/2)𝑑𝐑.$$
(36)
In momentum space the associated pair center-of-mass and relative momentum distributions can be defined:
$$n^{(2)}(𝐊)=n^{(2)}(𝐊/2+𝐤,𝐊/2𝐤)𝑑𝐤,$$
(37)
$$n^{(2)}(𝐤)=n^{(2)}(𝐊/2+𝐤,𝐊/2𝐤)𝑑𝐊.$$
(38)
The physical meaning of $`\rho ^{(2)}(𝐬)`$ and $`n^{(2)}(𝐤)`$ is the probability to find two particles displaced of a certain relative distance $`𝐬=𝐫_1𝐫_2`$ or moving with relative momentum $`𝐤=(𝐤_1𝐤_2)/2`$, respectively, while $`\rho ^{(2)}(𝐑)`$ and $`n^{(2)}(𝐊)`$ represent the probability to find a pair of particles with center-of-mass coordinate $`𝐑=\left(𝐫_1+𝐫_2\right)/2`$ or center-of-mass momentum $`𝐊=𝐤_1+𝐤_2`$.
## III Results and Discussion
### A The ground and first excited collective state energies
The Hill-Wheeler equation (2) is solved using a discretization procedure similar to that of Refs. . The values of the Skyrme-force parameters in (6) in the case of square-well construction potential are the same as in . They are determined to give an optimal fit to the binding energies of <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca obtained from the Hill-Wheeler equation (2). The parameter set values $`t_0=2765,t_1=383.94,t_2=38.04,t_3=15865,\sigma =1/6`$ lead to the energies of the ground $`E_0`$ and the first monopole excited $`E_1`$ states (without the Coulomb energy) shown in Table I. In the case of the harmonic-oscillator and Woods-Saxon construction potentials SkM\* parameter set values ($`t_0=2645,t_1=410,t_2=135,t_3=15595,\sigma =1/6`$) giving realistic binding energies are used. It can be seen from Table I that the energies obtained with WS1 are close to that obtained with HO, while the energies corresponding to WS2 are closer to that calculated with SW construction potential.
The values of the excitation energies $`\mathrm{\Delta }E=E_1E_0`$ of the first monopole state ($`I^\pi =0^+`$) calculated within the GCM are given and compared with other calculations and some experimental data in Table II. It is seen that the GCM results with HO construction potential and in the WS1 case are in good agreement with the result of Brink and Nash for <sup>16</sup>O nucleus and they are closer to the experimental values than the GCM results with the SW construction potential. The energy of the first excited $`0^+`$ in <sup>4</sup>He obtained in the present work is compared in Table II with the recent refined calculations of Bishop et al within the TIC method as well as with the results of the coupled rearrangement of the channels method .
### B The one-body density matrix $`\rho (𝐫,𝐫^{})`$ of the ground and first excited monopole states
#### 1 The nucleon density distribution $`\rho (𝐫)`$
The one-body distribution function corresponding to the density distribution $`\rho _i(𝐫)`$ from (14) is given by the expression:
$$g_i(r)=4\pi r^2\rho _i(r),i=0,1.$$
(39)
The function $`g_i(r)`$ of the ground (i=0) and the first excited monopole (i=1) states of <sup>4</sup>He calculated in the GCM are compared in Figure 1 with those obtained in the TIC method. It can be seen that there is a large difference in the behaviour of the function $`g(r)`$ for both states. This is due to the strong increase of the size of the nucleus, as is shown in Table III: the rms radius increases from 1.89 fm for the ground state to 3.00 fm for the first excited monopole state in the case of the HO, from 1.77 fm to 2.86 fm in the case of the SW and from 1.85 fm to 3.29 fm when using WS2. The significantly larger values of rms radii for the first excited state in $`{}_{}{}^{4}He`$ nucleus in respect to their ground state values support the breathing-mode interpretation. At the same time, the nature of the first excited monopole state in $`{}_{}{}^{4}He`$ is not still well understood. An attempt to solve this problem is made in Ref. where it is concluded that the $`i=1`$ state is dominated by the $`{}_{}{}^{3}H+p`$ configuration. The rms radii of the ground and of the first excited monopole states in <sup>16</sup>O and <sup>40</sup>Ca are also presented in Table III. We note that the rms radii of the ground state in these nuclei obtained with Woods-Saxon construction potential are very close to the experimental values.
For the first excited monopole state the one-body distribution function $`g_1(r)`$ calculated within the GCM using the WS2 potential and that one calculated within the TIC method are in good agreement. In contrast with this state, for a ground state both methods yield functions $`g_0(r)`$ which differ significantly. This is due to the spurious centre-of-mass motion correction for <sup>4</sup>He which is accounted for in the TIC method . At the same time effect doesn’t influence substantially the first excited monopole-state density.
The nucleon density distributions of <sup>16</sup>O and <sup>40</sup>Ca of the ground $`\rho _0`$ and the first excited monopole $`\rho _1`$ states obtained within the GCM are presented in Figures 2 and 3, respectively. For the first excited monopole state the density distribution $`\rho _1(r)`$ decreases in the central region of both nuclei and increases for large $`r`$. This behaviour corresponds to the increase of the rms radius in the first excited state. It can be seen also that the difference in the rms radii $`\mathrm{\Delta }r_{rms}=r_{rms}^{(1)}r_{rms}^{(0)}`$ decreases with the increase of the mass number.
#### 2 The nucleon momentum distribution $`n(𝐤)`$
The nucleon momentum distribution $`n(𝐤)`$ is one of the nuclear quantities which are sensitive to various types of nucleon-nucleon correlations. The momentum distributions for the ground and first excited monopole states of <sup>16</sup>O have been calculated in the GCM using Eqs.(15) and (17) and are shown in Figure 4. As can be seen there is no essential difference between the behaviour of the $`n(𝐤)`$ in the ground and in the first excited monopole states.
It is seen from Figure 4 that the nucleon momentum distribution $`n(𝐤)`$ obtained in the GCM depends on the construction potentials considered. The use of SW potential leads to an existence of high-momentum tail of $`n(𝐤)`$ which is not the case when the HO and WS potentials are used. Here we would like to note that:
(i) the single-particle wave functions corresponding to the SW potential contain themselves high-momentum components due to the particular form of the potential and they are reflected in the GCM result,
(ii) the lack of substantial high-momentum tail in $`n(𝐤)`$ when using HO and WS construction potential leads to the conclusion that the nucleon-nucleon correlations included in the GCM approach are not of short-range type. An additional study of the correlations accounted for in the GCM concerns some two-body characteristics and we will discuss the results for them in subsection C.
#### 3 The natural orbitals and occupation numbers
The natural orbitals and occupation numbers have been calculated using Eqs.(21)-(24). A discretization procedure with respect to both $`r`$ and $`r^{}`$ has been applied solving Eq.(23) and the resulting matrix eigenvalue problem has been solved numerically. The natural orbitals obtained in the coordinate space are plotted in Figure 5 for the <sup>16</sup>O nucleus. The corresponding rms radii and the natural occupation numbers of <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca are shown in Tables 4 and 5, respectively.
One can see from Table IV and from the shape of the NO’s given in Figure 5 that the rms radii of the natural hole-states increase from the ground state to the first excited state. The occupation numbers of the natural hole-states decrease in the first excited monopole state, while those corresponding to the natural particle-states increase (see Table V). As can be seen from the same Table V, the depletion of the Fermi sea increases in the first excited state. This is strongly expressed for the <sup>4</sup>He nucleus. The calculated depletion decreases with the increase of the mass number.
### C The pair center-of-mass and relative density and momentum distributions
The pair center-of-mass and relative density and momentum distributions Eqs.(35)-(38) are calculated by using of HO and SW potentials and are given in Figure 6 for the ground state of <sup>16</sup>O nucleus. They are compared with the results obtained within the low-order approximation of the Jastrow correlation method (JCM) where the short-range correlations are explicitely involved. As it is seen from Figure 6($`a^{}`$) the relative distributions $`\rho ^{(2)}(𝐬)`$ obtained in GCM and JCM have different behaviour in the region from 0 to 0.4 fm. This fact shows that the GCM doesn’t take into account short-range but other kind of correlations which are obviously related with the collective motion of the nucleons. The curves of $`n^{(2)}(𝐊)`$ calculated for both potentials decrease rapidly down at K$`>`$1fm<sup>-1</sup> while the result obtained within the JCM shows a high-momentum tail.
## IV Conclusion
In the present work the characteristics of the ground and the first excited monopole states of the three nuclei <sup>4</sup>He, <sup>16</sup>O and <sup>40</sup>Ca are studied within the GCM using different construction potentials. Though the results are sensitive to the type of the construction potentials used, some general conclusions can be summarized as follows
(i) There is an increase of the rms radius in the first excited monopole state as a consequence of the increase of the density distribution for large $`r`$.
(ii) There is not an essential difference between the behaviour of the nucleon momentum distribution in the ground and in the first excited monopole state.
(iii) The use of the natural orbital representation makes it possible to study the effective single-particle picture not only of the ground, but also of the first excited monopole state. It is established that the depletion of the Fermi sea increases considerably in the first excited monopole state. This is strongly expressed in the case of <sup>4</sup>He. The calculated depletion decreases with the increase of the mass number.
(iv) The study of the natural orbitals in the ground and the first excited monopole state shows that there are not substantial changes in their forms for a given $`nl`$-state when they are calculated for the ground and the first excited monopole state. The rms radii of the hole-state natural orbitals are larger in the case of the first excited monopole state in comparison with those for the ground state.
(v) The results on the one- and two-body density and momentum distributions, occupation probabilities and natural orbitals obtained within the GCM using various construction potentials show that the nucleon-nucleon correlations accounted in the approach are different from the short-range ones but are rather related to the collective motion of the nucleons. It turns out that the latter are also important in calculations of one- and two-body overlap functions which are necessary in the calculations of the cross-sections of one- and two-nucleon removal reactions. The work on the applications of the GCM overlap functions for the description of cross-section of $`(p,d)`$, $`(e,e^{}p)`$ and $`(\gamma ,p)`$ reactions is in progress.
## V Acknowledgments
This work was partly supported by the Bulgarian National Science Foundation under the Contract No.$`\mathrm{\Phi }`$-809.
Figure Captions
One-body distribution function (Eq.(39)) $`g(r)`$ of <sup>4</sup>He calculated within the GCM using harmonic-oscillator (a), square-well (b) and Woods-Saxon (c) potentials and within the TIC method . The normalization is: $`g_i(r)𝑑r=1(i=0,1).`$
Nucleon density distributions of the ground ($`\rho _0`$) and the first excited monopole ($`\rho _1`$) states of <sup>16</sup>O calculated within the GCM using harmonic-oscillator (a), square-well (b) and Woods-Saxon (c) potentials. The normalization is: $`\rho _i(𝐫)𝑑𝐫=16(i=0,1)`$.
The same as in Figure 2 but for <sup>40</sup>Ca.
Nucleon momentum distributions of the ground ($`n_0`$) and the first excited monopole ($`n_1`$) states of <sup>16</sup>O calculated within the GCM using harmonic-oscillator (a), square-well (b) and Woods-Saxon (c) potentials. The normalization is: $`n_i(k)k^2𝑑k=16(i=0,1).`$
Natural orbitals for the 1s-hole (a, a, a<sup>′′</sup>), 1p-hole (b, b, b<sup>′′</sup>) and 2s-particle (c, c, c<sup>′′</sup>) states of the ground ($`i=0`$) and of the first excited monopole ($`i=1`$) states of <sup>16</sup>O in coordinate space calculated within the GCM using harmonic-oscillator, square-well and Woods-Saxon potentials, respectively.
The pair center-of-mass $`\rho ^{(2)}(𝐑)`$ (a) and relative $`\rho ^{(2)}(𝐬)`$ (a) density distributions and the pair center-of-mass $`n^{(2)}(𝐊)`$ ($`b`$) and relative $`n^{(2)}(𝐤)`$ (b) momentum distributions of the ground state of <sup>16</sup>O calculated within the GCM using harmonic-oscillator and square-well potentials. The results obtained within the Jastrow correlation method are also given. The normalization is: $`n^{(2)}(𝐊)𝑑𝐊=1`$, $`n^{(2)}(𝐤)𝑑𝐤=1`$, $`\rho ^{(2)}(𝐑)𝑑𝐑=A(A1)/2`$, $`\rho ^{(2)}(𝐬)𝑑𝐬=A(A1)/2`$ |
warning/0001/gr-qc0001068.html | ar5iv | text | # Gravitational waves from quasi-spherical black holes
## Abstract
A quasi-spherical approximation scheme, intended to apply to coalescing black holes, allows the waveforms of gravitational radiation to be computed by integrating ordinary differential equations.
04.30.-w, 04.25.-g, 04.70.Bw, 04.20.Ha
The coalescence of binary black holes is expected to be one of the main astrophysical sources for upcoming gravitational-wave detectors. The initial phase of inspiral and the final phase of ringdown are understood in terms of post-Newtonian and close-limit approximations respectively, but the coalescence is only qualitatively understood and generally thought to be tractable only by numerical methods. Considerable problems have been encountered and currently there are no reliable predictions of waveforms.
This article presents an approximation scheme to compute the gravitational waveforms for space-times close to spherical symmetry. This is intended to apply to binary black holes once they have coalesced, i.e. when a marginal surface encloses both sources. The quasi-spherical approximation will be best when the angular momentum is small, but note that even the maximally rotating Kerr black hole is $`70\%`$ spherically symmetric according to the ratio of the areal and equatorial radii of the horizon. Thus rough estimates may still be possible even for appreciable angular momentum.
The basic idea of a quasi-spherical approximation is to make a 2+2 decomposition of the space-time and linearize only those parts of the extrinsic curvature which vanish in spherical symmetry, cf. Bishop et al.. Thus when the linearized fields vanish, spherical symmetry is recovered in full. This can be a highly dynamical situation; there will be no assumption of quasi-stationarity. Likewise, there will be no assumption of an exactly spherical background. Unlike previous work on null-temporal formulations, a dual-null formulation is adopted here, i.e. a decomposition of the space-time by two intersecting foliations of null hypersurfaces. This is adapted to the radiation problem in that the imposition of no ingoing radiation and the extraction of the outgoing radiation are immediate. It also allows a remarkable simplification from partial to ordinary differential equations.
A general Hamiltonian theory of dual-null dynamics has been applied to Einstein gravity and is summarized as follows. Denoting the space-time metric by $`g`$ and labelling the null hypersurfaces by $`x^\pm `$, the normal 1-forms $`n^\pm =dx^\pm `$ therefore satisfy
$$g^1(n^\pm ,n^\pm )=0.$$
(1)
The relative normalization of the null normals may be encoded in a function $`f`$ defined by
$$e^f=g^1(n^+,n^{}).$$
(2)
Then the induced metric on the transverse surfaces, the spatial surfaces of intersection, is found to be
$$h=g+2e^fn^+n^{}$$
(3)
where $``$ denotes the symmetric tensor product. The dynamics is described by Lie transport along two commuting evolution vectors $`u_\pm `$:
$$[u_+,u_{}]=0.$$
(4)
Specifically, the evolution derivatives, to be discretized in a numerical code, are
$$\mathrm{\Delta }_\pm =L_{u_\pm }$$
(5)
where $``$ indicates projection by $`h`$ and $`L`$ denotes the Lie derivative. There are two shift vectors
$$s_\pm =u_\pm .$$
(6)
In a coordinate basis $`(u_+,u_{};e_i)`$ such that $`u_\pm =/x^\pm `$, where $`e_i=/x^i`$ is a basis for the transverse surfaces, the metric takes the form
$`g`$ $`=`$ $`h_{ij}(dx^i+s_+^idx^++s_{}^idx^{})`$ (8)
$`(dx^j+s_+^jdx^++s_{}^jdx^{})2e^fdx^+dx^{}.`$
Then $`(h,f,s_\pm )`$ are configuration fields and the independent momentum fields are found to be linear combinations of
$`\theta _\pm `$ $`=`$ $`L_\pm 1`$ (9)
$`\sigma _\pm `$ $`=`$ $`L_\pm h\theta _\pm h`$ (10)
$`\nu _\pm `$ $`=`$ $`L_\pm f`$ (11)
$`\omega `$ $`=`$ $`\frac{1}{2}e^fh([l_{},l_+])`$ (12)
where $``$ is the Hodge operator of $`h`$ and $`L_\pm `$ is shorthand for the Lie derivative along the null normal vectors
$$l_\pm =u_\pm s_\pm =e^fg^1(n^{}).$$
(13)
Then the functions $`\theta _\pm `$ are the expansions, the traceless bilinear forms $`\sigma _\pm `$ are the shears, the 1-form $`\omega `$ is the twist, measuring the lack of integrability of the normal space, and the functions $`\nu _\pm `$ are the inaffinities, measuring the failure of the null normals to be affine. The fields $`(\theta _\pm ,\sigma _\pm ,\nu _\pm ,\omega )`$ encode the extrinsic curvature of the dual-null foliation. These extrinsic fields are unique up to duality $`\pm `$ and diffeomorphisms which relabel the null hypersurfaces, i.e. $`dx^\pm e^{\lambda _\pm }dx^\pm `$ for functions $`\lambda _\pm (x^\pm )`$.
It is also useful to decompose $`h`$ into a conformal factor $`r`$ and a conformal metric $`k`$ by
$$h=r^2k$$
(14)
such that
$$\mathrm{\Delta }_\pm \widehat{}1=0$$
(15)
where $`\widehat{}`$ is the Hodge operator of $`k`$, satisfying $`1=\widehat{}r^2`$. Denoting the covariant derivative of $`h`$ by $`D`$, the Ricci scalar of $`h`$ is found to be
$$R=2r^2(1D^2\mathrm{ln}r)$$
(16)
by using the coordinate freedom on a given surface to fix $`k`$ as the metric of a unit sphere.
The dual-null Hamilton equations and integrability conditions for vacuum Einstein gravity have been given previously in a slightly different notation, so will not be repeated here. They are linear combinations of the vacuum Einstein equation and a first integral of the contracted Bianchi identity. This is the vacuum Einstein system in first-order dual-null form. The vacuum case suffices for the application, outside the black holes.
In spherical symmetry, $`(s_\pm ,\sigma _\pm ,\omega ,D)`$ vanish, while $`(h,f,\theta _\pm ,\nu _\pm ,\mathrm{\Delta }_\pm )`$ are generally non-zero, e.g.. The quasi-spherical approximation will therefore consist of linearizing in $`(s_\pm ,\sigma _\pm ,\omega ,D)`$. In practice, one truncates the equations by setting to zero any second-order terms in $`(s_\pm ,\sigma _\pm ,\omega ,D)`$. This greatly simplifies the equations, leaving the momentum definitions as
$`\mathrm{\Delta }_\pm r`$ $`=`$ $`\frac{1}{2}r\theta _\pm `$ (17)
$`\mathrm{\Delta }_\pm f`$ $`=`$ $`\nu _\pm `$ (18)
$`\mathrm{\Delta }_\pm k`$ $`=`$ $`r^2\sigma _\pm `$ (19)
$`\mathrm{\Delta }_+s_{}\mathrm{\Delta }_{}s_+`$ $`=`$ $`2e^fh^1(\omega )`$ (20)
and the remaining equations as
$`\mathrm{\Delta }_\pm \theta _\pm `$ $`=`$ $`\nu _\pm \theta _\pm \frac{1}{2}\theta _\pm ^2`$ (21)
$`\mathrm{\Delta }_\pm \theta _{}`$ $`=`$ $`\theta _+\theta _{}e^fr^2`$ (22)
$`\mathrm{\Delta }_\pm \nu _{}`$ $`=`$ $`\frac{1}{2}\theta _+\theta _{}e^fr^2`$ (23)
$`\mathrm{\Delta }_\pm \sigma _{}`$ $`=`$ $`\frac{1}{2}(\theta _\pm \sigma _{}\theta _{}\sigma _\pm )`$ (24)
$`\mathrm{\Delta }_\pm \omega `$ $`=`$ $`\theta _\pm \omega \pm \frac{1}{2}(D\nu _\pm D\theta _\pm \theta _\pm Df).`$ (25)
This follows immediately from the full equations, the above expression for $`R`$ and the fact that
$$\mathrm{\Delta }_\pm =L_\pm $$
(26)
in this truncation. One may take quasi-spherical coordinates $`x^i=(\vartheta ,\phi )`$ on the transverse surfaces such that $`\widehat{}1=\mathrm{sin}\vartheta d\vartheta d\phi `$, the standard area form of a unit sphere. Then $`r`$ is the quasi-spherical radius.
The shear equations, composed into a second-order equation for $`k`$, become
$$\mathrm{}k=0$$
(27)
where $`\mathrm{}`$ is the quasi-spherical wave operator:
$$\mathrm{}\varphi =2e^f\left(\mathrm{\Delta }_{(+}\mathrm{\Delta }_)\varphi +2r^1\mathrm{\Delta }_{(+}r\mathrm{\Delta }_)\varphi \right).$$
(28)
Thus the conformal metric $`k`$ satisfies the quasi-spherical wave equation. Then $`k`$ may be interpreted as encoding the gravitational radiation. In particular, fixing $`u_+`$ to be the outgoing direction, the Bondi news at future null infinity $`\mathrm{}^+`$ is essentially $`r^1\sigma _{}`$, as described explicitly below. Likewise, the no-ingoing-radiation condition is just $`r^1\sigma _+=0`$ at past null infinity $`\mathrm{}^{}`$. That $`k`$ generally encodes the free gravitational data was suggested by d’Inverno & Stachel and has been rediscovered by various authors, e.g..
The dual-null initial-data formulation is based on a spatial surface $`\mathrm{\Sigma }`$ and the null hypersurfaces $`\mathrm{\Sigma }_\pm `$ locally generated from $`\mathrm{\Sigma }`$ in the $`u_\pm `$ directions. The structure of the field equations shows that one may specify $`(h,f,\theta _\pm ,\omega )`$ on $`\mathrm{\Sigma }`$, $`(\sigma _+,\nu _+)`$ on $`\mathrm{\Sigma }_+`$, $`(s_{},\sigma _{},\nu _{})`$ on $`\mathrm{\Sigma }_{}`$ and $`s_+`$ in $`U`$, a region to the future of $`\mathrm{\Sigma }_\pm `$. In particular, the initial data is freely specifiable. There are no constraints as in the 3+1 formulation; these have been converted into evolution equations along $`\mathrm{\Sigma }_\pm `$, which even in the general (non-quasi-spherical) case can be solved in closed form.
A numerical integration scheme runs as follows, as depicted in Fig.1. First integrate the $`\mathrm{\Delta }_+`$ equations from $`\mathrm{\Sigma }`$ to obtain the full data on $`\mathrm{\Sigma }_+`$. Then integrate the $`\mathrm{\Delta }_{}`$ equations one step along each ingoing null hypersurface, generating a new null hypersurface $`\mathrm{\Sigma }_+^{}`$. Then repeat: integrate the $`\mathrm{\Delta }_+`$ equations along $`\mathrm{\Sigma }_+^{}`$ to obtain the full data on $`\mathrm{\Sigma }_+^{}`$, and so on. In practice, some interpolation between the two integrations is useful. There are many ways to perform the integrations in a different order, allowing flexibility which can be used, for instance, to avoid singularities. Any such scheme gives two estimates of $`(r,k,f,\theta _\pm ,\omega )`$ at each point, since some of the equations play the role of integrability conditions. Thus one could ignore such equations to obtain a free rather than constrained integration scheme. This allows numerous internal checks on the accuracy of the numerical code, analogous to those of 3+1 integration schemes, e.g. Choptuik.
The equations for $`\mathrm{\Delta }_\pm (r,f,\theta _\pm ,\nu _\pm )`$, the quasi-spherical equations, decouple from the remaining equations. Thus there is a quasi-spherical background which may be found by integrating the quasi-spherical initial data. Since this background is independent of the linearized part, one may economize when computing different evolutions on the same background. It should be stressed that the quasi-spherical background is neither fixed in advance nor necessarily spherical, e.g. $`Dr0`$ in general.
To compute the outgoing radiation, one now needs only to integrate the equations for $`(\mathrm{\Delta }_\pm k,\mathrm{\Delta }_\pm \sigma _{})`$, i.e. the quasi-spherical wave equation for $`k`$. It is remarkable that this entire integration scheme involves only ordinary differential equations. The equations for $`\mathrm{\Delta }_\pm \omega `$ are partial differential equations, containing transverse $`D`$ derivatives, but the other equations decouple from the equations for $`(s_\pm ,\omega )`$, which therefore need not be solved for the radiation problem. In short, most of the complexity of the system has been isolated and sidestepped.
Moreover, one may use a conformal transformation to obtain a scheme which is more accurate at large distances. Using the conformal factor
$$\mathrm{\Omega }=r^1$$
(29)
the rescaled expansions and shears
$`\vartheta _\pm `$ $`=`$ $`r\theta _\pm `$ (30)
$`\varsigma _\pm `$ $`=`$ $`r^1\sigma _\pm `$ (31)
are finite and generally non-zero at $`\mathrm{}^{}`$ for an asymptotically flat space-time. Rewriting the relevant equations yields the quasi-spherical equations
$`\mathrm{\Delta }_\pm \mathrm{\Omega }`$ $`=`$ $`\frac{1}{2}\mathrm{\Omega }^2\vartheta _\pm `$ (32)
$`\mathrm{\Delta }_\pm f`$ $`=`$ $`\nu _\pm `$ (33)
$`\mathrm{\Delta }_\pm \vartheta _\pm `$ $`=`$ $`\nu _\pm \vartheta _\pm `$ (34)
$`\mathrm{\Delta }_\pm \vartheta _{}`$ $`=`$ $`\mathrm{\Omega }(\frac{1}{2}\vartheta _+\vartheta _{}+e^f)`$ (35)
$`\mathrm{\Delta }_\pm \nu _{}`$ $`=`$ $`\mathrm{\Omega }^2(\frac{1}{2}\vartheta _+\vartheta _{}+e^f)`$ (36)
and the linearized equations
$`\mathrm{\Delta }_\pm k`$ $`=`$ $`\mathrm{\Omega }\varsigma _\pm `$ (37)
$`\mathrm{\Delta }_\pm \varsigma _{}`$ $`=`$ $`\frac{1}{2}\mathrm{\Omega }\vartheta _{}\varsigma _\pm .`$ (38)
One may take $`\mathrm{\Sigma }_+`$ to be either part of $`\mathrm{}^{}`$, as depicted in Fig.2, or at sufficiently large distance for numerical purposes. Here, large distance means small $`\mathrm{\Omega }`$. For the quasi-spherical approximation to be valid at large distance, one may fix $`(f,\vartheta _\pm ,k)=(0,\pm \sqrt{2},ϵ)`$ on $`\mathrm{\Sigma }`$, where $`ϵ=d\vartheta d\vartheta +\mathrm{sin}^2\vartheta d\phi d\phi `$ is the standard metric of a unit sphere, and $`\nu _+=0`$ on $`\mathrm{\Sigma }_+`$. The remaining coordinate data is given by $`\nu _{}`$ on the ingoing null hypersurface $`\mathrm{\Sigma }_{}`$, which is left free so that one may adapt the foliation of $`\mathrm{\Sigma }_{}`$ to the surfaces which are most spherical.
The no-ingoing-radiation condition is $`\varsigma _+=0`$ at $`\mathrm{}^{}`$, leaving the gravitational initial data as $`\varsigma _{}`$ on $`\mathrm{\Sigma }_{}`$. The outgoing radiation is found by computing $`\varsigma _{}`$ at $`\mathrm{}^+`$, which is essentially the Bondi news. More precisely, the Bondi energy flux at $`\mathrm{}^{}`$ would be
$$\psi _\pm =\frac{e^f\vartheta _{}\varsigma _\pm ^2}{64\pi }$$
(39)
where $`\sigma ^2=k^{ab}k^{cd}\sigma _{ac}\sigma _{bd}`$ and such second-order terms are no longer being ignored. That is, the energy supply would be
$$\mathrm{\Delta }_\pm E=\widehat{}\psi _\pm $$
(40)
where $`E`$ is the Bondi energy. In summary, the outgoing waveforms and their energy may be computed by integrating 9 first-order ordinary differential equations and their duals, or a subset in the case of free evolution. For numerical purposes, this is a dramatic simplification. Numerical implementation of this scheme is in progress.
Before concluding, it should be noted that the domain of validity of the quasi-spherical approximation is not known in a precise sense. The guarantee is simply that spherically symmetric Einstein gravity is recovered in full when the linearized fields vanish. For the usual perturbative approximations, one may check successive orders of approximation to compare accuracy, but for the quasi-spherical approximation, the corresponding second-order approximation would be full Einstein gravity. If one wishes to know whether a given space-time is sufficiently spherical, the rough answer is that there should be a 2+2 decomposition such that the fields to be linearized are small compared to the remaining part. This depends on the choice of transverse surfaces, so that there will be some art to choosing the 2+2 foliation for optimal accuracy. For the Kerr black hole, the quasi-spherical null coordinates of Pretorius & Israel may be useful. For a coalesced black hole, one might base the foliation on the marginal surfaces which locally define it, i.e. use the coordinate freedom in $`\nu _{}`$ on $`\mathrm{\Sigma }_{}`$ so that the foliation contains a marginal surface. There are some general laws of black-hole dynamics in terms of marginal surfaces and the trapping horizons they generate, including that outer trapping horizons are achronal and therefore cannot causally influence $`\mathrm{}^+`$. Thus $`\mathrm{\Sigma }_{}`$ need not extend inside the trapped region in order for the domain $`U`$ of integration to reach all of $`\mathrm{}^+`$.
To conclude, the intended scenario is an asymptotically flat space-time containing coalescing black holes, with an ingoing null hypersurface $`\mathrm{\Sigma }_{}`$ chosen to intersect the coalesced black hole, i.e. the region of future trapped surfaces enclosing the original black holes. The initial data on $`\mathrm{\Sigma }_{}`$ may be determined by extracting the relevant data from a conventional 3+1 numerical computation from an initial spatial hypersurface $`\mathrm{\Sigma }_0`$, smoothed off to the past of $`\mathrm{\Sigma }_0`$. The smoothing may be expected not to affect the results significantly if $`\mathrm{\Sigma }_{}\mathrm{\Sigma }_0`$ is sufficiently outside the black holes; there will be some spurious radiation at $`\mathrm{}^+`$ at early times, but not at the relevant late times, as this would involve backscatter of backscatter.
The advantage of this procedure is that it avoids the outer-boundary problems which plague the conventional codes. As depicted in Fig.2, the outer boundary $`B`$ cannot causally influence $`\mathrm{\Sigma }_{}`$ if it intersects $`\mathrm{\Sigma }_0`$ inside $`B`$. Thus the scheme requires only clean data from the 3+1 computation, uncontaminated by outer-boundary problems. A code implementing the scheme may be regarded as a black box which, taking input from any other code from which the required data on $`\mathrm{\Sigma }_{}`$ can be extracted, computes approximate waveforms for the gravitational radiation.
This suggests a quite general proposal to compute outgoing gravitational radiation from a 3+1 computation by a conformal dual-null code which extracts data on an ingoing null hypersurface intersecting the initial spatial hypersurface inside its outer boundary. One might expect this to be simpler than the usual matching on a temporal hypersurface since the outer boundary is avoided and the problem is merely of extraction rather than dynamic matching. At present, there seems to be neither a general conformal dual-null code nor work on data extraction on a null hypersurface, though the null-temporal formulation can presumably be adapted. These are tractable projects which would allow accurate computation of gravitational waveforms from coalescing black holes.
Acknowledgements. The author thanks Pablo Laguna, Luis Lehner, Keith Lockitch, Hisa-aki Shinkai and Jeff Winicour for discussions, and Abhay Ashtekar and the Center for Gravitational Physics and Geometry for hospitality. Research supported by the National Science Foundation under award PHY-9800973. |
warning/0001/quant-ph0001050.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Nonlinear lattice equations are abound in physical applications and their study constitutes a well developed field of research. In many cases where models are described by classical lattice differential equations this is done as a first approximation to what is essentially a quantum mechanical lattice system. Quantized nonlinear lattice equations are the main target of the most successful method of the quantum inverse scattering method (QISM) . Recent attempts to treat quantum lattice equations with arbitrary number of lattice sites have led to such schemes as the so called number state method (NSM) , and the Hartree approximation (see also for a discussion about QISM and NSM). On the other hand it seems highly desirable when treating quantum versions of classical lattices to be able to utilize in some sense the available classical solutions of the nonlinear evolution equations and to keep a conceptual framework viz.phase spaces, Poisson brackets, coordinate transformations etc. akin to that of the classical models. It is in that direction that this work focuses and provides an alternative to the above mentioned methods.
Specifically we deal with lattices related to the nonintegrable generalized discrete nonlinear Schrödinger equation (DNLS) (also known as the generalized discrete self-trapping (GDST) equation) and with a modified version of it (MDNLS) . We also discuss a fermionic lattice associated to the fermionic polaron (FP) model . The DNLS models quantized by the canonical quantization method lead to a configuration of $`f`$ sites, described by tensoring the Weyl-Heisenberg (WH) algebra attached to each lattice site. The ordering rule adopted when the classical wave amplitude variables are substituted by boson operators is essential for the form of the obtained Hamiltonian operator. Our method of solving the resulting Schrödinger equation is based on the assumption that the lattice state vector can be factorized to a tensor product of coherent state vectors () each one living in the Hilbert space associated to each lattice site. This leads to a set of Hamiltonian equations for the CSV coordinates, which provide an approximate quasiclassical solution for the exact quantum dynamics. As the CS are the most classical state vectors for boson systems the resulting equations are quasiclassical in nature and govern the motion of CS wavepackets. To elucidate the meaning of the ansatz we recall that the boson CS labels are coordinates of the canonical phase plane and therefore the quasiclassical dynamics occurs in the $`f`$-fold cartesian product of harmonic oscillator phase space. That dynamics is the classical dynamics if the normal ordering of the bosons is used and diverges from that for symmetric ordering. Therefore the CS mean values for normally ordered observables coincide with their classical values in the former case while their quantum mechanical fluctuations are constant and minimum in the course of time evolution. To try the method to more general quantum systems we have also studied the FP model, which is first mapped by means of the Jordan-Wigner transformation to its equivalent XXZ model . The resulting spin Hamiltonian is expressed by tensoring the $`su(2)`$ algebra generators attached to each site of the model. By the same token as in the boson case the quantum dynamics is studied by assuming factorization of the state vector of the system to a product of spin CS. However the geometrical framework for the dynamics of the spin CSV’s is different in this case. The $`su(2)`$ CS space is a spherical surface with a noncanonical symplectic structure defined on it, as it results by looking at the so called ray metric of the CSV ; the classical Hamiltonian equations generate a flow in the $`f`$-fold cartesian product of spherical (or rather complex projective) coordinates.
The plan of the paper is as follows: in section two the general framework for the construction of CSV is outlined and the needed formulas for the special cases of the canonical and the spin CSV are provided. In the next section the DNLS model is quantized and the associated quasiclassical equations are derived and studied. Attention is paid to the problem of discrete self trapping (DST) phenomenon in the quantum regime, by illustrating the situation with plots of the Poisson distribution and the Q-function of the quantized modes of the system. Next section explains the physical meaning of the factoring of the lattice state vector as adopted for all models, and gives a validity measure for that approximation. In the following two appendices we illustrate the same idea by taking up first the MDNLS quantum lattice and then the FP model. The final section contains a number of conclusions and offers some prospects.
Throughout the text we consider $`\mathrm{}=1`$ and we denote CS expectation values of any generator with brackets around it.
## 2 Coherent state manifolds
The notion of coherent state vector was essentially introduced in the early days of quantum mechanics by Schrödinger , in his succesful attempt to construct a nonspreading Gaussian wavepacket for a quantum harmonic potential. The wavepacket center was evolving following the classical path while the dispersion was kept to a minimum compatible with the minimum uncertainty principle. The regeneration of the theory took place in the sixties and seventies when to a wealth of physical applications of CS a mathematical group theoretical foundation was also provided .
For our needs here a brief introduction of the CS concept goes as follows: consider a Lie group $`𝒢`$, with a unitary irreducible representation $`T(g)`$, $`g𝒢`$, in a Hilbert space $``$. We select a reference vector $`|\mathrm{\Psi }_0`$, to be called the ”vacuum” state vector, and let $`𝒢_0𝒢`$ be its isotropy subgroup, i.e. for $`h𝒢_0`$, $`T(h)|\mathrm{\Psi }_0=e^{i\phi (h)}|\mathrm{\Psi }_0`$. The map from the factor group $`=𝒢/𝒢_0`$ to the Hilbert space $``$, introduced in the form of an orbit of the vacuum state under a factor group element, defines a CSV $`|x=T(𝒢/𝒢_0)|\mathrm{\Psi }_0`$ labelled by points $`x`$ of the coherent state manifold. Coherent states form an (over)complete set of states, since by means of the Haar invariant measure of the group $`𝒢`$ viz.$`d\mu (x),x`$, they provide a resolution of unity, $`\mathrm{𝟏}=_{}d\mu (x)\left|xx\right|`$. As a consequence, any vector $`|\mathrm{\Psi }`$ is analyzed in the CS basis, $`|\mathrm{\Psi }=_{}𝑑\mu (x)\mathrm{\Psi }(x)|x`$, with coefficients $`\mathrm{\Psi }(x)=x|\mathrm{\Psi }`$.
What concerns us here mostly is the geometry of the CS manifold $``$. Indeed by its very construction $``$ inherits the structure of a Riemann manifold with in general non-constant curvature, which is also endowed with a complex structure of a Kähler manifold , namely it can be considered as a generalized phase space . In the sequel we restrict ourselves to the case of a two dimensonal surface $``$ for definiteness, although higher dimensional extensions of our statements are also possible . Also we shall assume a generator $`\mathrm{G}_+`$ creating from the vacuum state the CS vector, i.e.,
$$|\zeta =𝒩||\zeta )=𝒩\mathrm{exp}(\zeta \mathrm{G}_+)|,$$
(1)
where $`\zeta C`$ and $`𝒩=(\zeta ||||\zeta )^{\frac{1}{2}}`$ is the normalization factor. Occasionally we shall write $`\mathrm{G}_{}`$ for the Hermitean conjugate of $`\mathrm{G}_+`$; they should both belong to the Lie algebra of the group for which the CS is defined. Below we shall specialize to the cases of $`su(2)`$ and the Weyl-Heisenberg groups but for the moment we proceed with the present general framework.
The symplectic structure possessed by the state space $``$ is based on the existence of a canonical kinematical 1-form $`\theta =\zeta \left|d\right|\zeta ={\scriptscriptstyle \frac{1}{2}}(G_+d\zeta G_{}d\overline{\zeta })`$. The derivation $`d=_\zeta d\zeta +_{\overline{\zeta }}d\overline{\zeta }`$ acts on the state vectors, e.g., $`d|\zeta =(G_+{\scriptscriptstyle \frac{1}{2}}G_+)|\zeta d\zeta +({\scriptscriptstyle \frac{1}{2}}G_{})|\zeta d\overline{\zeta }`$. The symplectic 2-form $`\omega `$ of $``$ is derived from the canonical $`\theta `$ by derivation $`\omega =d\theta `$, and can be expressed in the form $`\omega =(G_{}G_+G_{}G_+)d\zeta d\overline{\zeta }`$.
Concerning the geometric features of our $`2D`$ phase space considered as a Riemannian surface with a distance function operating on it, we shall employ a meaningful metric tensor starting from the distance on the projective Hilbert space of rays $`\overline{|\zeta }`$, $`\zeta C`$ . By $`\overline{|\zeta }`$ we simply mean the set of CS $`e^{i\varphi }|\zeta `$, multiplied by an arbitrary phase factor. The choice of space of rays rather than the Hilbert space is in accordance with the quantum mechanical arbitrariness of phase of the state vectors. The finite distance $`𝒟(\overline{|\zeta _1},\overline{|\zeta _2})`$ between any two rays $`\overline{|\zeta _1}`$ and $`\overline{|\zeta _2}`$, which are associated with the normalized coherent states $`e^{i\phi _1}|\zeta _1`$ and $`e^{i\phi _2}|\zeta _2`$, is defined by
$$𝒟^2(\overline{|\zeta _1},\overline{|\zeta _2})=\underset{\phi _1,\phi _2}{inf}e^{i\phi _1}|\zeta _1e^{i\phi _2}|\zeta _2^2=22\left|\zeta _1|\zeta _2\right|.$$
(2)
This is a proper distance function as it is positive definite and non-degenerate, it satisfies the triangular property and it is also gauge invariant. In addition its infinitesimal form gives the distance of two nearby coherent state vectors and provides the metric tensor on the CS-manifold , i.e.
$`(ds)^2`$ $`=`$ $`{\scriptscriptstyle \frac{1}{2}}𝒟^2(\overline{|\zeta +d\zeta },\overline{|\zeta })=1\left|\zeta +d\zeta |\zeta \right|`$ (3)
$`=`$ $`\left[{\displaystyle \frac{_\zeta _{\overline{\zeta }}(\zeta ||||\zeta )}{(\zeta ||||\zeta )}}{\displaystyle \frac{_{\overline{\zeta }}(\zeta ||||\zeta )}{(\zeta ||||\zeta )}}{\displaystyle \frac{_\zeta (\zeta ||||\zeta )}{(\zeta ||||\zeta )}}\right]d\zeta d\overline{\zeta }`$
$`=`$ $`{\displaystyle \frac{}{\left|\zeta \right|^2}}\left[{\displaystyle \frac{\left|\zeta \right|^2}{(\zeta ||||\zeta )}}{\displaystyle \frac{(\zeta ||||\zeta )}{\left|\zeta \right|^2}}\right]d\zeta d\overline{\zeta }g_{\zeta \overline{\zeta }}d\zeta d\overline{\zeta },`$
or alternatively
$$(ds)^2=\left[\frac{(\zeta G_{}G_+\zeta )}{(\zeta ||||\zeta )}\frac{(\zeta G_{}\zeta )}{(\zeta ||||\zeta )}\frac{(\zeta G_+\zeta )}{(\zeta ||||\zeta )}\right]d\zeta d\overline{\zeta }.$$
(4)
The curvature scalar ,
$$R=g_{\zeta \overline{\zeta }}^1_\zeta _{\overline{\zeta }}(\mathrm{ln}g_{\zeta \overline{\zeta }}),$$
(5)
following from the metric $`g`$ involves higher correlations of the $`G_\pm `$ generators and is in general not constant. It is worth noticing that the basic geometric objects of both the symplectic and Riemannian structure endowed in $``$ are given in terms of the so called symbols of the operators or their products , i.e. the coherent state mean value $`\zeta \left|\mathrm{G}\right|\zeta \mathrm{G}`$ of the corresponding operator $`\mathrm{G}`$. Also the noncommutativity of the involved operators and their non-zero uncertainties in the coherent state basis is essential for the non-trivial geometric characteristics of the $``$ manifold. In effect, the CS-manifold $``$ captures some genuine quantum mechanical features despite its classical character; this property of $``$ is further manifested in the quantum mechanical evolution to be studied shortly.
As the models to be studied in the next sections are of boson and fermionic/spin type with dynamical symmetries related to the WH and the $`su(2)`$ algebras , we now exemplify the above construction for the CS of these algebras. For the boson algebra,
$$[a,a^{}]=\mathrm{𝟏}[N,a^{}]=a^{}[N,a]=a,$$
(6)
with $`N=a^{}a`$ the number operator, the vacuum state is the zero-photon state vector $`|=|0`$, and $`\mathrm{G}_+=a^{}`$ the creation operator. This defines the boson CS
$$|\alpha =e^{\alpha a^{}\overline{\alpha }a}|0=𝒩e^{\alpha a^{}}|0=e^{\frac{1}{2}\left|\alpha \right|^2}\underset{n=0}{\overset{\mathrm{}}{}}\frac{\alpha ^n}{\sqrt{n!}}|n.$$
(7)
It is an (over)complete set of states with respect to the measure $`d\mu (\alpha )=\frac{1}{\pi }e^{\left|\alpha \right|^2}d^2\alpha `$ for the non-normalized CS, and $`\alpha =WH/U(1)C`$ is the CS manifold. Since $`a|\alpha =\alpha |\alpha `$, which implies the symbols $`a=\alpha `$, $`a^{}=\overline{\alpha }`$, $`N=\left|\alpha \right|^2`$, $``$ is the flat canonical phase plane with the standard line element $`ds^2=d\alpha d\overline{\alpha }`$. Also the symplectic 2-form $`\omega =id\alpha d\overline{\alpha }`$ is associated to the canonical Poisson bracket $`\{f,g\}=i(_\alpha f_{\overline{\alpha }}g_{\overline{\alpha }}f_\alpha g)`$.
Next we come to the case of the $`su(2)`$ algebra with commutation relations,
$$[J_0,J_\pm ]=\pm J_\pm [J_+,J_{}]=2J_0.$$
(8)
The vacuum state $`||jj`$ is the extremal vector in the representation module of the algebra, with $`\{|jm\}_{m=j}^j`$, $`j=1/2,1,3/2,\mathrm{}`$, and $`jmj`$. The CS vector is obtained by displacing the vacuum with a $`SU(2)/U(1)`$ coset element:
$$|z=e^{\nu J_+\overline{\nu }J_{}}|jj=𝒩e^{zJ_+}|jj=𝒩\underset{m=j}{\overset{j}{}}\left(\begin{array}{c}2j\\ j+m\end{array}\right)^{\frac{1}{2}}z^{j+m}|jm,$$
(9)
with $`z=(\nu /|\nu |)\mathrm{tan}|\nu |`$ and normalization coefficient $`𝒩=(1+z\overline{z})^j`$. The CS manifold $`=SU(2)/U(1)\mathrm{S}^2\mathrm{CP}^1`$ is isomorphic to the 2-sphere $`\mathrm{S}^2`$, with $`0\theta \pi ,0\phi 2\pi `$ polar and azimuthal angles respectively. It is also parametrized as the complex projection plane $`\mathrm{CP}^1`$ with coordinate $`z=\mathrm{tan}({\scriptscriptstyle \frac{1}{2}}\theta )e^{i\phi }`$.
The symbols of the generators are
$`J_+`$ $`=`$ $`2j{\displaystyle \frac{\overline{z}}{1+z\overline{z}}},J_{}=2j{\displaystyle \frac{z}{1+z\overline{z}}}`$
$`J_0`$ $`=`$ $`j{\displaystyle \frac{1z\overline{z}}{1+z\overline{z}}},J_{}J_+={\displaystyle \frac{4j^2z\overline{z}+2j}{(1+z\overline{z})^2}}.`$ (10)
Based on the general scheme the above generator symbols give rise to the line element $`ds^2=(1+z\overline{z})^2dzd\overline{z}`$ for the complex tangent plane $``$. The (over)completeness of the non-normalized CS invokes the Haar invariant measure on $``$, $`d\mu (z)=\frac{2j+1}{\pi }(1+z\overline{z})^2d^2z`$. The CS vector space is now the constant curvature ($`R=1`$) sphere equipped with the canonical 1-form $`\theta =ij\frac{\overline{z}}{1+z\overline{z}}dz+\mathrm{cc}`$, which furnishes a canonical symplectic 2-form $`\omega =ij\frac{dzd\overline{z}}{(1+z\overline{z})^2}`$, and the associated Poisson bracket becomes $`\{f,g\}=\frac{i(1+z\overline{z})^2}{2j}(_zf_{\overline{z}}g_{\overline{z}}f_zg)`$.
## 3 Generalized discrete self-trapping equation
The classical Hamiltonian of the generalized discrete self-trapping (GDST) model reads
$$H_{\mathrm{CL}}=\underset{j=1}{\overset{f}{}}(\omega _0|A_j|^2\frac{\gamma }{m}|A_j|^{2m})\underset{jk}{\overset{f}{}}\lambda _{jk}\overline{A}_jA_k,$$
(11)
where $`j=1,2,\mathrm{},f`$ counts the number of complex mode amplitudes $`A_j(t)`$ and its complex conjugate $`\overline{A}_j(t)`$. The equation of motion
$$i\dot{A}_j=\omega _0A_j\underset{(jk)k=1}{\overset{f}{}}\lambda _{jk}A_k\gamma |A_j|^{2(m1)}A_j$$
(12)
is derived from the canonical Poisson brackets
$$\{A_j,\overline{A}_k\}=i\delta _{jk},\{A_j,A_k\}=\{\overline{A}_j,\overline{A}_k\}=0$$
(13)
using the equation of motion $`\dot{A}_j=\{H,A_j\}.`$
The quantization of this canonical Lie-Poisson algebra proceeds with the correspondance rule $`Ab,A^{}b^{}`$, where $`b^{}`$ and $`b`$ are the canonical creation and annihilation operators respectively. To account for problems of ordering, well known in the naive quantization rule, we shall here confine ourselves to two ordering rules, namely the normal ordering (NO) and the symmetric ordering (SO). SO is a classically motivated symmetrization of noncommuting operators and should be expected to give the right classical limit when $`\mathrm{}0`$. However here we study the quasiclassical limit of the quantum equations of motion by means of the CS symbol of operators and in this case the NO is the appropriate ordering. (For further discussion on the difference between semi- and quasiclassical ordering see refs. 89 and 142 in , ff p.64-65).
For the SO quantization rule the correspondence
$$|A|^{2m}\frac{m!}{2^m}L_m(2x),$$
(14)
is valid for quantization of arbitrary positive powers of the modulus of the complex amplitude; here $`x^k=:(b^{}b)^k:b^kb^k`$ and $`L_m()`$ is the Laguerre polynomial of zero order. Then, using the symmetric ordering quantization rule and after a constant shifting by $`\frac{1}{2}\omega _0\gamma \frac{(m1)!}{2^m}`$, the Hamiltonian of the GDST model takes the form
$$H_{\mathrm{SO}}=H_{\mathrm{NO}}\gamma \underset{j=1}{\overset{f}{}}\underset{n=1}{\overset{m1}{}}\mu _n^{(m)}b_{j}^{}{}_{}{}^{n}b_{j}^{}{}_{}{}^{n},$$
(15)
where $`\mu _n^{(m)}=\frac{(m1)!}{2^m}\left(\begin{array}{c}m\\ mn\end{array}\right)\frac{2^n}{n!}`$, and
$$H_{\mathrm{NO}}=\underset{j=1}{\overset{f}{}}(\omega _0b_j^{}b_j\frac{\gamma }{m}b_j^mb_j^m)\underset{jk}{\overset{f}{}}\lambda _{jk}b_j^{}b_k,$$
(16)
is the Hamiltonian obtained by the normal ordering quantization rule. We observe that the symmetric ordering rule adds $`m1`$ additional terms in the Hamiltonian with respect to its NO quantization. However since for $`m=2`$ it is reduced to the number operator which can be absorbed in the $`H_{\mathrm{NO}}`$ by simply redefining the coefficient of the corresponding number operator term, we see that in effect the symmetric ordering modifies the Hamiltonian only for $`m>2`$, i.e. only for higher order nonlinearity.
Next we obtain the Heisenberg equations of motion for the boson operators by using the Hamiltonian of eq. (15)
$$i\dot{b}_j=\omega _0b_j\gamma b_j^{m1}b_j^m\underset{k=1}{\overset{f}{}}\lambda _{jk}b_k\gamma (\underset{n=1}{\overset{m1}{}}\mu _n^{(m)}nb_{j}^{}{}_{}{}^{n1}b_{j}^{}{}_{}{}^{n}).$$
(17)
Making use of the factorized state vector $`|\beta >_{j=1}^f|\beta _j>`$, where $`b_j|\beta _j>=\beta _j|\beta _j>,j=1,\mathrm{},f`$, we evaluate the CS mean value of the last equation to obtain
$$i\dot{\beta }_j=\omega _0\beta _j\gamma |\beta _j|^{2m2}\beta _j\underset{(jk)k=1}{\overset{f}{}}\lambda _{jk}\beta _k\gamma \underset{n=1}{\overset{m1}{}}\mu _n^{(m)}n|\beta _j|^{2n2}\beta _j.$$
(18)
By means of the canonical Poisson brackets this set of equations can be derived from the Hamiltonian
$$=\beta |H|\beta =\underset{k=1}{\overset{f}{}}(\omega _0|\beta _k|^2\frac{\gamma }{m}|\beta _k|^{2m})\underset{kl}{\overset{f}{}}\lambda _{kl}\beta _k\overline{\beta }_l\gamma \underset{k=1}{\overset{f}{}}\underset{n=1}{\overset{m1}{}}\mu _n^{(m)}|\beta _k|^{2n}.$$
(19)
The above Hamiltonian and the norm $`N=_{j=1}^f|\beta _j|^2`$ are the constants of motion for the dynamics issued by eq. (18). We observe that the first two sums in eq. (19) combine to give the coherent state symbol of the $`H_{\mathrm{NO}}`$ of eq. (16), while the remaining part corresponds to the symbol of the terms appearing in eq. (15) due to the SO. This implies that NO quantization induces in the CS manifold $``$ an essentially classical time evolution while the quantization by SO, due to the additional terms, induces a dynamical evolution on $``$ which departs from classical dynamics. This of course makes a difference in the dynamics of the observables of the system as well as in the wavefunction distributions. At this point we should remark that the assumption of factorization of the state vector into CSVs implies that any normal ordered function of boson operators from each mode has CS mean value expressed by the same function, i.e.
$$\alpha |:f(a_j^{},a_j):|\alpha =f(\overline{\alpha }_j,\alpha _j).$$
(20)
Also for each oscillator mode the Hermitian combinations $`x=\frac{1}{\sqrt{2}}(a^{}+a)`$ and $`p=\frac{i}{\sqrt{2}}(a^{}a)`$, i.e. the position and momentum operators, we obtain the minimum uncertainty property of the boson CSV,
$$\mathrm{\Delta }x\mathrm{\Delta }p=\frac{1}{2},$$
(21)
for the uncertainties $`(\mathrm{\Delta }x)^2=(xx)^2`$ and $`(\mathrm{\Delta }p)^2=(pp)^2`$. We shall recall this property shortly for the Q-function.
To illustrate these differences in dynamics due to ordering we shall employ the quantum mechanical Q-distribution function and the Poissonian distribution of the boson excitations of a CSV in the number state basis . For the Q-distribution of a $`\rho `$ density matrix we define
$$Q_j=\frac{1}{\pi }\beta _r\left|\rho _j\right|\beta _r=\frac{1}{\pi }\left|\beta _r|\beta _j\right|^2,$$
(22)
with $`j=1,\mathrm{},f`$. Here we have assumed a pure state for each mode i.e. $`\rho _j=\left|\beta _j\beta _j\right|`$, and the reference CS $`|\beta _r`$ is labelled by the complex coordinate $`\beta _r=x_r+iy_r`$. The Q-function then takes the form of a displaced Gaussian function viz.
$$Q_j(x_r,y_r)=\frac{1}{\pi }e^{(x_r\mathrm{Re}\beta _j)^2(y_r\mathrm{Im}\beta _j)^2}.$$
(23)
The center of the Gaussian $`(\mathrm{Re}\beta _j,\mathrm{Im}\beta _j)`$ is the expectation value of the position and momentum operators in the CS basis, i.e. $`(\mathrm{Re}\beta _j,\mathrm{Im}\beta _j)=(\beta |x_j|\beta ,\beta |p_j|\beta )`$, for $`j=1,\mathrm{},f`$. However there is a nonzero quantum mechanical uncertainty around this point since $`\mathrm{\Delta }x_j=\mathrm{\Delta }p_j=\frac{1}{\sqrt{2}}`$, albeit it is the minimum acceptable one (c.f . eq.(21)).
In Figs. 1(a,b) the Gaussian bells of the Q-functions for the quintic $`(m=3)`$ GDST trimer $`(f=3)`$ system are plotted at a certain instant of time for SO and NO respectively. Here and in the following we use periodic boundary conditions and choose $`\lambda _{jk}=1`$ for nearest neighbouring sites, and zero otherwise. The centers of the Gaussian functions of Fig. 1 are plotted in Fig. 2 for the SO (Fig. 2a,b) and the NO (Fig. 2 c,d) cases, respectively. We note that in the SO case the trajectory of the bell corresponding to the initially excited site will always be confined in a region far from the origin, while the bells corresponding to the other two sites will stay close to the origin. In our framework, this corresponds to the well known self-trapping of the classical system. It is obvious that the appearance of the extra terms in the SO-equation enhances the self-trapping compared to the NO-case.
Next we consider the Poissonian distribution of each boson mode in the coherent state,
$$P_n^j=\left|\beta _j|n\right|^2=e^{\left|\beta _j\right|^2}\frac{\left|\beta _j\right|^{2n}}{n!},$$
(24)
where $`n=0,1,\mathrm{}`$ enumerates the number states and $`\left|\beta _j\right|^2=\beta _j|b_j^{}b_j|\beta _j`$ is the expectation value of the number operator which gives the average number of quanta in the CS basis. An illustration of how the self-trapping reflects itself in the Poissonian distribution and how it gets enhanced in SO quantization, is shown in Fig. 3a and 3b for the SO and NO cases respectively. The system is here a rather large $`(f=21)`$ quintic $`(m=3)`$ GDST-system, and initially one single site $`(j=21)`$ is excited with the total excitation number $`N=10`$. As can be seen, for the particular parameter values chosen here the main part of the excitation stays at the initially excited site in the SO-case, while it spreads more or less equally among the modes in the NO case.
To end this section, we note that in the simple case of the GDST dimer $`(f=2)`$ with cubic $`(m=2)`$ or quintic $`(m=3)`$ nonlinearity, it is possible to derive an analytic expression for the critical value,$`\gamma _{\mathrm{cr}}`$, of the nonlinear coupling coefficient for the appearance of self-trapping in eq. (18). This is done in the standard way by considering the variable $`r=|\beta _1|^2|\beta _2|^2`$. For the ordinary GDST equation (12), it was found in ref. for the cubic case and in ref. for the quintic case, that $`\ddot{r}`$ could be expressed as a cubic polynomial in $`r`$, and solutions in terms of Jacobi elliptic functions were obtained. In the case of eq. (18), containing both cubic and quintic nonlinearity, we find the same equation for $`\ddot{r}`$ as in ref. , provided that we replace $`\gamma `$ with $`\gamma ^{(2)}+N\gamma ^{(3)}`$, where $`\gamma ^{(2)}`$ and $`\gamma ^{(3)}`$ are the coefficients of the cubic and quintic nonlinear terms, respectively. Thus, defining $`\gamma _{\mathrm{cr}}`$ in the usual way as the smallest value of $`\gamma `$ for which $`\left|\beta _{j_0}\right|^2`$ is always larger than $`N/2`$, given that the excitation initially was localized on site $`j_0`$, we find the self-trapping condition $`\frac{\gamma ^{(2)}+N\gamma ^{(3)}}{\lambda _{jk}}>\frac{4}{N}`$ from the explicit solution for $`r(t)`$ just as in ref. . Using the explicit expressions for $`\gamma ^{(2)}`$ and $`\gamma ^{(3)}`$ in (18) and $`\lambda _{jk}=1`$, the following expressions for $`\gamma _{\mathrm{cr}}`$ are obtained in the NO and SO cases, respectively:
$$\gamma _{\mathrm{cr}}^{\mathrm{NO}}=\frac{4}{N^2},\gamma _{\mathrm{cr}}^{\mathrm{SO}}=\frac{4}{N(N+3)}.$$
(25)
These expressions are plotted in Fig. 4 as a function of the total number of lattice quanta $`N`$. In the limit of large number of excitations (phonons) the quantum lattice model, according to Bohr’s correspondance principle, will behave classically. In particular the critical value of $`\gamma `$ for the onset of self-trapping, which as we have also seen from previous plots is different for SO and NO, is expected to become equal in the classical limit of the quantum lattice models. In this limit the ordering of the operators, a genuine quantum characteristic, should not be of importance. Indeed the analytic expressions (25) for the critical $`\gamma `$’s of the GDST dimer show, that when the number of quanta becomes large, the phenomenon of DST occurs asymptotically for the same values i.e. $`\gamma _{\mathrm{cr}}^{\mathrm{NO}}\gamma _{\mathrm{cr}}^{\mathrm{SO}}`$, independently of the ordering rule.
## 4 Validity of the factorization ansatz and quantum correlations
Let a composite quantum system be described by a general pure state $`|\psi `$ living in $`=_{i=1}^f_i`$, the tensor product of Hilbert spaces of its $`f`$ subsystems. In general the form of such a generic vector $`|\psi `$ can be of three different types, each one signifying a specific kind of correlation among the subsystems. For vectors of the product form, $`|\psi =|\varphi ^1|\varphi ^2\mathrm{}|\varphi ^f`$, the subsystems are decorrelated. If $`|\psi `$ is a convex combination of product states i.e. $`|\psi =_i^n\lambda _i|\varphi _i^1|\varphi _i^2\mathrm{}|\varphi _i^f`$, with $`_i^n\lambda _i=1`$, the subsystems are said to be classically correlated . For inseparable or entangled systems, the third case, the state vector $`|\psi `$ is a general superposition of states from its subsystems with no additional properties, that are further characterized by e.g. violation of Bell inequalities ; this is the case when genuine quantum correlations are developed among the subsystems.
Although the existence and the oddities of quantum nonlocal correlations and of entanglement have been pointed out since the early days of Quantum Mechanics , and have remained an active subject of research since then , it is only recently (due mainly to some striking applications of entanglement such as quantum computation, cryptography, teleportation etc; see and refs. therein ), that measures that quantify the amount of correlations of a given state of a composite quantum system have been studied. In the above framework of quantum correlations it is clear that our factorization ansatz is equivalent to the assumption that no quantum correlations will be developed in the course of time evolution. Still, as we have seen, due to the nonlinear form of the quantum Hamiltonian the equations of motions derived from that approximate ansatz differ from the entirely classical ones. This difference captures some of the quantum features of the system, so we would like to have a figure of merit of our approximation. To this end we introduce a single index, the correlation index, that measures the distance of the true state of the model from the factorized one that is assumed, and so gauges the validity of our approximation.
Recall that the CS-vectors form a basis for the Hilbert space of a single site, so for the $`f`$ sites of the boson chain model a general state vector can be expanded as
$$|\psi =P(\{\alpha _i\},\{\overline{\alpha }_i\})_{i=1}^f|\alpha _i𝑑\mu (\{\alpha _i\})$$
(26)
with $`d\mu (\{\alpha _i\})=_{i=1}^fd^2\alpha _i`$. By means of the relations (),
$`f(\overline{\beta })`$ $`=`$ $`{\displaystyle f(\overline{\alpha })\mathrm{exp}[{\scriptscriptstyle \frac{1}{2}}(\left|\alpha \right|^2+\left|\beta \right|^2)+\overline{\beta }\alpha ]d^2\alpha },`$ (27)
$`f(\overline{\beta })`$ $`=`$ $`{\displaystyle f(\overline{\alpha })\beta |\alpha d^2\alpha },`$
which imply that the overlap $`\beta |\alpha `$ of CS behaves as a delta function for the integrable functions of the Bargmann-Hilbert space, we can compute that $`|\psi `$ is normalized provided that $`\left|P(\{\alpha _i\},\{\overline{\alpha }_i\})\right|^2𝑑\mu (\{\alpha _i\})=1.`$ This shows that if $`P(\{\alpha _i\},\{\overline{\alpha }_i\})=\delta (\alpha _1\alpha _1^r)\mathrm{}\delta (\alpha _f\alpha _f^r)`$, then the general vector $`|\psi `$ is reduced to the factorized state $`|\psi _{ref}=_{i=1}^f|\alpha _i^r.`$
We now introduce the correlation index $`ϵ`$ (c.f eq. (2) ),
$`ϵ`$ $`=`$ $`𝒟^2(\overline{|\psi },\overline{|\psi _{ref}})=22\left|\psi |\psi _{ref}\right|`$
$`=`$ $`22\left|{\displaystyle 𝑑\mu (\{\alpha _i\})P(\{\alpha _i\},\{\overline{\alpha }_i\})\underset{i=1}{\overset{f}{}}\mathrm{exp}[{\scriptscriptstyle \frac{1}{2}}(\left|\alpha _i^r\right|^2+\left|\alpha _i\right|^2)+\overline{\alpha }_i^r\alpha _i]}\right|.`$
Initially $`ϵ(t=0)=0`$, since we set $`|\psi (t=0)=|\psi _{ref}`$. In any subsequent time interval $`atb`$, for which $`ϵ(t=0)0`$, our ansatz is justified. To determine $`ϵ(t)`$, we need to know the kernel $`P(\{\alpha _i\},\{\overline{\alpha }_i\})`$. Substituting the expansion (26) in the Schrödinger equation we find that the kernel $`P`$ satisfies the Fokker-Planck equation,
$$i\mathrm{}_tP(\{\alpha _i\},\{\overline{\alpha }_i\})=(\alpha _i,_{\alpha _i}+{\scriptscriptstyle \frac{1}{2}}\overline{\alpha }_i)P(\{\alpha _i\},\{\overline{\alpha }_i\}),$$
(29)
where $``$ is the analytic image of the Hamiltonian operator under the substitutions $`a_i\alpha _i,a_i^{}\overline{}_{\alpha _i}`$. There are a number of approximate and analytic solution techniques available for such equations , most accessible e.g. is the technique of continued fractions. However the problem of quantum correlations in nonlinear lattice models is interesting even in the dimer case, and as far as we know it has not been addressed. In such a case the conservation of the total number of excitation quanta in the lattice sites will make the number of terms in a continued fraction method of solving the ensuing Fokker-Planck equation finite, since the dimensionality of the available Hilbert space and of the $`\rho `$-density matrix is determined by the number of quanta available. This makes possible an explicit calculation of the correlation index and allows to set a number of interesting questions concerning the relation of correlation/decorrelation versus selftrapping/non-selftrapping in the case of the quantum dimer (see e.g. for a full quantum treatment of the quantum dimer); however more work should be anticipated along these lines . Finally, we note that for non-boson lattices, as the ones of the last appendix, the introduction of the correlation index can be done again based on the distance function defined for the respective CSV, c.f eq.(2).
## 5 Conclusions
This work was inspired by the fact that for a linear quantum mechanical oscillator there is a coherent state vector defined in its Hilbert space, predicting that the motion will be centered around the classical path with the minimum uncertainty. Here, this simple idea has been extended to some quantum nonlinear chain models. For quantum chain models with an oscillator attached at each site, the obvious assumption which has been followed here is that each site is in a CS ; the same idea however can be applied for multispin-like models with some advantages as we have shown. The essential geometric nature of our assumption has also been stated in terms of the symplectic and Riemannian geometry of the CSV spaces. Other types of special states, such as e.g. even/odd CSV (symmetric/antisymmetric combination of boson CS) which are associated with different geometries of their CS spaces, could have been considered; the final choice depends on the dynamical symmetry of the given Hamiltonian and on the specific form of the nonlinear interaction terms.
Also, the treatment of the DST in the quantum regime as has been presented here shows that this is a phenomenon which continues to exist after quantization of the classical DST equations. However, as has been demonstrated by means of the CS method, its exact parameter dependence is crucially affected by the quantization scheme. Contrary to the NO case where the situation is essentially classical, in the SO rule DST gets even more pronounced from its classical form. We note that a similar statement can be made also concerning the Hartree approximation treated in , since this ansatz results into similar approximate dynamics as obtained here. This can be seen by comparing our eq. (18) with eqs. (4.3) and (4.4) of ref. . However, the Hartree ansatz is basically different from the factorization ansatz used here. Namely, geometrically it can be said that the two ansätze confine the dynamics in different subspaces of the total Hilbert space of the GDST model, moreover the Hartree wavefunction is not fully factorized.
Finally, we should note that although the factorization hypothesis for the state vector has been effective in treating the dynamical evolution in the model-cases studied here, its validity should be questioned. This is a crucial point, since factoring the state vector of a compound quantum system implies a loss of the quantum correlations which may be developed during the course of the time evolution. The failure of a factorized state to account for such entanglement of quantum subsystems can however be quantified by means of a correlation index, as we have shown. In future work we plan to present a study of the entanglement phenomenon versus DST in a quantum dimer model.
## 6 Acknowledgements
We thank A. C. Scott for valuable comments. One of us (D. E.) acknowledges support from the EU HCM programme under grant No. CHRX-CT93-0331, he is also grateful to the Institute of Mathematical Modelling (IMM) for hospitality and to G. P. Tsironis for discussions. M. J. acknowledges financial support from the Swedish Foundation for International Cooperation in Research and Higher Education. Constructive comments by the referee are also acknowledged.
## Appendix 1: Modified discrete nonlinear Schrödinger equation
The ability of the CS-method to extract the quantum mechanical characteristics of the time evolution for a given lattice Hamiltonian model is very much depending on the special form of the Hamiltonian itself. Occasionally the derived equations of motion will be reducible to the classical ones indifferently of the quantization ordering rule (at least for the common rules); let us take up the following case. The so called modified discrete nonlinear Schrödinger equation (MDNLS) has been recently introduced as a model describing in the adiabatic limit polaron dynamics when the excitation is coupled to an acoustic chain of oscillators. Compared to ordinary DNLS dynamics of an excitation coupled to optical oscillators, the MDNLS contains some new formal features such as off-diagonal nonlinearity, and for a ring with $`f`$-sites reads
$$i\dot{A}_j=V(A_{j+1}+A_{j1})X(|A_{j+1}|^2+|A_{j1}|^2+2|A_j|^2)A_j.$$
(30)
Using canonical Poisson brackets, these equations can be derived from the Hamiltonian
$$H_{\mathrm{CL}}=V\underset{j=1}{\overset{f}{}}(A_j\overline{A}_{j+1}+\overline{A}_jA_{j+1})X\underset{j=1}{\overset{f}{}}(|A_{j+1}|^2|A_{j+2}|^2+|A_j|^4).$$
(31)
The quantization of the classical Hamiltonian for the MDNLS equation leads (after a constant shifting of the Hamiltonian) to the following expression
$$H_{\mathrm{SO}}=H_{\mathrm{NO}}{\scriptscriptstyle \frac{1}{2}}X\underset{j=1}{\overset{f}{}}(b_{j+2}^{}b_{j+2}+b_{j+1}^{}b_{j+1}+b_j^{}b_j),$$
(32)
where
$$H_{\mathrm{NO}}=V\underset{j=1}{\overset{f}{}}(b_jb_{j+1}^{}+b_j^{}b_{j+1})X\underset{j=1}{\overset{f}{}}(b_{j+1}^{}b_{j+1}b_{j+2}^{}b_{j+2}+b_j^^2b_j^2).$$
(33)
The first term in eq. (32) refers to the NO and the second one is obtained as before from the SO rule. Since the total Hamiltonian is written in terms of WH algebra elements defined at each site and by their products, we can follow exactly the same procedure as in the quantum DNLS case. Namely we assume the factorization of the state vector in terms of canonical CSVs. By the method of the previous section, and for the Hamiltonian $`H_{\mathrm{SO}}`$, this leads to derivation of a set of quasiclassical equations of motion satisfied by the CSV parameters $`A_j`$,
$$i\dot{A}_j=V(A_{j+1}+A_{j1})X(|A_{j+1}|^2+|A_{j1}|^2+2|A_j|^2)A_j+\frac{3}{2}XA_j.$$
(34)
The above expression shows that the quasiclassical equation differs by a constant shift of the coefficients of the onsite variables from the corresponding classical equation. With the transformation $`A_je^{\frac{3}{2}iXt}A_j`$ they become the classical equations. The formal similarity of classical and quasiclassical equations allows to conclude the following: the particular form of the MDNLS Hamiltonian after quantization by NO and SO generates, (on the geometrical space singled out by the factorization of the model’s quantum state in terms of boson CSV), an (almost) classical dynamics. This however as the previous sections show, should not be expected to be true in general.
## Appendix 2: Fermionic polaron model (FP)
To demonstrate the ability of the proposed method to study quantum dynamics for models others than quantized boson chains, we shall take up in this appendix the fermionic polaron (FP) model. It has been recently proposed to describe the interaction of electrons with optical phonons in one spatial dimension . This same model has been studied from the point of view of QISM and the NSM . Under the assumptions made in the Hamiltonian of the interacting electron-phonon system with periodic boundary conditions reads
$$H=ϵ\underset{j=1}{\overset{N}{}}a_j^{}a_jg\underset{j=1}{\overset{N}{}}(a_j^{}a_{j+1}+a_{j+1}^{}a_j)+V\underset{j=1}{\overset{N}{}}n_jn_{j+1},$$
(35)
where the fermion annihilation operator of a polaron at lattice site $`j`$, $`a_j`$ and its corresponding Hermitian conjugate $`a_j^{}`$ (the creation operator) generate the Grassmann algebra
$$[a_j,a_k]_+=[a_j^{},a_k^{}]_+=0,[a_j,a_k^{}]_+=\delta _{jk}.$$
(36)
The occupation number operator is defined by $`n_j=a_j^{}a_j`$, while $`g`$ is a parameter proportional to the overlapping integral and $`V`$ stands for the electron-phonon coupling constant (see for details). Using the equivalence of the FP model to the XXZ model provided in by means of a Jordan-Wigner transformation, we shall proceed to write the spin analogue of the fermionic model. The Grassmann algebra in eq. (36) is realized in terms of a tensor product of generators of the (spin $`1/2`$ Pauli) algebra by means of the following Jordan-Wigner transformation
$`a_j`$ $`=`$ $`\sigma _j^{}\mathrm{exp}\left(i\pi {\displaystyle \underset{k=1}{\overset{j1}{}}}(\sigma _k^z+{\scriptscriptstyle \frac{1}{2}}\mathrm{𝟏}_j)\right)`$
$`a_j^{}`$ $`=`$ $`\sigma _j^+\mathrm{exp}\left(i\pi {\displaystyle \underset{k=1}{\overset{j1}{}}}(\sigma _k^z+{\scriptscriptstyle \frac{1}{2}}\mathrm{𝟏}_j)\right)`$
$`n_j`$ $`=`$ $`\sigma _j^+\sigma _j^{}=({\scriptscriptstyle \frac{1}{2}}\mathrm{𝟏}_j+\sigma _j^z),`$ (37)
and its inverse
$$\sigma _j^{}=a_j\mathrm{exp}(i\pi \underset{k=1}{\overset{j1}{}}n_k),\sigma _j^+=a_j^{}\mathrm{exp}(i\pi \underset{k=1}{\overset{j1}{}}n_k)$$
$$\sigma _j^z=n_j{\scriptscriptstyle \frac{1}{2}}\mathrm{𝟏}_j,$$
(38)
where $`\sigma _j^\pm =\sigma _j^x\pm i\sigma _j^y`$, and $`\sigma _j^x,\sigma _j^y,\sigma _j^z`$ are the Pauli matrices and $`\mathrm{𝟏}_j`$ is the unit matrix at site $`j`$.
Using eq. (37) the transformed Hamiltonian of eq. (35) reads,
$$H=\frac{N}{2}(ϵ+\frac{V}{2})+(ϵ+V)\underset{j=1}{\overset{N}{}}\sigma _j^z+H_{\mathrm{XXZ}},$$
(39)
where
$$H_{\mathrm{XXZ}}=V\underset{j=1}{\overset{N}{}}\sigma _j^z\sigma _{j+1}^z2g\underset{j=1}{\overset{N}{}}(\sigma _j^x\sigma _{j+1}^x+\sigma _j^y\sigma _{j+1}^y),$$
(40)
or
$$H_{\mathrm{XXZ}}=V\underset{j=1}{\overset{N}{}}\sigma _j^z\sigma _{j+1}^zg\underset{j=1}{\overset{N}{}}(\sigma _j^+\sigma _{j+1}^{}+\sigma _j^{}\sigma _{j+1}^+).$$
(41)
We now proceed with the $`H_{\mathrm{XXZ}}`$ part of the Hamiltonian as the other terms are constants of motion. Following same tactics as for the boson models we acknowledge the fact that the $`H_{\mathrm{XXZ}}`$ Hamiltonian is embedded in the $`_{j=1}^Nsu_j(2)`$ algebra and make an appropriate ansatz about the form of the wavefunction. Namely, we assume it is written in the form $`|z=_{j=1}^N|z_j>`$, where $`|z_j>`$ is the $`su(2)`$ coherent state correspoding to the $`j`$-site of the chain. To determine the dynamics of the complex amplitudes $`z`$’s we use the relation $`zjJ_0=J_{}`$, which is obtained from eq. (2). Then the time derivative of this relation for $`j=1/2`$, gives the evolution equation of the complex amplitude at each site by means of the CS symbol of the Heisenberg equations, viz.
$`\dot{z}_j`$ $`=`$ $`{\displaystyle \frac{d/dt\sigma _j^{}1/2\sigma _j^z\sigma _j^{}d/dt1/2\sigma _j^z}{1/2\sigma _j^z^2}}`$ (42)
$`=`$ $`{\displaystyle \frac{i([\sigma _j^{},H]1/2\sigma _j^z+\sigma _j^{}[\sigma _j^z,H])}{1/2\sigma _j^z^2}}.`$
Straightforward evaluation of the CS symbols of the involved operators gives the evolution equation,
$`i\dot{z}_j`$ $`=`$ $`{\displaystyle \frac{1}{(1+|z_{j1}|^2)(1+|z_{j+1}|^2)}}`$ (43)
$`[Vz_j(12|z_{j1}|^2|z_{j+1}|^2)+(2g3g|z_j|^2)[z_{j1}(1+|z_{j+1}|^2)+z_{j+1}(1+|z_{j1}|^2)]`$
$`+`$ $`gz_j^2[\overline{z}_{j1}(1+|z_{j+1}|^2)+\overline{z}_{j+1}(1+|z_{j1}|^2)]]`$
Returning to the original model we compute the CS mean value of the fermion operators by use of eqs. (2) and (37). This yields
$$z(t)|a_j^{}|z(t)=\underset{k=1}{\overset{j1}{}}(2)z_k(t)|\sigma _k^z|z_k(t)z_j(t)|\sigma _j^+|z_j(t)=C_j(t)e^{i\theta _j(t)},$$
(44)
where
$$C_j(t)=\frac{\left|z_j(t)\right|}{1+\left|z_j(t)\right|^2}\underset{k=1}{\overset{j1}{}}\frac{1\left|z_k(t)\right|^2}{1+\left|z_k(t)\right|^2},$$
(45)
with $`z_j(t)=\left|z_j(t)\right|e^{i\theta _j(t)}`$.
Also, $`z(t)|a_k|z(t)=\overline{z(t)|a_k^{}|z(t)}`$, $`k=1,\mathrm{},N`$, while the average value of the fermion number operator is given by
$$z(t)|n_j|z(t)=z(t)|\sigma _j^z|z(t)+\frac{1}{2}=\frac{\left|z_j(t)\right|^2}{1+\left|z_j(t)\right|^2}.$$
(46)
A final remark concerns the applicability of the CS method to the XXZ model; it has recently come to our attention that this method also can treat the XXZ model of higher spins.
Figure captions
Fig. 1. The Q-functions (22) for the quintic $`(m=3)`$ GDST-trimer $`(f=3)`$ plotted at the time-instant $`t=76.4`$ for the cases of symmetric (a) resp. normal (b) ordering of the boson operators. Dashed line corresponds to the initially excited site $`j=2`$, while solid line corresponds to the sites $`j=1`$ and $`j=3`$. (They are equal for all times due to the symmetric initial condition.) In both cases, the total excitation number $`N=10`$, the nonlinear parameter $`\gamma =0.055`$, and the linear coupling coefficient $`\lambda _{jk}`$ is unity.
Fig. 2. The trajectories of the centers of the Gaussian bells in Fig. 1 for times $`t<260`$. The parameter-values are the same as in Fig. 1. Fig. 2(a,b) (Fig. 2(c,d)) corresponds to sites $`1`$ resp. $`2`$ for the SO (NO) case.
Fig. 3. The Poissonian probability distribution $`P_n^j`$ from eq. (24) plotted as a function of excitation number $`n`$ and site index $`j`$ for the quintic GDST $`(m=3)`$ with $`f=21`$ at the time-instant $`t=50`$. Fig. 3(a) shows the case of SO, while (b) corresponds to NO. In both cases, the site $`j=21`$ is initially excited, and the total excitation number is $`N=10`$. The value of the nonlinear parameter is $`\gamma =0.05`$, which gives self-trapping in the SO-equation, but not in the NO-equation.
Fig. 4. The analytic expressions of $`\gamma _{\mathrm{cr}}`$ eq. (25) for a GDST quantum dimer with quintic nonlinearity (i.e. $`f=2,m=3`$) plotted as a function of the total excitation number $`N`$ for normal (NO) and symmetric (SO) ordering. The value of the linear coupling coefficients is set equal to $`1`$ or to $`1/2`$ if we consider periodic boundary conditions. We notice that the two $`\gamma `$’s become asymptotically equal for the moderate value of $`N=10`$. |
warning/0001/math-ph0001002.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Classical charges interact through Coulomb forces, as one learns in every course on electromagnetism. Presumably the best realization in nature is a strongly ionized gas, for which the Darwin correction to the Coulomb forces is of importance, since under standard conditions the velocities cannot be considered small as compared to the velocity of light, cf. \[7, §65\]. Thus, given $`N`$ charges, with positions $`r_\alpha `$, velocities $`u_\alpha `$, charges $`e_\alpha `$, and masses $`m_\alpha `$, $`\alpha =1,\mathrm{},N`$, their motion is governed by the Lagrangian
$`_\mathrm{D}`$ $`=`$ $`{\displaystyle \underset{\alpha =1}{\overset{N}{}}}\left({\displaystyle \frac{1}{2}}m_\alpha u_\alpha ^2+{\displaystyle \frac{1}{8c^2}}m_\alpha ^{}u_\alpha ^4\right){\displaystyle \frac{1}{2}}{\displaystyle \underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}}{\displaystyle \frac{e_\alpha e_\beta }{4\pi |r_\alpha r_\beta |}}`$ (1.1)
$`+{\displaystyle \frac{1}{4c^2}}{\displaystyle \underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}}{\displaystyle \frac{e_\alpha e_\beta }{4\pi |r_\alpha r_\beta |}}\left(u_\alpha u_\beta +|r_\alpha r_\beta |^2(u_\alpha [r_\alpha r_\beta ])(u_\beta [r_\alpha r_\beta ])\right),`$
$`c`$ denoting the velocity of light. The first term is the kinetic energy with a $`u_\alpha ^4`$-correction of a strength $`m_\alpha ^{}`$ depending on the precise model ($`m_\alpha ^{}=m_\alpha `$ for a relativistic particle). The second term is the Coulomb potential, whereas the third term is the Darwin potential, which decays as the Coulomb potential and has a velocity dependent strength.
On a more fundamental level, the forces between the charges are mediated through the electromagnetic field. The instantaneous Coulomb-Darwin interaction is a derived concept only. To understand the emergence of such an interaction, in this paper we will investigate the coupled system, charges and Maxwell field, and we will prove that in a certain limit the motion of the charges is well approximated by the Lagrange equations for $`_\mathrm{D}`$.
Let us first describe how the charges are coupled to the Maxwell field. To avoid short-distance singularities, we assume that the charge is spread out over a distance $`R_\phi `$, which physically is of order of the classical electron radius. Thus charge $`\alpha `$ has a charge distribution $`\rho _\alpha `$ which for simplicity we take to be of the form
$$\rho _\alpha (x)=e_\alpha \phi (x),x\mathrm{IR}^3,$$
where the form factor $`\phi `$ satisfies
$$0\phi C_0^{\mathrm{}}(\mathrm{IR}^3),\phi (x)=\phi _r(|x|),\phi (x)=0\text{for}|x|R_\phi .$$
$`(C)`$
To distinguish the true solution from the approximation (1.1), the position of a charge $`\alpha `$ in the coupled system is denoted by $`q_\alpha `$ and its velocity by $`v_\alpha `$, $`\alpha =1,\mathrm{},N`$. The charges then generate the charge distribution $`\rho `$ and the current $`j`$ given by
$$\rho (x,t)=\underset{\alpha =1}{\overset{N}{}}\rho _\alpha (xq_\alpha (t))\text{and}j(x,t)=\underset{\alpha =1}{\overset{N}{}}\rho _\alpha (xq_\alpha (t))v_\alpha (t),$$
(1.2)
which satisfy charge conservation by fiat. The Maxwell field, consisting of the electric field $`E`$ and the magnetic field $`B`$, evolves according to
$$c^1\frac{}{t}B(x,t)=E(x,t),c^1\frac{}{t}E(x,t)=B(x,t)c^1j(x,t)$$
(1.3)
with the constraints
$$E(x,t)=\rho (x,t),B(x,t)=0.$$
(1.4)
The charges generate the electromagnetic field which in turn determines the forces on the charges through the Lorentz force equation
$$\frac{d}{dt}\left(m_{\mathrm{b}\alpha }\gamma _\alpha v_\alpha (t)\right)=d^3x\rho _\alpha (xq_\alpha (t))\left[E(x,t)+v_\alpha (t)B(x,t)\right],t\mathrm{IR},$$
(1.5)
for $`\alpha =1,\mathrm{},N`$. Here $`m_{\mathrm{b}\alpha }`$ is the bare mass of charge $`\alpha `$ and $`\gamma _\alpha `$ the relativistic factor $`\gamma _\alpha =(1v_\alpha ^2/c^2)^{1/2}`$, which ensures $`|v_\alpha |<c`$. Note that there are no direct forces acting between the particles. Eqns. (1.2)–(1.5) are known as Abraham model for $`N`$ charges.
We define the energy function by
$$(E,B,\stackrel{}{q},\stackrel{}{v})=\underset{\alpha =1}{\overset{N}{}}m_{\mathrm{b}\alpha }\gamma _\alpha +\frac{1}{2}d^3x[E^2(x)+B^2(x)],$$
(1.6)
with $`\stackrel{}{q}=(q_1,\mathrm{},q_N)`$ and $`\stackrel{}{v}=(v_1,\mathrm{},v_N)`$. It then may be seen that the initial value problem corresponding to (1.2)–(1.5) has a unique weak solution of finite energy and that $``$ is conserved by this solution, compare with for the case of a single particle.
We assume that initially the particles are very far apart on the scale set by $`R_\phi `$. Thus we require, for $`\alpha \beta `$, that
$$|q_\alpha (0)q_\beta (0)|\epsilon ^1R_\phi $$
(1.7)
with $`\epsilon >0`$ small. If particles would come together as close as $`R_\phi `$, our equations of motion are not trustworthy anyhow. In addition, we require that the initial velocities be small compared to the speed of light,
$$|v_\alpha (0)|\sqrt{\epsilon }c.$$
(1.8)
Subject to these restrictions, in essence, the initial electromagnetic field is chosen such as to minimize the energy function $``$ from (1.6), cf. Section 5.1 for precise statements and estimates. With these initial conditions, for the particles to travel a distance of order $`\epsilon ^1R_\phi `$ it will take a time of order $`\epsilon ^{3/2}R_\phi /c`$, which will be the time scale of interest. Thus physically we consider slow particles that are far apart, and we want to follow their motion over long times.
Next note that it takes a time of order $`\epsilon ^1R_\phi /c`$ for a signal to travel between the particles. This means that on the time scale of interest, retardation effects are small. If particles interact through Coulomb forces, as will have to be proved, the strength of the forces is of order $`\epsilon ^2`$ since the distance is of order $`\epsilon ^1R_\phi `$. Followed over a time span $`\epsilon ^{3/2}R_\phi /c`$, this yields a change in velocity of order $`\sqrt{\epsilon }c`$. On this basis we expect the orders of magnitude (1.7) and (1.8) to remain valid over times of order $`\epsilon ^{3/2}R_\phi /c`$. There is one subtle point here, however. The self-interaction of a charge with the fields renormalizes its mass. Thus in (1.1) the quantity $`m_\alpha `$ cannot be the bare mass of the charge, the electromagnetic mass has to be added.
In theoretical physics it is common practice to count the post-Coulombian corrections in orders of $`v/c`$ relative to the motion through pure Coulomb forces. Thus the Darwin term is the first correction and of order $`(v/c)^2`$. The next correction is of order $`(v/c)^3`$ and accounts for damping through radiation. If we push the Taylor expansion in Section 3 one term further, one obtains
$$\frac{d}{dt}\left(\frac{_\mathrm{D}}{u_\alpha }\right)=\frac{_\mathrm{D}}{r_\alpha }+(e_\alpha /\mathrm{\hspace{0.17em}6}\pi c^3)\underset{\beta =1}{\overset{N}{}}e_\beta \ddot{v}_\beta ,$$
(1.9)
$`\alpha =1,\mathrm{},N`$. The physical solution has to be on the center manifold for (1.9). At the present level of precision it suffices to substitute the Hamiltonian dynamics to lowest order, which yields
$`{\displaystyle \frac{d}{dt}}\left({\displaystyle \frac{_\mathrm{D}}{u_\alpha }}\right)={\displaystyle \frac{_\mathrm{D}}{r_\alpha }}`$
$`+{\displaystyle \frac{e_\alpha }{6\pi c^3}}{\displaystyle \frac{1}{2}}{\displaystyle \underset{\stackrel{\beta ,\beta ^{}=1}{\beta \beta ^{}}}{\overset{N}{}}}\left({\displaystyle \frac{e_\beta }{m_\beta }}{\displaystyle \frac{e_\beta ^{}}{m_\beta ^{}}}\right){\displaystyle \frac{e_\beta e_\beta ^{}}{4\pi |r_\beta r_\beta ^{}|^3}}\left((u_\beta u_\beta ^{})3{\displaystyle \frac{(r_\beta r_\beta ^{})(u_\beta u_\beta ^{})}{|r_\beta r_\beta ^{}|^2}}(r_\beta r_\beta ^{})\right).`$
Note that if the ratio $`e_\alpha /m_\alpha `$ does not depend on $`\alpha `$, then the radiation reaction vanishes and the system does not emit dipole radiation. The next order correction is $`(v/c)^4`$ and of Lagrangian form. It is discussed in and .
In general relativity, there is a huge effort to obtain corrections to the Newtonian orbits, which as a problem is similar to the one discussed here. The most famous example is the Hulse-Taylor binary pulsar, where two highly compact neutron stars of roughly solar mass revolve around each other with a period of 7.8 h . In this case $`(v/c)=10^3`$. For gravitational systems there is only quadrupole radiation which is of order $`(v/c)^5`$. To this order the theory agrees with the observed radio signals within 0,3%. In newly designed experiments one expects highly improved precision which will require corrections up to order $`(v/c)^{11}`$.
## 2 Main results
We recall the initial conditions for the Abraham model (1.2)–(1.5), where we set $`c=1`$ throughout for simplicity. For the initial positions $`q_\alpha ^0=q_\alpha (0)`$ we require
$$C_1\epsilon ^1|q_\alpha ^0q_\beta ^0|C_2\epsilon ^1,\alpha \beta ,$$
(2.1)
for some constants $`C_1,C_2>0`$. For the initial velocities $`v_\alpha ^0=v_\alpha (0)`$ we assume
$$|v_\alpha ^0|C_3\sqrt{\epsilon }$$
(2.2)
with $`C_3>0`$. The initial fields are a sum over charge solitons,
$$E(x,0)=E^0(x)=\underset{\alpha =1}{\overset{N}{}}E_{v_\alpha ^0}(xq_\alpha ^0)\text{and}B(x,0)=B^0(x)=\underset{\alpha =1}{\overset{N}{}}B_{v_\alpha ^0}(xq_\alpha ^0).$$
(2.3)
Here
$$E_v(x)=\varphi _v(x)+(v\varphi _v(x))v\text{and}B_v(x)=v\varphi _v(x)$$
(2.4)
and the Fourier transform of $`\varphi _v`$ is given by
$$\widehat{\varphi }_v(k)=e\widehat{\phi }(k)/[k^2(kv)^2],$$
(2.5)
where it is understood that in $`\varphi _{v_\alpha ^0}`$ we have to set $`e=e_\alpha `$. For this choice of data, the constraints (1.4) are satisfied for $`t=0`$ and therefore for all $`t`$. In case $`N=1`$, the particle would travel freely, $`q_1(t)=q_1^0+v_1^0t`$, $`t0`$, and the co-moving electromagnetic fields would maintain their form (2.3).
In spirit, the bounds (2.1) and (2.2) should propagate in time and the form (2.3) of the electromagnetic fields, at least in approximation. On the other hand, for two particles with opposite charge one particular solution is the head on collision which violates the lower bound in (2.1). Considerably more delicate are solutions where some particles reach infinity in finite time, . Thus we simply require that for given constants $`C_{},C^{}>0`$ the bound
$$C_{}\epsilon ^1\underset{t[0,T\epsilon ^{3/2}]}{sup}|q_\alpha (t)q_\beta (t)|C^{}\epsilon ^1,\alpha \beta ,$$
(2.6)
holds, which implicitly defines the first time, $`T`$, at which (2.6) is violated. In fact (2.6) looks like an uncheckable assumption. But, as to be shown, the optimal $`T`$ can be computed on the basis of the approximation dynamics generated by the Lagrangian (1.1).
Under the assumption (2.6) the velocity bound propagates through the conservation of energy. We define the electrostatic energy of the charge distributions as
$$_{\mathrm{stat}}=\underset{\alpha =1}{\overset{N}{}}e_\alpha ^2\left(\frac{1}{2}d^3k|\widehat{\phi }(k)|^2k^2\right).$$
(2.7)
and compute the energy (1.6) for the given initial data. Then
$$(0):=(t=0)=\underset{\alpha =1}{\overset{N}{}}m_{\mathrm{b}\alpha }\gamma (v_\alpha ^0)+_{\mathrm{stat}}+𝒪(\epsilon )$$
with $`\gamma (v)=(1v^2)^{1/2}`$. We minimize the electromagnetic field energy $`_\mathrm{f}(t)=\frac{1}{2}d^3x[E^2(x,t)+B^2(x,t)]`$ at time $`t`$ for given $`\rho `$ and $`j`$, i.e., for given positions $`\stackrel{}{q}(t)`$ and velocities $`\stackrel{}{v}(t)`$. Using (2.6) it may be shown that
$$(t)\underset{\alpha =1}{\overset{N}{}}m_{\mathrm{b}\alpha }\gamma (v_\alpha (t))+_{\mathrm{stat}}+𝒪(\epsilon ).$$
Since by energy conservation $`(0)=(t)`$ and since the dominant contributions $`_{\mathrm{stat}}`$ cancel exactly, we thus will continue to have the bound $`|v_\alpha (t)|C\sqrt{\epsilon }`$. (We refer to Section 5.1 in Appendix A for the complete argument). Therefore
$$\underset{t[0,T\epsilon ^{3/2}]}{sup}|v_\alpha (t)|C_v\sqrt{\epsilon }$$
(2.8)
with some constant $`C_v>0`$.
As a next step we solve the inhomogeneous Maxwell equations for the fields and insert them into the Lorentz force equations. According to the retarded part of the fields, retarded positions $`q_\alpha (s)`$, $`s[0,t]`$, will show up. To control the Taylor expansion of $`q_\alpha (t)q_\alpha (s)`$ and thus of the retarded force, including the Darwin term, we will need bounds not only on positions and velocities, but also on $`\dot{v}_\alpha `$ and $`\ddot{v}_\alpha `$. Implicitly they use that the true fields remain close to the fields of the form (2.3) evaluated at current positions and velocities.
###### Lemma 2.1
Let the initial data for the Abraham model satisfy (2.1), (2.2), and (2.3). Moreover, assume
$$C_{}\epsilon ^1\underset{t[0,T\epsilon ^{3/2}]}{sup}|q_\alpha (t)q_\beta (t)|,\alpha \beta ,$$
(2.9)
for some $`T>0`$. Then there exist constants $`C^{},C_v>0`$ such that (2.6) and (2.8) hold. In particular, $`sup_{t[0,T\epsilon ^{3/2}]}|v_\alpha (t)|\overline{v}<1`$ for some $`\overline{v}`$. In addition, we find $`C>0`$ and $`\overline{e}>0`$ such that
$$\underset{t[0,T\epsilon ^{3/2}]}{sup}|\dot{v}_\alpha (t)|C\epsilon ^2\text{and}\underset{t[0,T\epsilon ^{3/2}]}{sup}|\ddot{v}_\alpha (t)|C\epsilon ^{7/2}$$
(2.10)
in case that $`|e_\alpha |\overline{e}`$, $`\alpha =1,\mathrm{},N`$. In the estimates (2.6), (2.8), and (2.10), $`C`$ and $`\overline{e}`$ do depend only on $`T`$ and the bounds for the initial data, but not on $`\epsilon `$.
The proof of this lemma is rather technical and will be given in Appendix A. Using the bounds of Lemma 2.1, we expand the Lorentz force up to an error of order $`\epsilon ^{7/2}`$, cf. Lemma 3.5, which is the order of radiation damping (the Coulomb force is order $`\epsilon ^2`$ and radiation damping a relative order $`\epsilon ^{3/2}`$ smaller). The terms up to order $`\epsilon ^3`$ then can be collected in the form of the Darwin Lagrangian (1.1). We set
$$m_\alpha =m_{\mathrm{b}\alpha }+\frac{4}{3}e_\alpha ^2m_e\text{and}m_\alpha ^{}=m_{\mathrm{b}\alpha }+\frac{16}{15}e_\alpha ^2m_e$$
with the electromagnetic mass $`m_e=\frac{1}{2}d^3k|\widehat{f}(k)|^2k^2`$ and the Darwin Lagrangian
$`_\mathrm{D}(\stackrel{}{r},\stackrel{}{u})`$ $`=`$ $`{\displaystyle \underset{\alpha =1}{\overset{N}{}}}\left({\displaystyle \frac{1}{2}}m_\alpha u_\alpha ^2+{\displaystyle \frac{\epsilon }{8}}m_\alpha ^{}u_\alpha ^4\right){\displaystyle \frac{1}{2}}{\displaystyle \underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}}{\displaystyle \frac{e_\alpha e_\beta }{4\pi |r_\alpha r_\beta |}}`$
$`+{\displaystyle \frac{\epsilon }{4}}{\displaystyle \underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}}{\displaystyle \frac{e_\alpha e_\beta }{4\pi |r_\alpha r_\beta |}}\left(u_\alpha u_\beta +|r_\alpha r_\beta |^2(u_\alpha [r_\alpha r_\beta ])(u_\beta [r_\alpha r_\beta ])\right)`$
for $`\stackrel{}{r}=(r_1,\mathrm{},r_N)`$ and $`\stackrel{}{u}=(u_1,\mathrm{},u_N)`$. The comparison dynamics is then
$$\frac{d}{dt}\left(\frac{_\mathrm{D}}{u_\alpha }\right)=\frac{_\mathrm{D}}{r_\alpha },\alpha =1,\mathrm{},N.$$
(2.11)
It conserves the energy
$$_\mathrm{D}(\stackrel{}{r},\stackrel{}{u})=\underset{\alpha =1}{\overset{N}{}}\left(\frac{1}{2}m_\alpha u_\alpha ^2+\epsilon \frac{3}{8}m_\alpha ^{}u_\alpha ^4\right)+\frac{1}{2}\underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}\frac{e_\alpha e_\beta }{4\pi |r_\alpha r_\beta |}.$$
(2.12)
Because of the Coulomb singularity, in general the solutions to (2.11) will exist only locally in time, the only exception being when all charges have the same sign, in which case energy conservation yields global existence. In the corresponding gravitational problem, for a set of positive phase space measure, mass can be transported to infinity in a finite time, . We do not know whether this can happen also for the Coulomb problem.
We set
$$q_\alpha ^0=\epsilon ^1r_\alpha ^0\text{and}v_\alpha ^0=\sqrt{\epsilon }u_\alpha ^0,\alpha =1,\mathrm{},N,$$
(2.13)
with $`r_\alpha ^0r_\beta ^0`$ for $`\alpha \beta `$. Then (2.1) and (2.2) are satisfied. During the initial time slip of order $`\epsilon ^1`$ the fields build up the forces between particles and adjust to their motion. Thus during that period the dynamics of the particles is not well approximated by the Darwin Lagrangian and we correct the initial data of the comparison dynamics to the true positions and velocities only at the end of the initial time slip. To take into account that the comparison dynamics will have no global solutions in time, in general, we define $`\tau ]0,\mathrm{}]`$ to be the first time when either $`lim_{t\tau ^{}}|r_\alpha (t)r_\beta (t)|=0`$ for some $`\alpha \beta `$ or $`lim_{t\tau ^{}}|r_\alpha (t)|=\mathrm{}`$ for some $`\alpha `$ holds for the comparison dynamics (2.11).
As our main approximation result we state
###### Theorem 2.2
Let $`T>0`$ be fixed. Define $`\tau ]0,\mathrm{}]`$ as above and fix some $`\delta _0]0,\tau [`$. For the Abraham model let the initial data be given by (2.13) and (2.3). Furthermore we require $`|e_\alpha |\overline{e}`$, with $`\overline{e}=\overline{e}(T,\mathrm{data})>0`$ from Lemma 2.1. Let $`t_0=4(R_\phi +C^{}\epsilon ^1)`$. We adjust the initial data of the comparison dynamics such that $`q_\alpha (t_0)=\epsilon ^1r_\alpha (\epsilon ^{3/2}t_0)`$ and $`v_\alpha (t_0)=\sqrt{\epsilon }u_\alpha (\epsilon ^{3/2}t_0)`$, $`\alpha =1,\mathrm{},N`$.
Then there exists a constant $`C>0`$ such that for all $`t[t_0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$ we have
$$|q_\alpha (t)\epsilon ^1r_\alpha (\epsilon ^{3/2}t)|C\sqrt{\epsilon },|v_\alpha (t)\sqrt{\epsilon }u_\alpha (\epsilon ^{3/2}t)|C\epsilon ^2,\alpha =1,\mathrm{},N.$$
(2.14)
Remarks (i) If we are satisfied with the precision from the pure Coulomb dynamics, then in (2.14) we loose one power in $`\epsilon `$. In this case, we can adjust the initial data of the comparison dynamics at time $`t=0`$, and then (2.14) holds for all $`t[0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$.
(ii) In fact the initial data need not to be adjusted exactly at $`t=t_0`$, a bound
$$|q_\alpha (t_0)\epsilon ^1r_\alpha (\epsilon ^{3/2}t_0)|\sqrt{\epsilon }\text{and}|v_\alpha (t_0)\sqrt{\epsilon }u_\alpha (\epsilon ^{3/2}t_0)|\epsilon ^2$$
would be sufficient.
## 3 Self–action and mutual interaction
In this section we expand the Lorentz force term
$$F_\alpha (t)=d^3x\rho _\alpha (xq_\alpha (t))\left[E(x,t)+v_\alpha (t)B(x,t)\right].$$
(3.1)
Since the fields $`(E,B)`$ are a solution to the inhomogeneous Maxwell’s equations, we may decompose them in the initial and the retarded fields,
$$E(x,t)=E^{(0)}(x,t)+E^{(r)}(x,t)\text{and}B(x,t)=B^{(0)}(x,t)+B^{(r)}(x,t),$$
where
$`\widehat{E}^{(0)}(k,t)`$ $`=`$ $`\mathrm{cos}|k|t\widehat{E}(k,0)i{\displaystyle \frac{\mathrm{sin}|k|t}{|k|}}k\widehat{B}(k,0),`$
$`\widehat{B}^{(0)}(k,t)`$ $`=`$ $`\mathrm{cos}|k|t\widehat{B}(k,0)+i{\displaystyle \frac{\mathrm{sin}|k|t}{|k|}}k\widehat{E}(k,0),`$
$`\widehat{E}^{(r)}(k,t)`$ $`=`$ $`{\displaystyle _0^t}𝑑s\mathrm{cos}|k|(ts)\widehat{j}(k,s)+i{\displaystyle _0^t}𝑑s{\displaystyle \frac{\mathrm{sin}|k|(ts)}{|k|}}\widehat{\rho }(k,s)k,`$
$`\widehat{B}^{(r)}(k,t)`$ $`=`$ $`i{\displaystyle _0^t}𝑑s{\displaystyle \frac{\mathrm{sin}|k|(ts)}{|k|}}k\widehat{j}(k,s),`$
cf. \[6, Section 4\], with $`j(x,t)`$ and $`\rho (x,t)`$ from (1.2). Accordingly we can rewrite $`F_\alpha (t)`$ in (3.1) as
$`F_\alpha (t)`$ $`=`$ $`{\displaystyle d^3x\rho _\alpha (xq_\alpha (t))[E^{(0)}(x,t)+v_\alpha (t)B^{(0)}(x,t)]}`$ (3.2)
$`+{\displaystyle d^3x\rho _\alpha (xq_\alpha (t))[E^{(r)}(x,t)+v_\alpha (t)B^{(r)}(x,t)]}`$
$`=`$ $`F_\alpha ^{(0)}(t)+F_\alpha ^{(r)}(t).`$
First we consider $`F_\alpha ^{(0)}(t)`$.
###### Lemma 3.1
For $`t[t_0,T\epsilon ^{3/2}]`$, with $`t_0=4(R_\phi +C^{}\epsilon ^1)`$, we have $`F_\alpha ^{(0)}(t)=0`$.
Proof : If $`S(t)`$ denotes the solution group generated by the free wave equation in $`D^{1,2}(\mathrm{IR}^3)L^2(\mathrm{IR}^3)`$, it follows from (2.3) through Fourier transform that
$`\left(\begin{array}{c}E^{(0)}(x,t)\\ \dot{E}^{(0)}(x,t)\end{array}\right)=\left[S(t)\left(\begin{array}{c}E^{(0)}(,0)\\ \dot{E}^{(0)}(,0)\end{array}\right)\right](x)={\displaystyle \underset{\beta =1}{\overset{N}{}}}e_\beta {\displaystyle _{\mathrm{}}^0}ds[S(ts)\mathrm{\Phi }_E^\beta (q_\beta ^0v_\beta ^0s)](x),`$
where $`\mathrm{\Phi }_E^\beta (x)=(\phi (x)v_\beta ^0,\phi (x))`$. The analogous formula is valid for $`B^{(0)}(x,t)`$, with $`\mathrm{\Phi }_E^\beta `$ to be replaced with $`\mathrm{\Phi }_B^\beta (x)=(0,v_\beta ^0\phi (x))`$. For fixed $`1\beta N`$ and $`x\mathrm{IR}^3`$ with $`|xq_\beta ^0|tR_\phi `$ assumption $`(C)`$ yields $`[S(ts)\mathrm{\Phi }_E^\beta (q_\beta ^0v_\beta ^0s)]_1(x)=0`$ for all $`s0`$ by means of Kirchhoff’s formula and Lemma 2.1, $`[\mathrm{}]_1`$ denoting the first component. As for $`t[t_0,T\epsilon ^{3/2}]`$ and $`|xq_\beta ^0|>tR_\phi `$ we obtain
$`|xq_\alpha (t)|`$ $``$ $`|xq_\beta ^0||q_\alpha (t)q_\beta (t)||q_\beta (t)q_\beta ^0|tR_\phi C^{}\epsilon ^1C\sqrt{\epsilon }t`$
$``$ $`t_0/2R_\phi C^{}\epsilon ^1R_\phi `$
for $`\epsilon `$ small by Lemma 2.1, the claim follows. $`\mathrm{}`$
Turning then to $`F_\alpha ^{(r)}(t)`$ in (3.2), we write this term in Fourier transformed form and use (1.2) to obtain
$$F_\alpha ^{(r)}(t)=e_\alpha ^2F_{\alpha \alpha }^{(r)}(t)+\underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}e_\alpha e_\beta F_{\alpha \beta }^{(r)}(t),$$
(3.4)
with
$`F_{\alpha \beta }^{(r)}(t)`$ $`=`$ $`{\displaystyle _0^t}ds{\displaystyle }dk|\widehat{\phi }(k)|^2e^{ik[q_\alpha (t)q_\beta (s)]}\{\mathrm{cos}|k|(ts)v_\beta (s)+i{\displaystyle \frac{\mathrm{sin}|k|(ts)}{|k|}}k`$ (3.5)
$`i{\displaystyle \frac{\mathrm{sin}|k|(ts)}{|k|}}v_\alpha (t)(kv_\beta (s))\},`$
$`\alpha ,\beta =1,\mathrm{},N`$. The term $`F_{\alpha \alpha }^{(r)}(t)`$ accounts for the self-force, whereas $`F_{\alpha \beta }^{(r)}(t)`$ for $`\beta \alpha `$ represents the mutual interaction force between particle $`\alpha `$ and particle $`\beta `$. These both contributions are dealt with separately in the following two subsections.
Before going on to this, we state an auxiliary result.
###### Lemma 3.2
Let $`1\alpha ,\beta N`$, $`\alpha \beta `$. For $`t[t_0,T\epsilon ^{3/2}]`$ we have
* $`{\displaystyle _0^t}𝑑s{\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik[q_\alpha (t)q_\beta (s)]}\mathrm{cos}|k|(ts)v_\beta (s)}`$
$`={\displaystyle _0^{\mathrm{}}}𝑑\tau {\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\mathrm{cos}|k|\tau \left\{v_\beta i\tau (kv_\beta )v_\beta \tau \dot{v}_\beta \right\}}+𝒪(\epsilon ^{7/2})`$,
* $`i{\displaystyle _0^t}𝑑s{\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik[q_\alpha (t)q_\beta (s)]}\frac{\mathrm{sin}|k|(ts)}{|k|}k}`$
$`=i{\displaystyle _0^{\mathrm{}}}𝑑\tau {\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\frac{\mathrm{sin}|k|\tau }{|k|}k\left\{1ik\left[\tau v_\beta \frac{1}{2}\tau ^2\dot{v}_\beta \right]\frac{1}{2}\tau ^2(kv_\beta )^2\right\}}+𝒪(\epsilon ^{7/2})`$,
* $`(i){\displaystyle _0^t}𝑑s{\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik[q_\alpha (t)q_\beta (s)]}\frac{\mathrm{sin}|k|(ts)}{|k|}v_\alpha (t)}(kv_\beta (s))`$
$`=(i){\displaystyle _0^{\mathrm{}}}𝑑\tau {\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\frac{\mathrm{sin}|k|\tau }{|k|}v_\alpha }(kv_\beta )+𝒪(\epsilon ^{7/2})`$.
Here $`v_\alpha =v_\alpha (t)`$, etc., and $`\xi _{\alpha \beta }=q_\alpha (t)q_\beta (t)`$.
The proof is somewhat tedious and given in Appendix B.
### 3.1 Self–action
For $`t[t_0,T\epsilon ^{3/2}]`$ we have
$`F_{\alpha \alpha }^{(r)}(t)`$ $`=`$ $`{\displaystyle _0^{\mathrm{}}}d\tau {\displaystyle }dk|\widehat{\phi }(k)|^2e^{i(kv_\alpha )\tau }(1+{\displaystyle \frac{i}{2}}(k\dot{v}_\alpha )\tau ^2)\{\mathrm{cos}|k|\tau [v_\alpha \dot{v}_\alpha \tau ]+i{\displaystyle \frac{\mathrm{sin}|k|\tau }{|k|}}k`$ (3.6)
$`i{\displaystyle \frac{\mathrm{sin}|k|\tau }{|k|}}v_\alpha (k[v_\alpha \dot{v}_\alpha \tau ])\}+𝒪(\epsilon ^{7/2}).`$
The rigorous proof of this relation is omitted since it is very similar to the proof of Lemma 3.2 given in Appendix B. It once more relies on the fact that we may Taylor expand
$$q_\alpha (s)q_\alpha v_\alpha \tau +\frac{1}{2}\dot{v}_\alpha \tau ^2+𝒪(\epsilon ^{7/2}),v_\alpha (s)v_\alpha \dot{v}_\alpha \tau +𝒪(\epsilon ^{7/2})$$
by Lemma 2.1, with $`q_\alpha =q_\alpha (t)`$ etc. and $`\tau =ts`$, whence
$$e^{ik[q_\alpha (t)q_\alpha (s)]}e^{i(kv_\alpha )\tau }\left(1+\frac{i}{2}(k\dot{v}_\alpha )\tau ^2\right)+𝒪(\epsilon ^{7/2}).$$
Introducing
$$I_p=_0^{\overline{t}}𝑑\tau \frac{\mathrm{sin}(|k|\tau )}{|k|}e^{i(kv_\alpha )\tau }\tau ^p,J_p=_0^{\overline{t}}𝑑\tau \mathrm{cos}(|k|\tau )e^{i(kv_\alpha )\tau }\tau ^p,p\mathrm{IN}_0,$$
Equ. (3.6) may be rewritten as
$`F_{\alpha \alpha }^{(r)}(t)`$ $`=`$ $`\underset{\overline{t}\mathrm{}}{lim}({\displaystyle }dk|\widehat{\phi }(k)|^2\{v_\alpha J_0\dot{v}_\alpha J_1+{\displaystyle \frac{i}{2}}(k\dot{v}_\alpha )v_\alpha J_2\}`$ (3.7)
$`+{\displaystyle }dk|\widehat{\phi }(k)|^2\{i[(1v_\alpha ^2)k+(kv_\alpha )v_\alpha ]I_0+i[(v_\alpha \dot{v}_\alpha )k(kv_\alpha )\dot{v}_\alpha ]I_1`$
$`{\displaystyle \frac{1}{2}}(k\dot{v}_\alpha )[(1v_\alpha ^2)k+(kv_\alpha )v_\alpha ]I_2\})+𝒪(\epsilon ^{7/2}),`$
since $`\dot{v}_\alpha ^2=𝒪(\epsilon ^4)`$. Denote the term containing the $`J_p`$ by $`𝒥`$ and the one containing the $`I_p`$ by $``$. To evaluate the limits $`\overline{t}\mathrm{}`$, we can rely on the results from \[6, Section 4\]. We first recall that
$$𝑑k|\widehat{\phi }(k)|^2J_00,𝑑k|\widehat{\phi }(k)|^2J_12m_\mathrm{e}\gamma _\alpha ^2\text{as}\overline{t}\mathrm{},$$
with $`\gamma _\alpha =(1v_\alpha ^2)^{1/2}`$ and $`m_\mathrm{e}=\frac{1}{2}𝑑k|\widehat{\phi }(k)|^2k^2`$. Moreover, $`_vJ_1=ikJ_2`$, and therefore
$`𝒥`$ $``$ $`(2m_\mathrm{e}\gamma _\alpha ^2)\dot{v}_\alpha +{\displaystyle \frac{1}{2}}\dot{v}_\alpha _v(2m_\mathrm{e}\gamma _\alpha ^2)v_\alpha =2m_\mathrm{e}\gamma _\alpha ^2\left(\dot{v}_\alpha +\gamma _\alpha ^2(v_\alpha \dot{v}_\alpha )v_\alpha \right)`$ (3.8)
$`=`$ $`2m_\mathrm{e}\left((1+v_\alpha ^2)\dot{v}_\alpha +(v_\alpha \dot{v}_\alpha )v_\alpha \right)+𝒪(\epsilon ^4)`$
as $`\overline{t}\mathrm{}`$, the latter equality according to the expansion $`\gamma _\alpha ^2=1+v_\alpha ^2+𝒪(v_\alpha ^4)=1+v_\alpha ^2+𝒪(\epsilon ^2)`$ and $`\gamma _\alpha ^4=1+𝒪(\epsilon )`$.
What concerns $``$, we know from that $`𝑑k|\widehat{\phi }(k)|^2kI_00`$,
$$𝑑k|\widehat{\phi }(k)|^2I_02m_\mathrm{e}|v_\alpha |^1\mathrm{arth}|v_\alpha |,𝑑k|\widehat{\phi }(k)|^2(k\dot{v}_\alpha )kI_22m_\mathrm{e}\mu (v_\alpha )\dot{v}_\alpha $$
as $`\overline{t}\mathrm{}`$, where
$$\mu (v)z=\left(\frac{\gamma ^2}{v^2}|v|^3\mathrm{arth}|v|\right)z+\left(\frac{\gamma ^4}{v^4}(5v^23)+3|v|^5\mathrm{arth}|v|\right)(vz)v$$
for $`|v|<1`$ and $`z\mathrm{IR}^3`$. Consequently, since $`s^1\mathrm{arth}(s)=1+s^2/3+s^4/5+𝒪(s^6)`$ for $`s`$ close to zero, it thus follows after some calculation that
$``$ $``$ $`(v_\alpha \dot{v}_\alpha )_v\left(2m_\mathrm{e}|v_\alpha |^1\mathrm{arth}|v_\alpha |\right)+\dot{v}_\alpha v_\alpha _v\left(2m_\mathrm{e}|v_\alpha |^1\mathrm{arth}|v_\alpha |\right)`$ (3.9)
$`{\displaystyle \frac{1}{2}}(1v_\alpha ^2)\left(2m_\mathrm{e}\mu (v_\alpha )\dot{v}_\alpha \right){\displaystyle \frac{1}{2}}v_\alpha v_\alpha \left(2m_\mathrm{e}\mu (v_\alpha )\dot{v}_\alpha \right)`$
$`=`$ $`\left({\displaystyle \frac{2}{3}}+{\displaystyle \frac{22}{15}}v_\alpha ^2\right)m_\mathrm{e}\dot{v}_\alpha +{\displaystyle \frac{14}{15}}m_\mathrm{e}(v_\alpha \dot{v}_\alpha )v_\alpha +𝒪(\epsilon ^4).`$
Summarizing (3.7), (3.8), and (3.9), we arrive at
###### Lemma 3.3
For $`t[t_0,T\epsilon ^{3/2}]`$ we have
$$F_{\alpha \alpha }^{(r)}(t)=\left(\frac{4}{3}+\frac{8}{15}v_\alpha ^2\right)m_\mathrm{e}\dot{v}_\alpha \frac{16}{15}m_\mathrm{e}(v_\alpha \dot{v}_\alpha )v_\alpha +𝒪(\epsilon ^{7/2}).$$
### 3.2 Mutual interaction
In this section we expand $`F_{\alpha \beta }^{(r)}(t)`$ from (3.5) with $`\beta \alpha `$. For $`p\mathrm{IN}_0`$ we have that
$$A_p:=_0^{\mathrm{}}𝑑\tau 𝑑k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\frac{\mathrm{sin}|k|\tau }{|k|}\tau ^p=(4\pi )^1𝑑x𝑑y\phi (x)\phi (y)|\xi _{\alpha \beta }+xy|^{p1}$$
and
$`B_p`$ $`:=`$ $`{\displaystyle _0^{\mathrm{}}}𝑑\tau {\displaystyle 𝑑k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\mathrm{cos}(|k|\tau )\tau ^p}`$
$`=`$ $`(p)(4\pi )^1{\displaystyle 𝑑x𝑑y\phi (x)\phi (y)|\xi _{\alpha \beta }+xy|^{p2}}=(p)A_{p1},`$
as may be seen through Fourier transform. We hence obtain from Lemma 3.2 that for $`\beta \alpha `$ and $`t[t_0,T\epsilon ^{3/2}]`$
$`F_{\alpha \beta }^{(r)}(t)`$ $`=`$ $`v_\beta (v_\beta _\xi )B_1+\dot{v}_\beta B_1_\xi A_0+{\displaystyle \frac{1}{2}}(\dot{v}_\beta _\xi )_\xi A_2{\displaystyle \frac{1}{2}}(v_\beta _\xi )^2_\xi A_2`$ (3.10)
$`+(v_\alpha v_\beta )_\xi A_0v_\beta (v_\alpha _\xi )A_0+𝒪(\epsilon ^{7/2}),`$
taking also into account that $`A_1=(4\pi )^1`$, thus $`_\xi A_1=0`$. As a consequence of $`|\xi _{\alpha \beta }|=𝒪(\epsilon ^1)`$, cf. Lemma 2.1, of assumption $`(C)`$, and of Lemma 2.1, it follows that in (3.10) we have $`_\xi A_0=𝒪(\epsilon ^2)`$, while all other terms are $`𝒪(\epsilon ^3)`$. Since e.g.
$$\left|(v_\alpha v_\beta )_\xi A_0(v_\alpha v_\beta )\left(\frac{\xi _{\alpha \beta }}{4\pi |\xi _{\alpha \beta }|^3}\right)\right|C\epsilon ^4,$$
with an obvious similar estimate for the other terms besides $`_\xi A_0`$, we find from (3.10) and after some calculation that for $`\beta \alpha `$ and $`t[t_0,T\epsilon ^{3/2}]`$
$`F_{\alpha \beta }^{(r)}(t)`$ $`=`$ $`v_\beta (v_\beta _\xi )\left({\displaystyle \frac{1}{4\pi |\xi _{\alpha \beta }|}}\right)\dot{v}_\beta \left({\displaystyle \frac{1}{4\pi |\xi _{\alpha \beta }|}}\right)_\xi A_0+{\displaystyle \frac{1}{2}}(\dot{v}_\beta _\xi )_\xi \left({\displaystyle \frac{|\xi _{\alpha \beta }|}{4\pi }}\right)`$
$`{\displaystyle \frac{1}{2}}(v_\beta _\xi )^2_\xi \left({\displaystyle \frac{|\xi _{\alpha \beta }|}{4\pi }}\right)+(v_\alpha v_\beta )_\xi \left({\displaystyle \frac{1}{4\pi |\xi _{\alpha \beta }|}}\right)v_\beta (v_\alpha _\xi )\left({\displaystyle \frac{1}{4\pi |\xi _{\alpha \beta }|}}\right)+𝒪(\epsilon ^{7/2})`$
$`=`$ $`_\xi A_0{\displaystyle \frac{1}{8\pi |\xi _{\alpha \beta }|}}\dot{v}_\beta {\displaystyle \frac{(\dot{v}_\beta \xi _{\alpha \beta })}{8\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{v_\beta ^2}{8\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }{\displaystyle \frac{3(v_\beta \xi _{\alpha \beta })^2}{8\pi |\xi _{\alpha \beta }|^5}}\xi _{\alpha \beta }`$
$`{\displaystyle \frac{(v_\alpha v_\beta )}{4\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{(v_\alpha \xi _{\alpha \beta })}{4\pi |\xi _{\alpha \beta }|^3}}v_\beta +𝒪(\epsilon ^{7/2}).`$
Finally, to deal with the lowest-order term we observe that with $`\stackrel{}{n}=\xi _{\alpha \beta }/|\xi _{\alpha \beta }|`$
$$\left|_\xi A_0+\frac{\xi _{\alpha \beta }}{4\pi |\xi _{\alpha \beta }|^3}\right|=\frac{1}{4|\xi _{\alpha \beta }|^2}\left|𝑑x𝑑y\phi (x)\phi (y)\left(\frac{\stackrel{}{n}+\frac{xy}{|\xi _{\alpha \beta }|}}{\left|\stackrel{}{n}+\frac{xy}{|\xi _{\alpha \beta }|}\right|^3}\stackrel{}{n}\right)\right|.$$
(3.11)
Defining $`R=(xy)/|\xi _{\alpha \beta }|=𝒪(\epsilon )`$ for $`|x|,|y|R_\phi `$, we can expand $`\psi (R)=(\stackrel{}{n}+R)/|\stackrel{}{n}+R|`$ to obtain that $`\psi (R)=\stackrel{}{n}+R3(\stackrel{}{n}R)\stackrel{}{n}+𝒪(\epsilon ^2)`$. As $`𝑑x𝑑y\phi (x)\phi (y)(xy)=0`$, we hence conclude that the right-hand side of (3.11) is $`𝒪(\epsilon ^4)`$. Thus we can summarize our estimates on the mutual interaction force as follows.
###### Lemma 3.4
For $`\beta \alpha `$ and $`t[t_0,T\epsilon ^{3/2}]`$ we have
$`F_{\alpha \beta }^{(r)}(t)`$ $`=`$ $`{\displaystyle \frac{\xi _{\alpha \beta }}{4\pi |\xi _{\alpha \beta }|^3}}{\displaystyle \frac{1}{8\pi |\xi _{\alpha \beta }|}}\dot{v}_\beta {\displaystyle \frac{(\dot{v}_\beta \xi _{\alpha \beta })}{8\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{v_\beta ^2}{8\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }{\displaystyle \frac{3(v_\beta \xi _{\alpha \beta })^2}{8\pi |\xi _{\alpha \beta }|^5}}\xi _{\alpha \beta }`$
$`{\displaystyle \frac{(v_\alpha v_\beta )}{4\pi |\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{(v_\alpha \xi _{\alpha \beta })}{4\pi |\xi _{\alpha \beta }|^3}}v_\beta +𝒪(\epsilon ^{7/2}).`$
### 3.3 Summary of the estimates
By (3.1), (3.2), and Lemma 3.1 we find $`F_\alpha (t)=F_\alpha ^{(r)}(t)`$ for $`t[t_0,T\epsilon ^{3/2}]`$. According to (3.4) and Lemmas 3.3 and 3.4 we hence have obtained the following expansion of the Lorentz force in (3.1). For $`t[t_0,T\epsilon ^{3/2}]`$ we have
$`F_\alpha (t)`$ $`=`$ $`\left({\displaystyle \frac{4}{3}}+{\displaystyle \frac{8}{15}}v_\alpha ^2\right)m_\mathrm{e}\dot{v}_\alpha {\displaystyle \frac{16}{15}}m_\mathrm{e}(v_\alpha \dot{v}_\alpha )v_\alpha +G_\alpha (\stackrel{}{q},\stackrel{}{v},\stackrel{}{\dot{v}})+𝒪(\epsilon ^{7/2}),`$
$`G_\alpha (\stackrel{}{q},\stackrel{}{v},\stackrel{}{\dot{v}})`$ $`=`$ $`{\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}}{\displaystyle \frac{e_\alpha e_\beta }{4\pi }}({\displaystyle \frac{\xi _{\alpha \beta }}{|\xi _{\alpha \beta }|^3}}{\displaystyle \frac{1}{2|\xi _{\alpha \beta }|}}\dot{v}_\beta {\displaystyle \frac{(\dot{v}_\beta \xi _{\alpha \beta })}{2|\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{v_\beta ^2}{2|\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }{\displaystyle \frac{3(v_\beta \xi _{\alpha \beta })^2}{2|\xi _{\alpha \beta }|^5}}\xi _{\alpha \beta }`$ (3.12)
$`{\displaystyle \frac{(v_\alpha v_\beta )}{|\xi _{\alpha \beta }|^3}}\xi _{\alpha \beta }+{\displaystyle \frac{(v_\alpha \xi _{\alpha \beta })}{|\xi _{\alpha \beta }|^3}}v_\beta ),`$
where $`t_0=4(R_\phi +C^{}\epsilon ^1)`$, $`\xi _{\alpha \beta }=q_\alpha (t)q_\beta (t)`$, $`v_\alpha =v_\alpha (t)`$, and $`v_\beta =v_\beta (t)`$. Due to the Lorentz equation $`\frac{d}{dt}(m_{\mathrm{b}\alpha }\gamma _\alpha v_\alpha )=F_\alpha (t)`$, cf. (1.5), we finally obtain the following lemma by calculating the right-hand side and expanding $`\gamma _\alpha `$.
###### Lemma 3.5
For $`t[t_0,T\epsilon ^{3/2}]`$ we have
$$M_\alpha (v_\alpha )\dot{v}_\alpha =G_\alpha (\stackrel{}{q},\stackrel{}{v},\stackrel{}{\dot{v}})+𝒪(\epsilon ^{7/2}),1\alpha N,$$
with $`G_\alpha `$ from (3.3) and $`M_\alpha (v)`$ the $`(3\times 3)`$-matrix $`M_\alpha (v)(z)=(m_\alpha +\frac{1}{2}m_\alpha ^{}v^2)z+m_\alpha ^{}(vz)v`$ for $`v,z\mathrm{IR}^3`$.
## 4 Proof of Theorem 2.2
We need to compare a solution $`(q_\alpha (t),v_\alpha (t))`$ of (1.2)–(1.5) with data (2.13) to $`(\stackrel{~}{r}_\alpha (t),\stackrel{~}{u}_\alpha (t))`$, where we let
$$\stackrel{~}{r}_\alpha (t)=\epsilon ^1r_\alpha (\epsilon ^{3/2}t),\stackrel{~}{u}_\alpha (t)=\sqrt{\epsilon }u_\alpha (\epsilon ^{3/2}t),$$
(4.1)
and where the $`(r_\alpha (t),u_\alpha (t))`$ are the solution to the system induced by (2.11) with data $`(r_\alpha ^0,u_\alpha ^0)`$.
A somewhat lengthy but elementary calculation shows that $`(\stackrel{~}{r}_\alpha (t),\stackrel{~}{u}_\alpha (t))`$ satisfy
$$M_\alpha (\stackrel{~}{u}_\alpha )\dot{\stackrel{~}{u}}_\alpha =G_\alpha (\stackrel{}{\stackrel{~}{r}},\stackrel{}{\stackrel{~}{u}},\stackrel{}{\dot{\stackrel{~}{u}}}),1\alpha N,$$
(4.2)
cf. Lemma 3.5 for the notation. Recalling that $`\tau ]0,\mathrm{}]`$ was defined to be the first time when either $`lim_{t\tau ^{}}|r_\alpha (t)r_\beta (t)|=0`$ for some $`\alpha \beta `$ or $`lim_{t\tau ^{}}|r_\alpha (t)|=\mathrm{}`$ for some $`\alpha `$ holds, we find that (4.2) is valid for $`t[0,(\tau \delta _0)\epsilon ^{3/2}]`$, for any $`\delta _0]0,\tau [`$ which we consider to be fixed throughout. This leads to some useful estimates on the effective dynamics.
###### Lemma 4.1
For suitable constants $`C_0,C^0,C>0`$ (depending on $`\tau `$, $`\delta _0`$, and the data) we have
$$C_0\epsilon ^1\underset{t[0,(\tau \delta _0)\epsilon ^{3/2}]}{sup}|\stackrel{~}{r}_\alpha (t)\stackrel{~}{r}_\beta (t)|C^0\epsilon ^1,\alpha \beta ,$$
(4.3)
and
$$\underset{t[0,(\tau \delta _0)\epsilon ^{3/2}]}{sup}|\stackrel{~}{u}_\alpha (t)|C\sqrt{\epsilon }.$$
(4.4)
Proof : The bounds in (4.3) follow from (4.1) and the fact that $`|r_\alpha (t)r_\beta (t)|\delta _1`$ and $`|r_\alpha (t)|C`$ on $`[0,\tau \delta _0]`$ for some $`\delta _1>0`$, $`C>0`$, by definition of $`\tau `$. Concerning (4.4), by conservation of the energy $`_\mathrm{D}`$ from (2.12) we obtain $`C_\mathrm{D}(\stackrel{}{r}(0),\stackrel{}{u}(0))=_\mathrm{D}(\stackrel{}{r}(t),\stackrel{}{u}(t))\frac{1}{2}m_\alpha u_\alpha ^2(t)`$ as long as the solution exists, in particular for $`t[0,\tau \delta _0]`$. $`\mathrm{}`$
To simplify the presentation, we henceforth omit the tilde and write $`(r,u)`$ instead of $`(\stackrel{~}{r},\stackrel{~}{u})`$ to denote the rescaled solution. Utilizing the bounds from Lemma 2.1 and from (4.3), (4.4), it may be seen after some calculation that
$$\left|G_\alpha (\stackrel{}{q},\stackrel{}{v},\stackrel{}{\dot{v}})(t)G_\alpha (\stackrel{}{r},\stackrel{}{u},\stackrel{}{\dot{u}})(t)\right|C\underset{\beta =1}{\overset{N}{}}\left(\epsilon ^3|q_\beta (t)r_\beta (t)|+\epsilon ^{5/2}|v_\beta (t)u_\beta (t)|+\epsilon |\dot{v}_\beta (t)\dot{u}_\beta (t)|\right)$$
(4.5)
for $`1\alpha N`$ and $`t[0,T\epsilon ^{3/2}][0,(\tau \delta _0)\epsilon ^{3/2}]=[0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$. Note that the term $`\epsilon ^3|q_\beta r_\beta |`$ appears through comparison of $`\xi _{\alpha \beta }/|\xi _{\alpha \beta }|^3`$ to $`r_{\alpha \beta }/|r_{\alpha \beta }|^3`$, cf. the form of $`G_\alpha `$ in (3.3).
Next, a general $`(3\times 3)`$-matrix $`M(v)=a(v)\mathrm{id}+b(vv)`$ has the inverse
$$M(v)^1=a(v)^1\mathrm{id}+\frac{b}{a(v)[a(v)+bv^2]}(vv).$$
This remark shows $`|M_\alpha (v_\alpha )^1|=𝒪(1)`$ and $`|M_\alpha (v_\alpha )^1M_\alpha (u_\alpha )^1|C\sqrt{\epsilon }|v_\alpha u_\alpha |`$ for $`t[0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$. Since $`|G_\alpha (\stackrel{}{q},\stackrel{}{v},\stackrel{}{\dot{v}})|=𝒪(\epsilon ^2)`$ it follows from Lemma 3.5, (4.2), and (4.5) that
$$|\dot{v}_\alpha (t)\dot{u}_\alpha (t)|C\underset{\beta =1}{\overset{N}{}}\left(\epsilon ^3|q_\beta (t)r_\beta (t)|+\epsilon ^{5/2}|v_\beta (t)u_\beta (t)|+\epsilon |\dot{v}_\beta (t)\dot{u}_\beta (t)|\right)+𝒪(\epsilon ^{7/2})$$
for $`1\alpha N`$ and $`t[t_0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$. Summation over $`\alpha `$ and choosing $`\epsilon >0`$ sufficiently small this results in
$$\underset{\alpha =1}{\overset{N}{}}|\dot{v}_\alpha (t)\dot{u}_\alpha (t)|C\underset{\alpha =1}{\overset{N}{}}\left(\epsilon ^3|q_\alpha (t)r_\alpha (t)|+\epsilon ^{5/2}|v_\alpha (t)u_\alpha (t)|\right)+𝒪(\epsilon ^{7/2})$$
(4.6)
for $`t[t_0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$. To use this basic estimate, we write $`d_\alpha (t)=q_\alpha (t)r_\alpha (t)`$ as
$$d_\alpha (t)=d_\alpha (t_0)+(tt_0)\dot{d}_\alpha (t_0)+_{t_0}^t(ts)\ddot{d}_\alpha (s)𝑑s,\dot{d}_\alpha (t)=\dot{d}_\alpha (t_0)+_{t_0}^t\ddot{d}_\alpha (s)𝑑s.$$
We then obtain for $`t[t_0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}]`$ from (4.6) that
$`D(t)`$ $``$ $`D(t_0)+(tt_0)\overline{D}(t_0)+C\epsilon ^3{\displaystyle _{t_0}^t}(ts)D(s)𝑑s`$ (4.7)
$`+C\epsilon ^{5/2}{\displaystyle _{t_0}^t}(ts)\overline{D}(s)𝑑s+C\sqrt{\epsilon },`$
$`\overline{D}(t)`$ $``$ $`\overline{D}(t_0)+C\epsilon ^3{\displaystyle _{t_0}^t}D(s)𝑑s+C\epsilon ^{5/2}{\displaystyle _{t_0}^t}\overline{D}(s)𝑑s+C\epsilon ^2,`$ (4.8)
where
$$D(t)=\underset{1\alpha N}{\mathrm{max}}\underset{s[t_0,t]}{\mathrm{max}}|d_\alpha (s)|\text{and}\overline{D}(t)=\underset{1\alpha N}{\mathrm{max}}\underset{s[t_0,t]}{\mathrm{max}}|\dot{d}_\alpha (s)|.$$
Application of Gronwall’s lemma to (4.8) yields
$$\overline{D}(t)C\left(\overline{D}(t_0)+\epsilon ^2+\epsilon ^3_{t_0}^tD(s)𝑑s\right),$$
(4.9)
and utilizing this in (4.7) implies
$$D(t)D(t_0)+(tt_0)\overline{D}(t_0)+C\sqrt{\epsilon }+C\epsilon ^{1/2}(\overline{D}(t_0)+\epsilon ^2)+C\epsilon ^3_{t_0}^t(ts)D(s)𝑑s.$$
Finally, $`(ts)C\epsilon ^{3/2}`$ yields upon a further application of Gronwall’s lemma that
$$D(t)C\left(D(t_0)+\epsilon ^{3/2}\overline{D}(t_0)+\sqrt{\epsilon }\right),t[t_0,\mathrm{min}\{\tau \delta _0,T\}\epsilon ^{3/2}].$$
(4.10)
By assumption $`D(t_0)=0=\overline{D}(t_0)`$. Therefore (4.10) and (4.9) imply (2.14). This completes the proof of Theorem 2.2. $`\mathrm{}`$
## 5 Appendix A: Proof of Lemma 2.1
This appendix concerns the proof of Lemma 2.1. We split the proof into three subsections.
### 5.1 Bounding the particle distances and the velocities
We intend to use energy conservation to show (2.8), and for that reason we calculate with (2.3) the field energy
$`_\mathrm{F}(0)`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle d^3x[E^2(x,0)+B^2(x,0)]}`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha =1}{\overset{N}{}}}{\displaystyle }d^3x[E_{v_\alpha ^0}^2(xq_\alpha ^0)+B_{v_\alpha ^0}^2(xq_\alpha ^0)]+{\displaystyle \frac{1}{2}}{\displaystyle \underset{\stackrel{\alpha ,\beta =1}{\alpha \beta }}{\overset{N}{}}}{\displaystyle }d^3x[E_{v_\alpha ^0}(xq_\alpha ^0)E_{v_\beta ^0}(xq_\beta ^0)`$
$`+B_{v_\alpha ^0}(xq_\alpha ^0)B_{v_\beta ^0}(xq_\beta ^0)].`$
According to (2.4) and \[6, Section 2\] the first term equals
$$_\mathrm{F}^{(1)}(0)=\underset{\alpha =1}{\overset{N}{}}e_\alpha ^2\left(\frac{1}{2}d^3k|\widehat{\phi }(k)|^2k^2\right)\left[\frac{1}{|v_\alpha ^0|}\mathrm{log}\frac{1+|v_\alpha ^0|}{1|v_\alpha ^0|}1\right].$$
Denoting the term in $`[\mathrm{}]`$ as $`\psi (|v_\alpha ^0|)`$, $`\psi (r)`$ is odd, and hence Taylor expansion implies $`\psi (r)=1+𝒪(r^2)`$ for $`r`$ small. Therefore (2.2) yields
$$_\mathrm{F}^{(1)}(0)=_{\mathrm{Coul}}+𝒪(\epsilon ),$$
with $`_{\mathrm{Coul}}`$ from (2.7). To deal with the contributions for $`\alpha \beta `$ in the second term, we obtain by passing to Fourier transformed form and observing (2.2) that e.g.
$$d^3xE_{v_\alpha ^0}(xq_\alpha ^0)E_{v_\beta ^0}(xq_\beta ^0)=e_\alpha e_\beta d^3k|\widehat{\phi }(k)|^2k^2e^{ik(q_\alpha ^0q_\beta ^0)}+𝒪(\epsilon )=𝒪(\epsilon ),$$
the latter with (2.1) and by passing to polar coordinates. Thus we have shown
$$_\mathrm{F}(0)=_{\mathrm{Coul}}+𝒪(\epsilon ).$$
(5.1)
Next we will investigate the field energy at time $`t>0`$. We claim that
$$_\mathrm{F}(t)=\frac{1}{2}d^3x[E^2(x,t)+B^2(x,t)]\frac{1}{2}d^3xE^2(x,t)\frac{1}{2}(\rho (,t),\mathrm{\Delta }^1\rho (,t))_{L^2(\mathrm{IR}^3)}.$$
(5.2)
The easiest way to see this is to introduce potentials $`A`$ and $`\varphi `$,
$$B(x,t)=A(x,t),E(x,t)=\varphi (x,t)F(x,t),\text{with}F(x,t)=\frac{A}{t}(x,t),$$
for the electromagnetic field. Then $`\rho =E=\mathrm{\Delta }\varphi F`$, and the estimate in (5.2) follows by passing to Fourier transformed form. On the other hand, substituting $`\rho `$ from (1.2) into
$$\frac{1}{2}(\rho (,t),\mathrm{\Delta }^1\rho (,t))_{L^2(\mathrm{IR}^3)}=\frac{1}{2}d^3k|\widehat{\rho }(k,t)|^2k^2,$$
by assumption (2.9) we can argue exactly as before to show that the terms with $`\alpha \beta `$ are $`𝒪(\epsilon )`$, and thus
$$_\mathrm{F}(t)\frac{1}{2}\underset{\alpha =1}{\overset{N}{}}d^3k|\widehat{\rho }_\alpha (k)|^2k^2+𝒪(\epsilon )=_{\mathrm{Coul}}+𝒪(\epsilon )$$
(5.3)
for $`t[0,T\epsilon ^{3/2}]`$. Consequently for $`t[0,T\epsilon ^{3/2}]`$ by energy conservation, cf. (1.6), by (5.1) and (5.3)
$`{\displaystyle \underset{\alpha =1}{\overset{N}{}}}m_{\mathrm{b}\alpha }\gamma (v_\alpha ^0)+_{\mathrm{Coul}}+𝒪(\epsilon )`$ $`=`$ $`{\displaystyle \underset{\alpha =1}{\overset{N}{}}}m_{\mathrm{b}\alpha }\gamma (v_\alpha ^0)+_F(0)={\displaystyle \underset{\alpha =1}{\overset{N}{}}}m_{\mathrm{b}\alpha }\gamma (v_\alpha (t))+_F(t)`$
$``$ $`{\displaystyle \underset{\alpha =1}{\overset{N}{}}}m_{\mathrm{b}\alpha }\gamma (v_\alpha (t))+_{\mathrm{Coul}}+𝒪(\epsilon ),`$
with $`\gamma (v)=(1v^2)^{1/2}`$. Thus
$$\underset{\alpha =1}{\overset{N}{}}m_{\mathrm{b}\alpha }\gamma (v_\alpha ^0)+C\epsilon \underset{\alpha =1}{\overset{N}{}}m_{\mathrm{b}\alpha }\gamma (v_\alpha (t)),t[0,T\epsilon ^{3/2}]$$
(5.4)
with some constant $`C`$ depending on $`C_1,C_3,C_{},T`$. This estimate now allows to prove (2.8). Define
$$I_+=\{\alpha \{1,\mathrm{},N\}:\gamma (v_\alpha (t))\gamma (v_\alpha ^0)\}\text{and}I_{}=\{\alpha \{1,\mathrm{},N\}:\gamma (v_\alpha (t))>\gamma (v_\alpha ^0)\}.$$
For $`\alpha I_+`$ we have $`|v_\alpha (t)||v_\alpha ^0|C_3\sqrt{\epsilon }`$ by (2.2). Thus for $`\epsilon `$ so small that $`C_3^2\epsilon 1/2`$, $`\gamma (v_\alpha ^0)\gamma (v_\alpha (t))\sqrt{2}|(v_\alpha ^0)^2(v_\alpha (t))^2|C\epsilon `$. Therefore by (5.4)
$$C\epsilon \underset{\alpha I_{}}{}m_{\mathrm{b}\alpha }\left(\gamma (v_\alpha (t))\gamma (v_\alpha ^0)\right).$$
Since $`m_{\mathrm{b}\alpha }>0`$ we deduce that
$$\gamma (v_\alpha (t))\gamma (v_\alpha ^0)+C\epsilon ,\alpha I_{},$$
and according to $`|v_\alpha ^0|C_3\sqrt{\epsilon }`$ it then follows that $`|v_\alpha (t)|C\sqrt{\epsilon }`$ also for $`\alpha I_{}`$. This concludes the proof of (2.8).
Using (2.1) and (2.8) it is finally easy to derive the upper bound in (2.6), since for $`t[0,T\epsilon ^{3/2}]`$ we have
$$|q_\alpha (t)q_\beta (t)||q_\alpha ^0q_\beta ^0|+|q_\alpha (t)q_\alpha ^0|+|q_\beta (t)q_\beta ^0|C_2\epsilon ^1+2C_vT\sqrt{\epsilon }\epsilon ^{3/2}=C^{}\epsilon ^1,$$
with $`C^{}=C_2+2C_vT`$. We remark that for the estimates in this section the smallness of the $`e_\alpha `$ was not needed.
### 5.2 Bounding $`|\dot{v}_\alpha (t)|`$
Since
$$\frac{d}{dt}\left(m_{\mathrm{b}\alpha }\gamma _\alpha v_\alpha (t)\right)=m_{0\alpha }(v_\alpha (t))\dot{v}_\alpha (t),$$
with the $`(3\times 3)`$-matrices $`m_{0\alpha }(v_\alpha )`$ given through $`m_{0\alpha }(v_\alpha )(z)=m_{\mathrm{b}\alpha }(\gamma _\alpha z+\gamma _\alpha ^3(v_\alpha z)v_\alpha )`$, $`z\mathrm{IR}^3`$, we obtain from (1.5) that for $`\alpha =1,\mathrm{},N`$
$`\dot{v}_\alpha `$ $`=`$ $`m_{0\alpha }(v_\alpha )^1{\displaystyle d^3x\rho _\alpha (xq_\alpha )\left([E(x)E_{v_\alpha }(xq_\alpha )]+v_\alpha [B(x)B_{v_\alpha }(xq_\alpha )]\right)}`$ (5.5)
$`=`$ $`m_{0\alpha }(v_\alpha )^1{\displaystyle d^3x\rho _\alpha (x)\left(Z_1(x+q_\alpha ,t)+v_\alpha Z_2(x+q_\alpha ,t)\right)}+R_\alpha (t),`$
where $`m_{0\alpha }(v_\alpha )^1z=m_{\mathrm{b}\alpha }^{}{}_{}{}^{1}\gamma _\alpha ^1(z(v_\alpha z)v_\alpha )`$, $`z\mathrm{IR}^3`$, is the matrix inverse of $`m_{0\alpha }(v_\alpha )`$. For (5.5) it is important to note that adding the $`E_{v_\alpha }(xq_\alpha )`$-term and the $`v_\alpha B_{v_\alpha }(xq_\alpha )`$-term does not change the integral, as may be seen through Fourier transform using (2.4) and (2.5). Moreover, in (5.5) we have set
$$R_\alpha (t)=m_{0\alpha }(v_\alpha )^1\left(\underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}d^3x\rho _\alpha (xq_\alpha )\left[E_{v_\beta }(xq_\beta )+v_\alpha B_{v_\beta }(xq_\beta )\right]\right)$$
(5.6)
and
$$Z(x,t)=\left(\begin{array}{c}Z_1(x,t)\\ Z_2(x,t)\end{array}\right)=\left(\begin{array}{c}E(x,t)_{\beta =1}^NE_{v_\beta (t)}(xq_\beta (t))\\ B(x,t)_{\beta =1}^NB_{v_\beta (t)}(xq_\beta (t))\end{array}\right).$$
Maxwell’s equations and the relations $`(v)E_v(x)=B_v(x)+e\phi (x)v`$, $`(v)B_v(x)=E_v(x)`$, $`e=e_\alpha `$ for index $`\alpha `$, yield
$$\dot{Z}(t)=𝒜Z(t)f(t),\text{with}𝒜=\left(\begin{array}{cc}0& \\ & 0\end{array}\right)$$
(5.7)
and
$$f(x,t)=\underset{\beta =1}{\overset{N}{}}\left(\begin{array}{c}(\dot{v}_\beta (t)_v)E_{v_\beta }(xq_\beta (t))\\ (\dot{v}_\beta (t)_v)B_{v_\beta }(xq_\beta (t))\end{array}\right).$$
(5.8)
The Maxwell operator $`𝒜`$ generates a $`C^0`$-group $`U(t)`$, $`t\mathrm{IR}`$, of isometries in $`L^2(\mathrm{IR}^3)^3L^2(\mathrm{IR}^3)^3`$; see \[2, p. 435; (H2)\]. Therefore we have the mild solution representation
$$Z(x,t)=[U(t)Z(,0)](x)_0^t𝑑s[U(ts)f(,s)](x).$$
(5.9)
According to (2.3), $`Z(0)=0`$, so the first term drops out. To estimate the remaining term, we first state and prove some auxiliary lemmas that will be used frequently.
###### Lemma 5.1
For given $`f=(f_1,f_2)`$ with $`f_1=0`$ and $`f_2=0`$ we have for $`W(t,s,x)=(W_1(t,s,x),W_2(t,s,x))=[U(ts)f(,s)](x)`$
$`W_1(t,s,x)`$ $`=`$ $`{\displaystyle \frac{1}{4\pi (ts)^2}}{\displaystyle _{|yx|=(ts)}}d^2y\left[(ts)f_2(y,s)+f_1(y,s)+((yx))f_1(y,s)\right],`$
$`W_2(t,s,x)`$ $`=`$ $`{\displaystyle \frac{1}{4\pi (ts)^2}}{\displaystyle _{|yx|=(ts)}}d^2y\left[(ts)f_1(y,s)+f_2(y,s)+((yx))f_2(y,s)\right].`$
Proof : See \[6, Lemma 8.1\]. $`\mathrm{}`$
###### Lemma 5.2
(a) Let $`\xi (s)0`$ be some function. Assume that for $`y\mathrm{IR}^3`$, $`s[0,t]`$, and some $`f(y,s)=(f_1(y,s),f_2(y,s))`$ with $`f_1=0=f_2`$
$`|f_1(y,s)|+|f_2(y,s)|`$ $``$ $`C\xi (s){\displaystyle \underset{\beta =1}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\beta (s)|^2}},`$ (5.10)
$`|f_1(y,s)|+|f_2(y,s)|`$ $``$ $`C\xi (s){\displaystyle \underset{\beta =1}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\beta (s)|^3}},`$ (5.11)
Then for each $`\alpha =1,\mathrm{},N`$, $`t[0,T\epsilon ^{3/2}]`$, and $`|x|R_\phi `$
$$\left|_0^t𝑑s[U(ts)f(,s)](x+q_\alpha (t))\right|C\left(\underset{s[0,t]}{sup}\xi (s)\right).$$
(b) Under the hypotheses of (a), if instead of (5.10) and (5.11) it holds for fixed $`1\alpha N`$ that
$`|f_1(y,s)|+|f_2(y,s)|`$ $``$ $`C\xi (s){\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\beta (s)|^3}},`$ (5.12)
$`|f_1(y,s)|+|f_2(y,s)|`$ $``$ $`C\xi (s){\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\beta (s)|^4}},`$ (5.13)
then for $`t[0,T\epsilon ^{3/2}]`$ and $`|x|R_\phi `$ we have even that
$$\left|_0^t𝑑s[U(ts)f(,s)](x+q_\alpha (t))\right|C\left(\underset{s[0,t]}{sup}\xi (s)\right)\epsilon .$$
(c) Let $`\xi (\tau ,s)0`$ be some function. Assume that for $`y\mathrm{IR}^3`$, $`\tau [0,t]`$, $`s[0,\tau ]`$, and some $`g(y,\tau ,s)=(g_1(y,\tau ,s),g_2(y,\tau ,s))`$ with $`g_1=0=g_2`$ that
$`|g_1(y,\tau ,s)|+|g_2(y,\tau ,s)|`$ $``$ $`C\xi (\tau ,s){\displaystyle \underset{\alpha =1}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\alpha (s)|^3}},`$ (5.14)
$`|g_1(y,\tau ,s)|+|g_2(y,\tau ,s)|`$ $``$ $`C\xi (\tau ,s){\displaystyle \underset{\alpha =1}{\overset{N}{}}}{\displaystyle \frac{1}{1+|yq_\alpha (s)|^4}}.`$ (5.15)
Then for each $`\alpha =1,\mathrm{},N`$, $`t[0,T\epsilon ^{3/2}]`$, and $`|x|R_\phi `$
$$\left|_0^t𝑑\tau _0^\tau 𝑑s[U(ts)g(,\tau ,s)](x+q_\alpha (t))\right|C\left(\underset{(\tau ,s)\mathrm{\Delta }_t}{sup}\xi (\tau ,s)\right),$$
where $`\mathrm{\Delta }_t=\{(\tau ,s):\tau [0,t],s[0,\tau ]\}`$.
In (a)–(c), all constants $`C`$ on the right-hand sides are independent of $`\alpha `$, $`t`$, and $`x`$.
Proof : (a) Define $`W`$ as in Lemma 5.1. We derive the estimates with $`W_1`$. Fix $`1\alpha N`$, $`t[0,T\epsilon ^{3/2}]`$, $`s[0,t]`$, and $`|x|R_\phi `$. According to Lemma 5.1, (5.10), and (5.11)
$$|W_1(t,s,x+q_\alpha (t))|C\frac{\xi (s)}{(ts)^2}\underset{\beta =1}{\overset{N}{}}I_{\alpha \beta }^{(2)}(t,s,x),$$
with
$$I_{\alpha \beta }^{(n)}(t,s,x)=_{|yxq_\alpha (t)|=(ts)}d^2y\left[\frac{(ts)}{1+|yq_\beta (s)|^{n+1}}+\frac{1}{1+|yq_\beta (s)|^n}\right].$$
(5.16)
In the sum in (5.16), with general $`n2`$, we first consider the term $`I_{\alpha \alpha }^{(n)}(t,s,x)`$, i.e., the one with $`\beta =\alpha `$. In this case according to (2.8), $`|yq_\beta (s)||yxq_\alpha (t)||x||q_\alpha (t)q_\alpha (s)|(ts)R_\phi C\sqrt{\epsilon }(ts)(ts)/2R_\phi `$ for $`\epsilon `$ small. Therefore $`|yq_\beta (s)|(ts)/4`$ for $`st4R_\phi `$. We hence obtain for $`\beta =\alpha `$ and $`st4R_\phi `$
$$I_{\alpha \alpha }^{(n)}(t,s,x)C\frac{(ts)^2}{1+(ts)^n}.$$
(5.17)
On the other hand, for $`s[t4R_\phi ,t]`$
$`I_{\alpha \alpha }^{(n)}(t,s,x)`$ $``$ $`C(ts)^2[(ts)+1]C(ts)^2[4R_\phi +1]C(ts)^2{\displaystyle \frac{1}{1+(4R_\phi )^n}}`$
$``$ $`C{\displaystyle \frac{(ts)^2}{1+(ts)^n}}.`$
Hence (5.17) shows that the latter estimate holds for any $`s[0,t]`$. Since
$$_0^t\frac{ds}{1+(ts)^2}C,_0^t𝑑\tau _0^\tau \frac{ds}{1+(ts)^3}C,$$
the term with $`\beta =\alpha `$ will satisfy the claimed estimates not only in (a), but also in (c).
Next we turn to deriving a bound for $`I_{\alpha \beta }^{(2)}(t,s,x)`$ with $`\beta \alpha `$. First note that for some portion of the interval $`[0,t]`$ the preceding argument applies again. For this, define $`t_0=4(R_\phi +C^{}\epsilon ^1)`$. Then for $`stt_0`$ we find by (2.8) for $`\epsilon `$ small that on the $`y`$-sphere
$`|yq_\beta (s)|`$ $``$ $`|yxq_\alpha (t)||x||q_\alpha (t)q_\beta (s)|`$
$``$ $`(ts)R_\phi |q_\alpha (t)q_\beta (t)||q_\beta (t)q_\beta (s)|`$
$``$ $`(ts)R_\phi C^{}\epsilon ^1C\sqrt{\epsilon }(ts)(ts)/2R_\phi C^{}\epsilon ^1(ts)/4.`$
Therefore as in (5.17) for general $`n2`$
$$I_{\alpha \beta }^{(n)}(t,s,x)C\frac{(ts)^2}{1+(ts)^n},s[0,tt_0],$$
(5.18)
and it remains to estimate $`I_{\alpha \beta }^{(2)}(t,s,x)`$ for $`\beta \alpha `$ and $`s[tt_0,t]`$. To do so, we note that an explicit computation shows for $`z_1,z_2\mathrm{IR}^3`$ and $`\gamma 0`$
$`{\displaystyle _{|yz_1|=\gamma }}d^2y{\displaystyle \frac{1}{(1+|yz_2|^2)}}`$ $`=`$ $`{\displaystyle \frac{\pi \gamma }{|z_1z_2|}}\mathrm{log}\left({\displaystyle \frac{1+(\gamma +|z_1z_2|)^2}{1+(\gamma |z_1z_2|)^2}}\right)`$ (5.19)
$`=`$ $`{\displaystyle \frac{\pi \gamma }{|z_1z_2|}}\mathrm{log}\left(1+{\displaystyle \frac{4\gamma |z_1z_2|}{1+(\gamma |z_1z_2|)^2}}\right)`$
$``$ $`{\displaystyle \frac{4\pi \gamma ^2}{1+(\gamma |z_1z_2|)^2}},`$
as $`\mathrm{log}(1+A)A`$ for $`A0`$. Similarly, for $`n2`$
$`{\displaystyle _{|yz_1|=\gamma }}d^2y{\displaystyle \frac{1}{(1+|yz_2|^{n+1})}}`$ (5.20)
$`=`$ $`2\pi \gamma ^2{\displaystyle _1^1}{\displaystyle \frac{dr}{1+\left(|z_1z_2|^2+2\gamma |z_1z_2|r+\gamma ^2\right)^{(n+1)/2}}}`$
$``$ $`C\gamma ^2{\displaystyle _1^1}{\displaystyle \frac{dr}{\left(1+|z_1z_2|^2+2\gamma |z_1z_2|r+\gamma ^2\right)^{(n+1)/2}}}`$
$`=`$ $`C_n{\displaystyle \frac{\gamma }{|z_1z_2|}}\left({\displaystyle \frac{1}{\left[1+(|z_1z_2|\gamma )^2\right]^{(n1)/2}}}{\displaystyle \frac{1}{\left[1+(|z_1z_2|+\gamma )^2\right]^{(n1)/2}}}\right).`$
So in particular
$$_{|yz_1|=\gamma }d^2y\frac{1}{(1+|yz_2|^{n+1})}C\frac{\gamma }{|z_1z_2|},n2.$$
(5.21)
Below we will also need some more refined estimates, and for this purpose we note that according to (5.20) also
$$_{|yz_1|=\gamma }d^2y\frac{1}{(1+|yz_2|^3)}C\frac{\gamma ^2}{1+(|z_1z_2|+\gamma )^2}C\frac{\gamma ^2}{|z_1z_2|^2}.$$
(5.22)
Analogously we obtain
$$_{|yz_1|=\gamma }d^2y\frac{1}{(1+|yz_2|^4)}C\frac{1}{1+(|z_1z_2|\gamma )^2}\mathrm{min}\{1,\frac{\gamma ^2}{|z_1z_2|^2}\}.$$
(5.23)
As to bound $`I_{\alpha \beta }^{(2)}(t,s,x)`$ for $`\beta \alpha `$ and $`s[tt_0,t]`$ we then use (5.21) and (5.19) with $`z_1=x+q_\alpha (t)`$, $`z_2=q_\beta (s)`$, and $`\gamma =ts`$ to obtain for $`s[tt_0,t]`$
$$I_{\alpha \beta }^{(2)}(t,s,x)C\left(\frac{(ts)^2}{|x+q_\alpha (t)q_\beta (s)|}+\frac{(ts)^2}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}\right).$$
(5.24)
Therefore by (5.18) and (5.24)
$`{\displaystyle _0^t}𝑑s{\displaystyle \frac{\xi (s)}{(ts)^2}}I_{\alpha \beta }^{(2)}(t,s,x)`$ (5.25)
$``$ $`\left(\underset{s[0,t]}{sup}\xi (s)\right)\left({\displaystyle _0^{tt_0}}{\displaystyle \frac{ds}{(ts)^2}}I_{\alpha \beta }^{(2)}(t,s,x)+{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{(ts)^2}}I_{\alpha \beta }^{(2)}(t,s,x)\right)`$
$``$ $`C\left(\underset{s[0,t]}{sup}\xi (s)\right)({\displaystyle _0^{tt_0}}{\displaystyle \frac{ds}{1+(ts)^2}}+{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{|x+q_\alpha (t)q_\beta (s)|}}`$
$`+{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}}).`$
The first of the three integrals is bounded by a constant. Concerning the second, we have
$$|x+q_\alpha (t)q_\beta (s)||q_\alpha (t)q_\beta (t)||x||q_\beta (t)q_\beta (s)|C_{}\epsilon ^1R_\phi C\sqrt{\epsilon }(ts)$$
by (2.6) and (2.8). In the domain of integration $`[tt_0,t]`$ it holds that $`tst_0C\epsilon ^1`$, whence
$$|x+q_\alpha (t)q_\beta (s)|C_{}\epsilon ^1R_\phi C\epsilon ^{1/2}(C_{}/2)\epsilon ^1,s[tt_0,t],\beta \alpha ,|x|R_\phi ,$$
(5.26)
for $`\epsilon `$ small. Therefore the second integral can be bound by $`C\epsilon _{tt_0}^t𝑑sC\epsilon t_0C`$. To estimate the last integral $`=:J`$ on the right-hand side of (5.25), we substitute $`\theta =ts`$ to obtain
$$J=_0^{t_0}\frac{d\theta }{1+[\theta r(\theta )]^2}$$
(5.27)
with $`r(\theta )=|x+q_\alpha (t)q_\beta (t\theta )|`$. Observe that $`|\dot{r}(\theta )||\dot{q}_\beta (t\theta )|C\sqrt{\epsilon }`$ by (2.8). Thus $`\theta \chi (\theta )=\theta r(\theta )`$ is strictly increasing, and we can substitute $`\theta =\theta (\chi )`$ to get
$$J=_{\chi (0)}^{\chi (t_0)}\frac{d\chi }{1\dot{r}(\theta )}\left(\frac{1}{1+\chi ^2}\right)C_{\mathrm{IR}}\frac{d\chi }{1+\chi ^2}C.$$
Summarizing these estimates we obtain the bound claimed in part (a) of the lemma.
(b) Defining $`I_{\alpha \beta }^{(n)}`$ as in (5.16), we need to show
$$_0^t\frac{ds}{(ts)^2}I_{\alpha \beta }^{(3)}(t,s,x)C\epsilon ,\beta \alpha .$$
(5.28)
By (5.18),
$$_0^{tt_0}\frac{ds}{(ts)^2}I_{\alpha \beta }^{(3)}(t,s,x)C_0^{tt_0}\frac{ds}{(ts)^2}\frac{(ts)}{1+(ts)^2}.$$
In the domain of integration, $`(ts)t_0C\epsilon ^1`$, and hence
$$_0^{tt_0}\frac{ds}{(ts)^2}I_{\alpha \beta }^{(3)}(t,s,x)C\epsilon _0^t\frac{ds}{1+(ts)^2}C\epsilon .$$
(5.29)
Thus it remains to estimate the part of the integral in (5.28) for $`s[tt_0,t]`$. Firstly, by (5.22),
$`{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{(ts)^2}}{\displaystyle _{|yxq_\alpha (t)|=(ts)}}d^2y{\displaystyle \frac{1}{(1+|yq_\beta (s)|^3)}}`$ (5.30)
$`C{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{|x+q_\alpha (t)q_\beta (s)|^2}}C\epsilon ^2{\displaystyle _{tt_0}^t}𝑑s=C\epsilon ^2t_0C\epsilon .`$
Here we have used $`|x+q_\alpha (t)q_\beta (s)|(C_{}/2)\epsilon ^1`$ for $`\epsilon `$ small, cf. (5.26). Reference to this is possible, since we again have that $`\beta \alpha `$. Analogously we infer from (5.23) that
$`{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{(ts)^2}}{\displaystyle _{|yxq_\alpha (t)|=(ts)}}d^2y{\displaystyle \frac{(ts)}{(1+|yq_\beta (s)|^4)}}`$
$``$ $`{\displaystyle _{tt_0}^t}𝑑s{\displaystyle \frac{(ts)}{|x+q_\alpha (t)q_\beta (s)|^2}}\left({\displaystyle \frac{1}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}}\right)`$
$``$ $`C\epsilon ^2t_0{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}}=C\epsilon JC\epsilon ,`$
with the bounded $`J`$ from (5.27). This together with (5.30) and (5.29) shows that (5.28) is satisfied.
(c) Due to the remarks in (a), (5.14), and (5.15) we only have to prove
$$_0^t𝑑\tau _0^\tau \frac{ds}{(ts)^2}I_{\alpha \beta }^{(3)}(t,s,x)C,\beta \alpha ,t[0,T\epsilon ^{3/2}],|x|R_\phi .$$
(5.31)
We decompose the domain of integration $`\mathrm{\Delta }_t=\{(\tau ,s):\tau [0,t],s[0,\tau ]\}`$ in $`\mathrm{\Delta }_{t,1}=\mathrm{\Delta }_t\{(\tau ,s):s[0,tt_0]\}`$ and $`\mathrm{\Delta }_{t,2}=\{(\tau ,s):\tau [tt_0,t],s[tt_0,\tau ]\}`$. On $`\mathrm{\Delta }_{t,1}`$ we can utilize (5.18) to get
$$_{\mathrm{\Delta }_{t,1}}𝑑\tau 𝑑s\frac{1}{(ts)^2}I_{\alpha \beta }^{(3)}(t,s,x)C_0^t𝑑\tau _0^\tau 𝑑s\frac{1}{1+(ts)^3}C.$$
(5.32)
Since again $`tst_0C\epsilon ^1`$ for $`(\tau ,s)\mathrm{\Delta }_{t,2}`$, by (5.26) and (5.21)
$`{\displaystyle _{\mathrm{\Delta }_{t,2}}𝑑\tau 𝑑s\frac{1}{(ts)^2}_{|yxq_\alpha (t)|=(ts)}\frac{d^2y}{1+|yq_\beta (s)|^3}}`$ (5.33)
$``$ $`C{\displaystyle _{\mathrm{\Delta }_{t,2}}𝑑\tau 𝑑s\frac{1}{(ts)}\frac{1}{|x+q_\alpha (t)q_\beta (s)|}}`$
$``$ $`C\epsilon {\displaystyle _{tt_0}^t}𝑑\tau {\displaystyle _{tt_0}^\tau }{\displaystyle \frac{ds}{ts}}=C\epsilon {\displaystyle _t^t}𝑑s=C\epsilon t_0C.`$
In addition, by (5.23)
$`{\displaystyle _{\mathrm{\Delta }_{t,2}}𝑑\tau 𝑑s\frac{1}{(ts)}_{|yxq_\alpha (t)|=(ts)}\frac{d^2y}{1+|yq_\beta (s)|^4}}`$ (5.34)
$``$ $`{\displaystyle _{\mathrm{\Delta }_{t,2}}𝑑\tau 𝑑s\frac{1}{(ts)}\frac{1}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}}`$
$`=`$ $`{\displaystyle _{tt_0}^t}{\displaystyle \frac{ds}{1+[(ts)|x+q_\alpha (t)q_\beta (s)|]^2}}C,`$
since the last integral is just $`J`$ from (5.27) and hence bounded. By (5.32), (5.33), and (5.34) we thus have proved (5.31). $`\mathrm{}`$
###### Lemma 5.3
Define $`\varphi _v(x)`$ through $`\widehat{\varphi }_v(k)=e\widehat{\phi }(k)/[k^2(kv)^2]`$. Then for $`x\mathrm{IR}^3`$ and $`|v|\overline{v}<1`$, with $`=_x`$,
$`|\varphi _v(x)|+|_v\varphi _v(x)|+|_v^2\varphi _v(x)|+|_v^3\varphi _v(x)|C|e|(1+|x|)^2,`$
$`|^2\varphi _v(x)|+|_v^2\varphi _v(x)|+|_v^2^2\varphi _v(x)|+|_v^3^2\varphi _v(x)|C|e|(1+|x|)^3,`$
$`|^3\varphi _v(x)|+|_v^3\varphi _v(x)|+|_v^2^3\varphi _v(x)|+|_v^3^3\varphi _v(x)|C|e|(1+|x|)^4,`$
$`|^4\varphi _v(x)|+|_v^4\varphi _v(x)|+|_v^2^4\varphi _v(x)|+|_v^3^4\varphi _v(x)|C|e|(1+|x|)^5.`$
Proof : Tedious calculations; see also the appendices of . $`\mathrm{}`$
Now we can estimate $`_0^t𝑑s[U(ts)f(,s)](x+q_\alpha (t))`$, cf. (5.9), for $`t[0,T\epsilon ^{3/2}]`$ and $`|x|R_\phi `$, using Lemma 5.1 and Lemma 5.2(a), with $`f=(f_1,f_2)`$ defined by (5.8). Since $`B_v=0`$, and $`E_v=e\phi `$ is independent of $`v`$, we have $`f_1=0=f_2`$. Concerning (5.10) and (5.11), note $`|_vE_v(x)|+|_vB_v(x)|C(|\varphi _v(x)|+|_v\varphi _v(x)|)C|e|(1+|x|)^2`$ and $`|_vE_v(x)|+|_vB_v(x)|C(|^2\varphi _v(x)|+|_v^2\varphi _v(x)|)C|e|(1+|x|)^3`$ by Lemma 5.3. Thus (5.10) and (5.11) are satisfied with $`\xi (s)=\left(\mathrm{max}_{1\beta N}|\dot{v}_\beta (s)|\right)\left(\mathrm{max}_{1\beta N}|e_\beta |\right)`$. As $`Z(x,0)=0`$, hence (5.9) in conjunction with Lemma 5.2(a) yields for $`\alpha =1,\mathrm{},N`$
$$|Z(x+q_\alpha (t),t)|C\left(\underset{s[0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\dot{v}_\beta (s)|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right),t[0,T\epsilon ^{3/2}],|x|R_\phi .$$
(5.35)
We will utilize this further in (5.5), and to this end we also need to bound $`R_\alpha (t)`$ from (5.6). For fixed $`\beta \alpha `$ one calculates for the interaction terms
$`\mathrm{\Psi }_{\alpha \beta }(t)`$ $`=`$ $`{\displaystyle d^3x\rho _\alpha (xq_\alpha (t))\varphi _{v_\beta (t)}(xq_\beta (t))}`$ (5.36)
$`=`$ $`(i)e_\alpha e_\beta {\displaystyle d^3kk\frac{|\widehat{\phi }(k)|^2}{k^2(kv_\beta (t))^2}e^{ik[q_\beta (t)q_\alpha (t)]}}`$
$`=`$ $`{\displaystyle \frac{e_\alpha e_\beta }{4\pi }}{\displaystyle d^3xd^3y\phi (xq_\alpha (t))\phi (yq_\beta (t))\zeta _{v_\beta (t)}(xy)},`$
with
$$\zeta _v(x)=\frac{1}{[(1v^2)x^2+(xv)^2]^{1/2}},\widehat{\zeta }_v(k)=\sqrt{\frac{2}{\pi }}\frac{1}{k^2(kv)^2},|v|<1.$$
(5.37)
Then $`sup_{t[0,T\epsilon ^{3/2}]}|\zeta _{v_\beta (t)}(x)|C(1+|x|)^2`$ due to (2.8). By $`(C)`$, in (5.36) we only need to integrate over $`(x,y)`$ that have $`|xq_\alpha (t)|R_\phi `$ and $`|yq_\beta (t)|R_\phi `$. Then by (2.6), $`|xy||q_\alpha (t)q_\beta (t)|2R_\phi C_{}\epsilon ^12R_\phi (C_{}/2)\epsilon ^1`$ for $`\epsilon `$ small. Therefore (5.36) shows
$$|\mathrm{\Psi }_{\alpha \beta }(t)|C\epsilon ^2,t[0,T\epsilon ^{3/2}],\alpha \beta .$$
(5.38)
By definition of $`B_v(x)`$ and $`E_v(x)`$ we have
$$R_\alpha (t)=m_{0\alpha }(v_\alpha (t))^1\underset{\beta \alpha }{}\left(\mathrm{\Psi }_{\alpha \beta }(t)+[v_\beta (t)\mathrm{\Psi }_{\alpha \beta }(t)]v_\beta (t)+v_\alpha (t)[v_\beta (t)\mathrm{\Psi }_{\alpha \beta }(t)]\right)$$
(5.39)
cf. (5.6), and therefore (5.38) together with (2.8) implies
$$|R_\alpha (t)|C\epsilon ^2,t[0,T\epsilon ^{3/2}].$$
(5.40)
Hence (5.5), (5.35), and (5.40) finally yield
$$|\dot{v}_\alpha (t)|C\left(\underset{s[0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\dot{v}_\beta (s)|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)+C\epsilon ^2,$$
for every $`\alpha =1,\mathrm{},N`$ and $`t[0,T\epsilon ^{3/2}]`$. Choosing $`\mathrm{max}_{1\beta N}|e_\beta |\overline{e}`$ with $`\overline{e}>0`$ sufficiently small, we find that for $`\alpha =1,\mathrm{},N`$
$$\underset{t[0,T\epsilon ^{3/2}]}{sup}|\dot{v}_\alpha (t)|C\epsilon ^2.$$
(5.41)
For later reference we also note that then according to (5.35)
$$|Z(x+q_\alpha (t),t)|C\epsilon ^2,\alpha =1,\mathrm{},N,t[0,T\epsilon ^{3/2}],|x|R_\phi .$$
(5.42)
### 5.3 Bounding $`|\ddot{v}_\alpha (t)|`$
By (2.8) we have in particular that
$$|v_\alpha (t)v_\beta (t)|C\sqrt{\epsilon },t[0,T\epsilon ^{3/2}].$$
(5.43)
In order to estimate the derivative of Equ. (5.5), first note that using the explicit form of $`m_{0\alpha }(v_\alpha )^1`$ we obtain from (5.41) that
$$\left|\frac{d}{dt}m_{0\alpha }(v_\alpha (t))^1\right|C|\dot{v}_\alpha (t)|C\epsilon ^2.$$
(5.44)
Hence by (5.5), (5.42) and (5.40)
$$|\ddot{v}_\alpha (t)|C\left(\epsilon ^4+|M_\alpha (t)|+|\dot{R}_\alpha (t)|\right),$$
(5.45)
with $`R_\alpha `$ defined in (5.6), and
$$M_\alpha (t)=d^3x\rho _\alpha (x)\left[(L_\alpha (t)Z_1)(x+q_\alpha (t),t)+v_\alpha (t)(L_\alpha (t)Z_2)(x+q_\alpha (t),t)\right],$$
(5.46)
where $`L_\alpha (t)\varphi =(v_\alpha (t))\varphi +_t\varphi `$ for a function $`\varphi =\varphi (x,t)`$. We first estimate $`M_\alpha (t)`$. Let $`\mathrm{\Sigma }_\alpha (x,t)=(L_\alpha (t)Z)(x,t)`$. Since generally $`\frac{d}{dt}[L_\alpha (t)\varphi ]=L_\alpha (t)\dot{\varphi }+(\dot{v}_\alpha )\varphi `$ and, see (5.7), $`\dot{Z}=𝒜Zf`$ with $`f`$ from (5.8), we obtain
$$\dot{\mathrm{\Sigma }}_\alpha =𝒜\mathrm{\Sigma }_\alpha +(\dot{v}_\alpha )ZL_\alpha (t)f.$$
According to (2.3) it may be shown that $`\mathrm{\Sigma }_\alpha (x,0)=0`$. We hence get
$$\mathrm{\Sigma }_\alpha (x+q_\alpha (t),t)=_0^t𝑑\tau \left[U(t\tau )\left((\dot{v}_\alpha (\tau ))Z(,\tau )L_\alpha (\tau )f(,\tau )\right)\right](x+q_\alpha (t)).$$
As a consequence of $`\frac{d}{dt}(Z)=(𝒜Zf)=𝒜(Z)f`$ and $`Z(x,0)=0`$, we obtain from the group property of $`U()`$ that
$`\mathrm{\Sigma }_{\alpha ,1}(x+q_\alpha (t),t)`$ $`=`$ $`{\displaystyle _0^t}𝑑\tau \left[U(t\tau )\left((\dot{v}_\alpha (\tau ))Z(,\tau )\right)\right](x+q_\alpha (t))`$
$`=`$ $`{\displaystyle _0^t}𝑑\tau {\displaystyle _0^\tau }𝑑s\left[U(ts)\left((\dot{v}_\alpha (\tau ))f(,s)\right)\right](x+q_\alpha (t)).`$
With $`g(y,\tau ,s)=\dot{v}_\alpha (\tau )f(y,s)`$ it follows from the definitions of $`f`$, $`E_v(x)`$, and $`B_v(x)`$ that $`g=0`$. Moreover, by (5.41) and Lemma 5.3 we find that (5.14) and (5.15) are satisfied with $`\xi (\tau ,s)=\epsilon ^4`$. Therefore Lemma 5.2(c) applies to yield for $`\alpha =1,\mathrm{},N`$
$$|\mathrm{\Sigma }_{\alpha ,1}(x+q_\alpha (t),t)|C\epsilon ^4,t[0,T\epsilon ^{3/2}],|x|R_\phi .$$
(5.47)
To estimate
$$\mathrm{\Sigma }_{\alpha ,2}(x+q_\alpha (t),t)=_0^t𝑑\tau \left[U(t\tau )\left(L_\alpha (\tau )f(,\tau )\right)\right](x+q_\alpha (t)),$$
observe that
$`[L_\alpha (\tau )f(,\tau )](x)`$ $`=`$ $`v_\alpha (\tau )f(x,\tau )+_tf(x,\tau )`$
$`=`$ $`{\displaystyle \underset{\beta =1}{\overset{N}{}}}\left\{(\ddot{v}_\beta _v)\mathrm{\Phi }_{v_\beta }(xq_\beta )+(\dot{v}_\beta _v)^2\mathrm{\Phi }_{v_\beta }(xq_\beta )\right\}`$
$`+{\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{\overset{N}{}}}_{xv}^2\mathrm{\Phi }_{v_\beta }(xq_\beta )(v_\alpha v_\beta ,\dot{v}_\beta )`$
$`=:`$ $`f^{\mathrm{}}(\tau ,y)+f^{\mathrm{}}(\tau ,y),`$
with all time arguments taken at time $`\tau `$, and $`\mathrm{\Phi }_v=(E_v,B_v)`$. Since $`B_v=0`$ and $`E_v=e\phi `$ is independent of $`v`$, we have that $`f^{\mathrm{}}=0=f^{\mathrm{}}`$. In addition, $`f^{\mathrm{}}`$ satisfies (5.10) and (5.11) with
$$\xi ^{\mathrm{}}(\tau )=\left(\underset{1\beta N}{\mathrm{max}}|\ddot{v}_\beta (\tau )|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)+\epsilon ^4.$$
Because $`f^{\mathrm{}}`$ has an additional $`x`$-derivative, moreover (5.12) and (5.13) hold for $`f^{\mathrm{}}`$, with
$$\xi ^{\mathrm{}}(\tau )=\left(\underset{1\beta N}{\mathrm{max}}|v_\alpha (\tau )v_\beta (\tau )|\right)\epsilon ^2,$$
as again follows from Lemma 5.3 and (5.41). Thus Lemma 5.2(a) and (b) imply that for all $`\alpha =1,\mathrm{},N`$, $`t[0,T\epsilon ^{3/2}]`$, and $`|x|R_\phi `$
$`|\mathrm{\Sigma }_{\alpha ,2}(x+q_\alpha (t),t)|`$
$``$ $`\left|{\displaystyle _0^t}𝑑\tau [U(t\tau )f^{\mathrm{}}(,\tau )](x+q_\alpha (t))\right|+\left|{\displaystyle _0^t}𝑑\tau [U(t\tau )f^{\mathrm{}}(,\tau )](x+q_\alpha (t))\right|`$
$``$ $`C\left(\underset{\tau [0,t]}{sup}\xi ^{\mathrm{}}(\tau )+\underset{\tau [0,t]}{sup}\xi ^{\mathrm{}}(\tau )\epsilon \right)`$
$``$ $`C\left[\epsilon ^4+\left(\underset{\tau [0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\ddot{v}_\beta (\tau )|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)+\left(\underset{\tau [0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|v_\alpha (\tau )v_\beta (\tau )|\right)\epsilon ^3\right].`$
Hence by (5.47) and (5.43) for $`\alpha =1,\mathrm{},N`$, $`t[0,T\epsilon ^{3/2}]`$, and $`|x|R_\phi `$,
$$|\mathrm{\Sigma }_\alpha (x+q_\alpha (t),t)|C\left[\epsilon ^{7/2}+\left(\underset{\tau [0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\ddot{v}_\beta (\tau )|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)\right].$$
According to the definition of $`M_\alpha (t)`$ in (5.46) we therefore have
$`|M_\alpha (t)|`$ $`=`$ $`\left|{\displaystyle _{|x|R_\phi }}d^3x\rho _\alpha (x)\left[\mathrm{\Sigma }_{\alpha ,1}(x+q_\alpha (t),t)+v_\alpha (t)\mathrm{\Sigma }_{\alpha ,2}(x+q_\alpha (t),t)\right]\right|`$ (5.48)
$``$ $`C\left[\epsilon ^{7/2}+\left(\underset{\tau [0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\ddot{v}_\beta (\tau )|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)\right].`$
To further estimate the right-hand side of (5.45), we have to bound $`\dot{R}_\alpha (t)`$, with $`R_\alpha (t)`$ from (5.6). Calculating $`\dot{R}_\alpha (t)`$ explicitly we obtain
$`\dot{R}_\alpha (t)`$
$`=`$ $`\left({\displaystyle \frac{d}{dt}}m_{0\alpha }(v_\alpha )^1\right)m_{0\alpha }(v_\alpha )R_\alpha (t)`$
$`+m_{0\alpha }(v_\alpha )^1\left({\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{}}{\displaystyle d^3x\rho _\alpha (xq_\alpha )\left[(\dot{v}_\beta _v)E_{v_\beta }(xq_\beta )+v_\alpha (\dot{v}_\beta _v)B_{v_\beta }(xq_\beta )\right]}\right)`$
$`+m_{0\alpha }(v_\alpha )^1({\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{}}{\displaystyle }d^3x\rho _\alpha (xq_\alpha )[((v_\alpha v_\beta ))E_{v_\beta }(xq_\beta )`$
$`+v_\alpha ((v_\alpha v_\beta ))B_{v_\beta }(xq_\beta )])`$
$`+m_{0\alpha }(v_\alpha )^1\left({\displaystyle \underset{\stackrel{\beta =1}{\beta \alpha }}{}}{\displaystyle d^3x\rho _\alpha (xq_\alpha )\dot{v}_\alpha }B_{v_\beta }(xq_\beta )\right)`$
$`=:`$ $`\dot{R}_{\alpha ,1}(t)+\dot{R}_{\alpha ,2}(t)+\dot{R}_{\alpha ,3}(t)+\dot{R}_{\alpha ,4}(t)`$
with all time arguments at time $`t`$. Firstly,
$$|\dot{R}_{\alpha ,1}(t)|=\left|\left(\frac{d}{dt}m_{0\alpha }(v_\alpha )^1\right)m_{0\alpha }(v_\alpha )R_\alpha (t)\right|C\epsilon ^4$$
(5.49)
for $`\alpha =1,\mathrm{},N`$ and $`t[0,T\epsilon ^{3/2}]`$ by (5.44) and (5.40). Since $`B_v(x)=v\varphi _v(x)`$, by (5.41), (2.8), and (5.38) also
$$|\dot{R}_{\alpha ,4}(t)|C\epsilon ^{9/2}.$$
(5.50)
What concerns $`\dot{R}_{\alpha ,2}(t)`$, we may repeat the calculation in (5.36) to obtain
$`_v\mathrm{\Psi }_{\alpha \beta }(t)`$ $`:=`$ $`{\displaystyle d^3x\rho (xq_\alpha (t))_{xv}^2\varphi _{v_\beta (t)}(xq_\beta (t))}`$
$`=`$ $`{\displaystyle \frac{1}{4\pi }}{\displaystyle d^3xd^3y\rho (xq_\alpha (t))\rho (yq_\beta (t))_{xv}^2\zeta _{v_\beta (t)}(xy)},`$
with $`\zeta _v(x)`$ from (5.37). Since $`sup_{t[0,T\epsilon ^{3/2}]}|_{xv}^2\zeta _{v_\beta (t)}(x)|C(1+|x|)^2`$, we get as before that
$$|_v\mathrm{\Psi }_{\alpha \beta }(t)|C\epsilon ^2,t[0,T\epsilon ^{3/2}],\alpha \beta ,$$
and hence by (5.41)
$$|\dot{R}_{\alpha ,2}(t)|C\epsilon ^4.$$
(5.51)
So finally we have to bound $`\dot{R}_{\alpha ,3}(t)`$, and this relies on a similar argument. Here we have
$`\mathrm{\Psi }_{\alpha \beta }(t)`$ $`:=`$ $`{\displaystyle d^3x\rho (xq_\alpha (t))^2\varphi _{v_\beta (t)}(xq_\beta (t))}`$
$`=`$ $`{\displaystyle \frac{1}{4\pi }}{\displaystyle d^3xd^3y\rho (xq_\alpha (t))\rho (yq_\beta (t))^2\zeta _{v_\beta (t)}(xy)},`$
and $`sup_{t[0,T\epsilon ^{3/2}]}|^2\zeta _{v_\beta (t)}(x)|C(1+|x|)^3`$. This in turn yields
$$|\mathrm{\Psi }_{\alpha \beta }(t)|C\epsilon ^3,t[0,T\epsilon ^{3/2}],\alpha \beta .$$
Using the explicit form of $`E_v(x)`$ and $`B_v(x)`$, as in (5.39), we then get for $`t[0,T\epsilon ^{3/2}]`$
$$|\dot{R}_{\alpha ,3}(t)|C\epsilon ^3|v_\alpha (t)v_\beta (t)|C\epsilon ^{7/2},$$
(5.52)
by (5.43). Summarizing (5.49), (5.50), (5.51), and (5.52) it follows that
$$|\dot{R}_\alpha (t)|C\epsilon ^{7/2},\alpha =1,\mathrm{},N,t[0,T\epsilon ^{3/2}].$$
(5.53)
Consequently, by (5.45), (5.48), and (5.53) for $`\alpha =1,\mathrm{},N`$ and $`t[0,T\epsilon ^{3/2}]`$
$$|\ddot{v}_\alpha (t)|C\left(\epsilon ^4+|M_\alpha (t)|+|\dot{R}_\alpha (t)|\right)C\left[\epsilon ^{7/2}+\left(\underset{\tau [0,t]}{sup}\underset{1\beta N}{\mathrm{max}}|\ddot{v}_\beta (\tau )|\right)\left(\underset{1\beta N}{\mathrm{max}}|e_\beta |\right)\right].$$
Choosing $`\mathrm{max}_{1\beta N}|e_\beta |\overline{e}`$ with $`\overline{e}`$ sufficiently small we hence obtain
$$\underset{t[0,T\epsilon ^{3/2}]}{sup}|\ddot{v}_\alpha (t)|C\epsilon ^{7/2},\alpha =1,\mathrm{},N.$$
This completes the proof of Lemma 2.1. $`\mathrm{}`$
## 6 Appendix B: Proof of Lemma 3.2
Here we give the proof of Lemma 3.2. We verify e.g. (b). To compare the left-hand side to the right-hand side of the assertion, we will insert some additional terms and estimate the corresponding differences $`D_j(t)`$, $`j=1,2,3`$, for $`t[t_0,T\epsilon ^{3/2}]`$, where $`t_0=4(R_\phi +C^{}\epsilon ^1)`$. First we introduce
$`D_1(t)`$ $`=`$ $`i{\displaystyle _0^t}𝑑\tau {\displaystyle d^3k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\left\{e^{ik[q_\beta (t)q_\beta (t\tau )]}e^{ik[\tau v_\beta \frac{1}{2}\tau ^2\dot{v}_\beta ]}\right\}\frac{\mathrm{sin}|k|\tau }{|k|}k}`$
$`=`$ $`_\xi {\displaystyle _0^t}𝑑\tau {\displaystyle d^3k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\left\{e^{ik[q_\beta (t)q_\beta (t\tau )]}e^{ik[\tau v_\beta \frac{1}{2}\tau ^2\dot{v}_\beta ]}\right\}\frac{\mathrm{sin}|k|\tau }{|k|}}`$
$`=`$ $`_\xi {\displaystyle d^3xd^3y\phi (x)\phi (y)}`$
$`\times {\displaystyle _0^t}d\tau \{\psi _\tau ([\xi _{\alpha \beta }+xq_\beta (t\tau )][yq_\beta (t)])\psi _\tau ([x{\displaystyle \frac{1}{2}}\tau ^2\dot{v}_\beta ][y\tau v_\beta ])\},`$
as follows through application of the Fourier transform, with $`\xi _{\alpha \beta }=q_\alpha (t)q_\beta (t)`$, and $`\psi _\tau (x)=(4\pi |x|)^1`$ for $`|x|=\tau `$ whereas $`\psi _\tau (x)=0`$ otherwise. We claim that for $`x,y\mathrm{IR}^3`$ with $`|x|,|y|R_\phi `$ and $`t[t_0,T\epsilon ^{3/2}]`$ there exists a unique $`\tau _0=\tau _0(x,y,t,\xi _{\alpha \beta })[0,t_0][0,t]`$ such that
$$\tau _0=\left|[\xi _{\alpha \beta }+xq_\beta (t\tau _0)][yq_\beta (t)]\right|.$$
(6.2)
To see this, observe with $`\theta (\tau )=\tau |[\xi _{\alpha \beta }+xq_\beta (t\tau )][yq_\beta (t)]|`$ that $`0\theta (0)(2R_\phi +C^{}\epsilon ^1)`$ and $`\theta ^{}(\tau )1C_v\sqrt{\epsilon }`$ by (2.6) and (2.8). For $`\epsilon `$ so small that $`1C_v\sqrt{\epsilon }1/2`$ we hence obtain $`\theta (t_0)(2R_\phi +C^{}\epsilon ^1)+t_0/2=2C^{}\epsilon ^1`$. This shows $`\theta `$ has a unique zero $`\tau _0[0,t_0]`$. Moreover (6.2) together with (2.6) implies
$$\tau _0|\xi _{\alpha \beta }||xq_\beta (t\tau _0)][yq_\beta (t)]|C_{}\epsilon ^12R_\phi C_v\sqrt{\epsilon }\tau _0,$$
whence also $`\tau _0C\epsilon ^1`$ for $`\epsilon `$ small. Similarly, we find a unique $`\tau _1=\tau _1(x,y,t,\xi _{\alpha \beta })`$ satisfying
$$\tau _1=\left|[\xi _{\alpha \beta }+x\frac{1}{2}\tau _1^2\dot{v}_\beta ][y\tau _1v_\beta ]\right|,$$
(6.3)
with $`\tau _1`$ having the same properties as $`\tau _0`$. By definition of $`\psi _\tau `$ we therefore may simply write
$$D_1(t)=d^3xd^3y\phi (x)\phi (y)_\xi \left(\tau _0^1\tau _1^1\right).$$
(6.4)
To estimate this, we calculate from (6.2) that
$`_\xi \tau _0^1`$ $`=`$ $`\tau _0^3\{([\xi _{\alpha \beta }+xq_\beta (t\tau _0)][yq_\beta (t)])`$
$`+([\xi _{\alpha \beta }+xq_\beta (t\tau _0)][yq_\beta (t)])v_\beta (t\tau _0)_\xi \tau _0\},`$
with an analogous expression for $`_\xi \tau _1^1`$. Therefore
$`\left|_\xi \left(\tau _0^1\tau _1^1\right)\right|`$ $``$ $`C(\tau _0^3|q_\beta (t\tau _0)q_\beta (t\tau _1)|[1+|v_\beta (t\tau _1)||_\xi \tau _1|]`$ (6.5)
$`+|\tau _0^3\tau _1^3|\left|[\xi _{\alpha \beta }+xq_\beta (t\tau _1)][yq_\beta (t)]\right|\left[1+|v_\beta (t\tau _1)||_\xi \tau _1|\right]`$
$`+\tau _0^2|v_\beta (t\tau _0)v_\beta (t\tau _1)||_\xi \tau _1|`$
$`+\tau _0^2|v_\beta (t\tau _0)||_\xi (\tau _0\tau _1)|).`$
¿From (6.2), (6.3), and according to the Taylor expansion
$$q_\beta (t\tau )=q_\beta (t)\tau v_\beta +\frac{1}{2}\tau ^2\dot{v}_\beta +𝒪(\epsilon ^{7/2}\tau ^3),$$
cf. Lemma 2.1, it follows that
$`|\tau _0\tau _1|`$ $``$ $`\left|\tau _0v_\beta {\displaystyle \frac{1}{2}}\tau _0^2\dot{v}_\beta \tau _1v_\beta +{\displaystyle \frac{1}{2}}\tau _1^2\dot{v}_\beta \right|+𝒪(\epsilon ^{7/2}\tau _0^3)`$
$``$ $`C\sqrt{\epsilon }|\tau _0\tau _1|+C\epsilon ^2(\tau _0+\tau _1)|\tau _0\tau _1|+𝒪(\epsilon ^{7/2}\tau _0^3),`$
whence
$$|\tau _0\tau _1|=𝒪(\sqrt{\epsilon }),|\tau _0^3\tau _1^3|=𝒪(\epsilon ^{9/2}),$$
recall $`C\epsilon ^1\tau _0,\tau _1t_0=𝒪(\epsilon ^1)`$. Differentiating (6.2) and (6.3) w.r. to $`\xi =\xi _{\alpha \beta }`$ we moreover get $`|_\xi \tau _0|+|_\xi \tau _1|=𝒪(1)`$, and after a longer calculation which we omit also $`|_\xi (\tau _0\tau _1)|C\left(\epsilon ^{3/2}+\sqrt{\epsilon }|_\xi (\tau _0\tau _1)|\right)`$, thus
$$|_\xi (\tau _0\tau _1)|C\epsilon ^{3/2}.$$
Utilizing these estimates and Lemma 2.1 in (6.5), we consequently obtain $`|_\xi (\tau _0^1\tau _1^1)|C\epsilon ^{7/2}`$. Hence (6.4) yields
$$\underset{t[t_0,T\epsilon ^{3/2}]}{sup}|D_1(t)|C\epsilon ^{7/2}$$
(6.6)
as desired. Next, with
$`D_2(t)`$ $`=`$ $`i{\displaystyle _0^t}𝑑\tau {\displaystyle d^3k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}}`$
$`\times \left\{e^{ik[\tau v_\beta \frac{1}{2}\tau ^2\dot{v}_\beta ]}\left(1ik\left[\tau v_\beta {\displaystyle \frac{1}{2}}\tau ^2\dot{v}_\beta \right]{\displaystyle \frac{1}{2}}\tau ^2(kv_\beta )^2\right)\right\}{\displaystyle \frac{\mathrm{sin}|k|\tau }{|k|}}k`$
it may be shown in a a similar way that
$$\underset{t[t_0,T\epsilon ^{3/2}]}{sup}|D_2(t)|C\epsilon ^{7/2}.$$
(6.7)
Finally we need to compare $`_0^t𝑑\tau (\mathrm{})`$ to the infinite $`d\tau `$-integral and thus let
$$D_3(t)=i_t^{\mathrm{}}𝑑\tau d^3k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}\left(1ik\left[\tau v_\beta \frac{1}{2}\tau ^2\dot{v}_\beta \right]\frac{1}{2}\tau ^2(kv_\beta )^2\right)\frac{\mathrm{sin}|k|\tau }{|k|}k.$$
With the notation
$$K_p=e^{ik\xi _{\alpha \beta }}_t^{\mathrm{}}𝑑\tau \frac{\mathrm{sin}|k|\tau }{|k|}\tau ^p,p=0,\mathrm{},2,$$
this may be rewritten as
$$D_3(t)=d^3k|\widehat{\phi }(k)|^2\left(_\xi K_0(v_\beta _\xi )_\xi K_1+\frac{1}{2}(\dot{v}_\beta _\xi )_\xi K_2\frac{1}{2}(v_\beta _\xi )^2_\xi K_2\right).$$
Thus we only need to estimate
$`{\displaystyle d^3k|\widehat{\phi }(k)|^2K_p}`$ $`=`$ $`{\displaystyle d^3k|\widehat{\phi }(k)|^2e^{ik\xi _{\alpha \beta }}_t^{\mathrm{}}𝑑\tau \frac{\mathrm{sin}|k|\tau }{|k|}\tau ^p}`$ (6.8)
$`=`$ $`{\displaystyle d^3xd^3y\phi (x)\phi (y)_t^{\mathrm{}}𝑑\tau \psi _\tau (\xi _{\alpha \beta }+xy)\tau ^p},`$
the latter equality follows analogously to (6). However, for $`|x|,|y|R_\phi `$ and $`t[t_0,T\epsilon ^{3/2}]`$ we obtain in case $`\tau =|\xi _{\alpha \beta }+xy|`$ from (2.6) the contradiction
$$4(R_\phi +C^{}\epsilon ^1)=t_0t\tau 2R_\phi +|\xi _{\alpha \beta }|2R_\phi +C^{}\epsilon ^1.$$
This shows the term in (6.8) is identically zero for $`t[t_0,T\epsilon ^{3/2}]`$, and thus $`D_3(t)=0`$ for $`t[t_0,T\epsilon ^{3/2}]`$. Together with (6.6) and (6.7) this completes the proof of Lemma 3.2(b). $`\mathrm{}`$
Acknowledgement: We are grateful to A. Komech for discussions. HS thanks G. Schäfer for useful hints on post-Newtonian corrections in general relativity and for insisting on (1.9). |
warning/0001/cond-mat0001249.html | ar5iv | text | # The Magnetic Excitation Spectrum and Thermodynamics of High-𝑇_𝑐 Superconductors
## Abstract
Inelastic neutron scattering was used to study the wavevector- and frequency-dependent magnetic fluctuations in single crystals of superconducting YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6+x</sub>. The spectra contain several important features, including a gap in the superconducting state, a pseudogap in the normal state and the much-discussed resonance peak. The appearance of the pseudogap determined from transport and nuclear resonance coincides with formation of the resonance in the magnetic excitations. The exchange energy associated with the resonance has the temperature and doping dependences as well as the magnitude to describe approximately the electronic specific heat near the superconducting transition temperature ($`T_c`$).
The parent compounds of the high-transition-temperature (high-$`T_c`$) cuprate superconductors are antiferromagnetic (AF) insulators characterised by a simple doubling of the crystallographic unit cells in the CuO<sub>2</sub> planes. When holes are doped into these planes, the long-range AF ordered phase disappears and the lamellar copper oxide materials become metallic and superconducting with persistent short-range AF spin correlations (fluctuations). Although spin fluctuations in cuprate superconductors are observed for materials such as YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6+x</sub> \[denote as (123)O<sub>6+x</sub>\] at all hole doping levels, $`x`$ , the role of such fluctuations in the pairing mechanism and superconductivity is still a subject of controversy . The most prominent feature in the fluctuations for (123)O<sub>6+x</sub> is a sharp resonance which for highly doped compositions appears below the superconducting transition temperature $`T_c`$ at an energy of 41 meV . We show that the temperature-dependent resonance intensity is correlated with the electronic part of the specific heat of (123)O<sub>6+x</sub>, $`C^{el}(x,T)`$ , including the extended fluctuation regime (“pseudogap phase”), and that the pseudogap temperature $`T^{}`$ determined from the onset of the resonance intensity agrees with results of transport and nuclear magnetic resonance (NMR) techniques. By making absolute intensity measurements of the magnetic resonance, $`S_{res}(q,\omega )`$, we estimate the contribution of its exchange energy to the $`C^{el}(x,T)`$ anomaly around $`T_c`$. We find that the temperature and doping dependence of the resonance exchange energy can account for the $`C^{el}(x,T)`$ anomaly , suggesting that a large part of this anomaly is due to spin fluctuations.
Our experiments were performed using triple-axis spectrometers at the High-Flux Isotope Reactor (HFIR) of the Oak Ridge National Laboratory and the MARI direct geometry chopper spectrometer at the ISIS pulsed spallation neutron source of the Rutherford Appleton Laboratory . The momentum transfer $`(q_x,q_y,q_z)`$ is measured in units of Å<sup>-1</sup> and reciprocal space positions are specified in reciprocal lattice units (rlu) $`(h,k,l)=(q_xa/2\pi ,q_yb/2\pi ,q_zc/2\pi )`$, where $`a`$, $`b`$ ($`a`$), and $`c`$ are the lattice parameters of the orthorhombic unit cell of (123)O<sub>6+x</sub>. For this study, we prepared four single-crystal samples by the melt texture growth technique . Subsequent annealing resulted in oxygen stoichiometries of $`x=0.6`$, 0.7, 0.8, 0.93 with superconducting transitions at $`T_c=62.7`$, 74, 82, 92.5 K, respectively.
The complete magnetic excitation spectra (Fig. 1) are obtained by integration of the magnetic neutron scattering over the two-dimensional Brillouin zone (BZ) of (123)O<sub>6+x</sub> for $`x=0.6`$ at several temperatures . Because the CuO<sub>2</sub> planes in (123)O<sub>6+x</sub> actually appear in coupled bilayers, magnetic fluctuations that are in-phase (acoustic or $`\chi _{ac}^{\prime \prime }`$) or out-of phase (optical or $`\chi _{op}^{\prime \prime }`$) with respect to the neighboring plane will have different spectra . That the optical excitations occur at higher energies than the acoustic excitations, and are characterized by a substantial gap even in the normal state at 80 K, demonstrates that the bilayer coupling remains antiferromagnetic on going from the insulating parent to the superconducting metal. Upon lowering the temperature from 290 K to 35 K, the most dramatic change in the local-susceptibility $`\chi ^{\prime \prime }(\omega )`$ for energies less than 100 meV is the appearance of a resonance at 34 meV in the acoustic mode as shown in the shaded areas of Fig. 1. For $`x=0.6`$, the resonance is clearly present not only below $`T_c`$ but also at 80 K, albeit in reduced form. Beyond the resonance, there is a broad feature extending at least to 220 meV, with a maximum at approximately 75 meV. This broad feature is reminiscent of the continuum, peaked at around 20 meV, seen in the single layer compound La<sub>1.86</sub>Sr<sub>0.14</sub>CuO<sub>4</sub> . Well below the resonance peak, scattering is suppressed and a true spin-gap with a value of about 20 meV develops in the superconducting state (Fig. 1A). At room temperature, the optical and acoustic spectra (Figs. 1C and 1F) show no maxima over the range of our measurements.
Because the most dramatic change in the magnetic spectrum is the change in the resonance intensity, we focus on the temperature and composition dependence of the resonance. The detailed momentum and frequency dependence of $`S_{res}(q,\omega )`$ obtained at HFIR at various temperatures around the resonance energy for (123)O<sub>6.6</sub> (Fig. 2) show that in the low temperature superconducting state (Fig. 2A), the spectrum is dominated by the resonance at 34 meV. At temperatures just above $`T_c`$ (Fig. 2B), the resonance broadens and decreases in intensity, consistent with the ISIS data of Fig. 1. The resonance peak intensity appears to shift smoothly toward higher energies in the normal state with increasing temperature. A signature of the resonance remains at 125 K and 150 K, but essentially vanishes at 200 K. Previous temperature dependent measurements for underdoped (123)O<sub>6+x</sub> indicate that the resonance peak intensity changes its characteristics at $`T_c`$ . While the intensity at the resonance energy has a noticeable upturn upon cooling through $`T_c`$, the scattering at frequencies above and below the resonance decrease below $`T_c`$ . This behavior can now be understood as arising from the narrowing of the resonance in energy from the normal to the superconducting states.
The temperature dependence of the resonance peak intensity \[$`S_{res}(q,\omega )`$\] for three (123)O<sub>6+x</sub> samples with different doping levels and transition temperatures is shown in Fig. 3, A throught C. For $`x=0.6`$, the temperature dependence of the momentum- and frequency- integrated resonance is found to be different from that of the peak intensity (Fig. 3A). For (123)O<sub>6.8</sub>, similar behavior also occurs, but because the resonance is weaker above $`T_c`$, the counting times required to obtain reliable integrals of the type found for $`x=0.6`$ are prohibitive, and so we are content with simply plotting the peak intensities. Finally, for $`x=0.93`$, no broadening in either energy or wave vector has been identified , and so in this case, the temperature dependence of the integrated spectral weight does actually follow the peak intensity. We define the mean-squared (fluctuating) moment associated with the resonance as $`m_{res}^2=3/(2\pi )d(\mathrm{}\omega )\chi _{res}^{\prime \prime }(\omega )/(1\mathrm{exp}(\mathrm{}\omega /kT))`$, where $`\mathrm{}`$ is the Planck’s constant divided by $`2\pi `$, $`k`$ is the Boltzmann’s constant, $`\chi _{res}^{\prime \prime }(\omega )`$ is the resonance part of the acoustic spectrum of Fig. 1, and the factor of $`1/2`$ arises from averaging $`\mathrm{sin}^2(q_zd/2)`$ over $`q_z`$. For (123)O<sub>6.6</sub> and (123)O<sub>6.93</sub> our measurements are expressed in absolute units obtained by scaling to the low temperature measurements performed on both compounds at ISIS .
The most obvious feature of Figs. 3, A throught C, is that as the doping level and $`T_c`$ decrease, there is a progressively larger pretransitional regime above $`T_c`$. Specifically, for the ideally doped sample ($`x=0.93`$), the onset of the resonance occurs at a temperature $`T^{}`$ which almost coincides with $`T_c`$. For $`x=0.8`$ and 0.6, $`T^{}`$ increases to approximately $`115\pm 15`$ and $`150\pm 20`$ K, respectively, even while both $`T_c`$ and the resonance energy itself are reduced. Thus, the weight of the temperature-dependent resonance joins the long list of properties which show pretransitional behavior in suboptimally doped (123)O<sub>6+x</sub>. The cross-over temperature $`T^{}`$ (Fig. 4A) coincides with the temperatures below which the temperature derivatives of the electrical resistivity $`d\rho (T)/dT`$ and the Cu nuclear $`(T_1T)^1`$ relaxation rate reaches a broad maxima. The anomalies occurring at $`T^{}`$ are generally associated with the opening of a pseudogap in the low-energy spin excitation spectrum, a supposition also supported by neutron scattering data such as those in Fig. 1.
According to thermodynamics, a metal undergoes a transition into the superconducting state because such a transition can lower its total free energy, $`F`$ . The difference in free energy of the system between the normal state, extrapolated to zero temperature, ($`F_N`$) and the superconducting state ($`F_S`$) is the condensation energy, that is, $`E_C=(F_NF_S)_{T=0}`$. In principle, the free energy of a system, and therefore $`E_C`$, can be derived from the temperature dependence of $`C^{el}(x,T)`$ . For (123)O<sub>6+x</sub>, $`C^{el}(x,T)`$ has been measured by Loram et al. . The key features of the corresponding data (Fig. 3E) are a sharp jump at $`T_c`$ for the optimally doped (123)O<sub>6.92</sub>, which becomes much suppressed upon reduction of the oxygen content. Although there is less entropy released at $`T_c`$, more seems to be released at temperatures above $`T_c`$, as $`x`$ is reduced below its optimal value. Thus, the specific heat tracks the temperature derivative (Fig. 3D) of the spectral weight of the resonance. Just as for the specific heat, decreasing doping reduces the maximum in $`dm_{res}^2/dT`$ at $`T_c`$ and introduces progressively broader high-temperature tails.
In the $`t`$-$`J`$ model , the Hamiltonian of the system consists of a nearest-neighbor hopping ($`t`$) term accounting for the kinetic energy ($`E_K`$) of the carriers and a magnetic exchange ($`J`$) term of the type describing the insulating antiferromagnetic parent compound . Magnetic fluctuations with a particular wavevector $`q`$ contribute to the exchange energy ($`E_J`$) via a product with $`J(q)`$, the Fourier transform of the exchange interactions . Although magnetic spectra as complicated as those in Fig. 1 will be difficult to describe within such a simple model, the exchange part of the $`t`$-$`J`$ Hamiltonian does provide a convenient way to quantitatively estimate the effect of magnetic fluctuations on the thermodynamic properties of high-$`T_c`$ superconductors. For (123)O<sub>6+x</sub>, which has two coupled CuO<sub>2</sub> layers per unit cell, the exchange interactions consist of the nearest-neighbor spin-spin coupling $`J`$ in the same CuO<sub>2</sub> plane and the coupling $`J_{}`$ between two CuO<sub>2</sub> planes within the unit cell. Since $`J`$ is much larger than $`J_{}`$ , the exchange energy is :
$`E_J{\displaystyle \frac{3}{(g\mu _\mathrm{B})^2}}({\displaystyle \frac{a}{2\pi }})^2{\displaystyle \frac{d(\mathrm{}\omega )}{\pi }𝑑q_x𝑑q_y\frac{1}{2}J[\mathrm{cos}(q_xa)+\mathrm{cos}(q_yb)]\frac{[\chi _{ac}^{\prime \prime }(q_x,q_y,\omega )+\chi _{op}^{\prime \prime }(q_x,q_y,\omega )]}{1\mathrm{exp}(\mathrm{}\omega /kT)}}`$ (1)
where $`g`$ is the Lande factor ($`2`$) and $`\mu _B`$ is the Bohr magneton. Thus, Eq. 1 gives the contribution of the exchange energy to the total energy. Since the heat capacity is the temperature derivative of the total energy, we can estimate how the magnetic fluctuations responsible for the resonance affect the thermodynamic properties of (123)O<sub>6+x</sub> . As the resonance grows, more spins fluctuating at the resonance frequency are correlated antiferromagnetically, and the exchange energy is correspondingly reduced. Assuming that the spins accounting for the resonance are not spatially correlated at high temperatures and therefore do not contribute to Eq. 1, the contribution of the exchange energy of the resonance ($`E_J^{res}`$) to the specific heat, or $`C_J^{res}(x,T)`$, is then
$$C_J^{res}(x,T)\frac{dE_J^{res}}{dT}=\frac{3}{4\mu _\mathrm{B}^2}J\frac{d}{dT}\left(\frac{d(\mathrm{}\omega )}{\pi }\frac{\chi _{res}^{\prime \prime }(\omega )}{1\mathrm{exp}(\mathrm{}\omega /kT)}\right)=\frac{J}{2\mu _B^2}\frac{dm_{res}^2}{dT}(2).$$
Therefore, $`C_J^{res}(x,T)`$ is approximately proportional to the temperature derivative of the spectral weight of the resonance, as we already discovered from comparison of our data with the measured specific heat. In deriving Eq. 2 from Eq. 1 we relied on the fact that the resonance is sharply peaked around $`(\pi /a,\pi /b)`$. The contribution of the resonance exchange energy to the heat capacity calculated using Eq. 2 with the measured exchange coupling $`J=125\pm 20`$ meV for (123)O<sub>6.6</sub> is shown in Fig. 3D (right hand scale) and can be compared to the measured $`C^{el}(x,T)`$ of Loram et al. in Fig. 3E . The model calculation with no adjustable parameters shows that the resonance can provide enough temperature-dependent exchange energy to yield the superconducting anomaly in the specific heat . This suggests that a large part of the specific heat near $`T_c`$ and its associated entropy are due to spin fluctuations.
Clearly, the measured specific heat includes contributions other than that due to the exchange energy of the resonance ($`E_J^{res}`$). While we included the most obvious temperature dependent feature in our calculation, the temperature evolution of the remainder of the spin excitation spectrum is neglected. Most notably we ignored the formation of the spin pseudogap below $`T^{}`$ and spin gap below $`T_c`$, both of which could increase $`E_J`$. Further, we have not considered the contribution of $`E_K`$ to $`C^{el}(x,T)`$. For superconductors, $`E_K`$ is expected to increase from the normal to the superconducting state . Both this effect and spin (pseudo)gap formation could partially compensate for the reduction of $`E_J`$ due to the appearance of the resonance. Such compensation might explain why the area of the calculated $`C_J^{res}(x,T)`$ anomaly is considerably larger than that of the $`C^{el}(x,T)`$ measurement.
In the next few years, the contributions missing from the present analysis should become calculable using better neutron and optical conductivity data, which would provide the necessary information on the temperature-dependent magnetic exchange and kinetic energy terms, respectively. It would also be of considerable interest to make further model calculations to estimate the change in the resonance exchange energy between the normal and superconducting states at zero temperature and hence the resonance exchange energy contribution to the condensation energy . Unfortunately, this comparison is difficult because the normal state $`S(q,\omega )`$ must be extrapolated to zero temperature. This cannot be done reliably with the present data and presents a challenge for the future.
In summary, we show that a pseudogap regime bounded by a temperature $`T^{}>T_c`$ determined from transport, NMR, and thermodynamic measurements can be associated with the appearance of the resonance peak and the suppression of the low frequency response in the magnetic excitation spectrum. Our complete spectra allow us to establish the weight of the resonance relative to that of other spectral features. A simple calculation of the exchange energy using the measured temperature dependence of the resonance shows that spin fluctuations can account for a large part of the electronic specific heat near the superconducting transition. |
warning/0001/cs0001017.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Curve/curve intersection is one of the fundamental problems of computational geometry. At the present time there exist several different approaches to this problem but the endeavor is to avoid difficulties in calculations which are mainly results of polynomial representation rational higher degree curves. It can be done in certain cases by using the relief perspective and Bezier clipping.
Bezier clipping, in the context of plane curve intersection in this paper, see e.g. \[Nishi90\] \[Nishi98\], is an interactive method which takes advantages of the convex hull property of Bezier curves. Regions of one curve which are guaranteed do not intersect a second curve can be identified and subdivided away.
Relief perspective is a mapping of space into space in order to correspond conditions of the human seeing. Images of all objects under the relief perspective are located in space between two parallel planes.
## 2 Relief perspective
The relief perspective is a special case of perspective collineation of the extended Euclidean space $`\overline{𝔼}_3`$. We remind that a collineation is a bijective mapping $`\phi :\overline{𝔼}_3\overline{𝔼}_3`$ that preserves collinearity of points. If there exists a plane $`\omega `$ such that $`\phi (X)=X`$ $`X\omega `$$`\phi `$ is a perspective collineation (for more details, see e.g. \[Bus53\] \[Cizm84\]). Then there exists a point $`O`$ (called the centre of perspective collineation) such that $`\phi (O)=O`$ and $`\phi (\alpha )=\alpha `$ $`\alpha ;O\alpha `$.
The relief perspective is a perspective collineation including some additional conditions:
* in order to produce correct images of 3D objects, it is important to respect necessary conditions of the human seeing. As in linear perspective, see e.g. \[Cenek59\], also in the relief perspective we suppose, that objects are located inside viewing circular cone. The cone has a vertex in the eye (the centre of projection) and the distance from the eye to the objects has to be at least 25 cm.
* the image plane $`\omega `$ (the set of invariant points ) does not contain the centre $`O`$ (i.e. $`\phi `$ is a homology) and determine elements of the mapping $`\phi `$ are: the centre $`O`$, the image plane $`\sigma `$ (the set of invariant points) and the vanishing plane $`\omega ^r`$ (the image of the plane at infinity).
* no objects are ideal, i.e. objects and their images in the relief perspective have not ideal points. In notions of $`𝔼_3`$ the image plane has to be placed between the point $`O`$ and the plane $`\omega ^r`$. The mapped object, that image we want to construct, is located in the semi-space, that it is opposite to the semi-space $`\stackrel{}{\sigma O}`$ ("behind the image plane $`\sigma `$").
Let $`\phi `$ be a relief perspective of space $`\overline{𝔼}_3`$. Let denote $`\overline{𝔼}_3`$ be a preimage space and $`\overline{𝔼}_3^r`$ be an image space ($`\overline{𝔼}_3`$ with upper index "r") such that $`\phi :\overline{𝔼}_3\overline{𝔼}_3^r`$. The image of the object $`𝒰\overline{𝔼}_3`$ under the relief perspective $`\phi `$ is called the relief of the object $`𝒰`$ and is denoted analogically $`𝒰^r`$. The relief perspective is given by the centre $`O`$, the image plane $`\sigma `$ and ordered pair of different points $`A,A^r=\phi (A)(A,\phi (A)O)`$ such that $`A,A^r\sigma `$ and $`O,A,A^r`$ are collinear.
Note therefore all mapped objects $`𝒰`$ and their images $`𝒰^r`$ under the relief perspective have no ideal points, we can study them in $`𝔼_3`$ or $`\overline{𝔼}_3`$ using affine and projective methods as well. It is also clear that the restriction $`\phi /𝒰`$ of the relief perspective $`\phi `$, which is a bijective mapping, is again a bijection.
## 3 Representation of the relief perspective in an analytical form
Suppose $`X`$ and $`X^{}`$ are two different points of $`𝔼_3\overline{𝔼}_3`$ with coordinates $`[x,y,z]`$ and $`[x^{},y^{},z^{}]`$ respectively. If $`\phi `$ is a collineation of the space $`\overline{𝔼}_3`$ we use the following equations:
$`x^{}={\displaystyle \frac{a_{11}x+a_{12}y+a_{13}z+a_{14}}{a_{41}x+a_{42}y+a_{43}z+a_{44}}}`$
$`y^{}={\displaystyle \frac{a_{21}x+a_{22}y+a_{23}z+a_{24}}{a_{41}x+a_{42}y+a_{43}z+a_{44}}}`$ (1)
$`z^{}={\displaystyle \frac{a_{31}x+a_{32}y+a_{33}z+a_{34}}{a_{41}x+a_{42}y+a_{43}z+a_{44}}},`$
where $`a_{i,j}`$ $`(i,j\{\mathrm{1..4}\})`$ are real numbers and $`det(a_{ij})0`$. In equations (3 Representation of the relief perspective in an analytical form) are used non-homogenic coordinates, which correspond to a restriction $`\overline{\phi }=\phi /\{\overline{𝔼}_3\alpha \}`$, where $`\alpha `$ is a plane $`a_{41}x_1+a_{42}x_2+a_{43}x_3+a_{44}x_0=0`$.
In order to obtain simple mapping equations for relief perspective from (3 Representation of the relief perspective in an analytical form) we assume, that the centre $`O`$ is a point with coordinates $`[0,0,0]`$ and $`\sigma `$ is a plane $`z1=0`$. In this special case we obtain the convenient representation of the relief perspective in the following form
$`x^{}={\displaystyle \frac{(1+k)x}{z+k}}`$
$`y^{}={\displaystyle \frac{(1+k)y}{z+k}}`$ (2)
$`z^{}={\displaystyle \frac{(1+k)z}{z+k}}`$
where the span $`k^+\{1\}`$ (the geometric representation of the parameter $`k`$ will be explained in the next lines).
## 4 Some properties of the relief perspective
* the vanishing plane $`\omega ^r1+k`$ and the neutral plane $`\nu k`$ (the preimage of the plane at infinity) are parallel and for distances $`O,\nu `$ and $`\omega ^r,\sigma `$ we have
$$O\nu =\omega ^r\sigma =k$$
where the parameter $`k`$ represents the span
* if a plane $`\alpha zc(c0)`$ is parallel to the image plane $`(\alpha \sigma ,\nu )`$, then the relief of the plane $`\alpha `$ is the plane $`\alpha ^r\frac{(1+k)c}{c+k}`$ and these planes are parallel (it is obvious that $`\alpha \alpha ^r\sigma `$)
* the relief of a point placed in the semi-space, that is opposite to the semi-space $`\stackrel{}{\sigma O}`$, is a point in a part of space with boundary planes $`z=1`$ and $`z=1+k`$ (in an intersection of spaces $`z>1`$ and $`z<1+k`$) (Fig. 4 Some properties of the relief perspective)
* the orthographic projection of the object $`𝒰^r`$ (the relief of the object $`𝒰`$) onto the image plane $`\sigma `$ is the central projection of the object $`𝒰`$ from the point $`S`$, where $`Sl\nu (l;Oll\sigma )`$
## 5 Rational Bezier curves and relief perspective
The relief perspective is a mapping of the extended Euclidean space $`\overline{𝔼}_3`$. Let $`V_i,i\{0,1,\mathrm{},n\}`$ be given points, $`[x_i,y_i]`$ be their coordinates of the space $`\overline{𝔼}_2`$ and positive real numbers $`w_i,i\{0,1,\mathrm{},n\}`$ be their weights. By these elements and Bernstein polynomials $`B_i^n(t)`$, see e.g. \[Farin93\], is defined a n-th planar rational Bezier curve
$`𝐏(t)=[{\displaystyle \frac{\underset{i=0}{\overset{n}{}}x_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}w_iB_i^n(t)}},{\displaystyle \frac{\underset{i=0}{\overset{n}{}}y_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}w_iB_i^n(t)}}],t0,1`$ (3)
and also the space nonrational Bezier curve
$`\overline{𝐏}(t)=[{\displaystyle \underset{i=0}{\overset{n}{}}}x_iw_iB_i^n(t),{\displaystyle \underset{i=0}{\overset{n}{}}}y_iw_iB_i^n(t),{\displaystyle \underset{i=0}{\overset{n}{}}}w_iB_i^n(t)],t0,1`$ (4)
with control points
$$V_i=[x_iw_i,y_iw_i,w_i].$$
(5)
Let $`\phi :\overline{𝔼}_3\overline{𝔼}_3^r`$ be the relief perspective of the space $`\overline{𝔼}_3`$ and $`X^r[x,y,z]`$ $`(X^r\sigma )`$ be the relief of the point $`X[a,b,c]\overline{𝔼}_3`$. From the equations (3 Representation of the relief perspective in an analytical form) we get
$$x=\frac{(1+k)a}{c+k},y=\frac{(1+k)b}{c+k},z=\frac{(1+k)c}{c+k}$$
(6)
and according to the relation $`z=\frac{(1+k)c}{c+k}`$ we obtain
$`c+k={\displaystyle \frac{k(1+k)}{1+kz}}`$
From the above results coordinates $`[a,b,c]`$ of the point $`X`$, that is the preimage of the relief $`X^r[x,y,z]`$, are computed according to $`a=\frac{kx}{1+kz}`$, $`b=\frac{ky}{1+kz}`$, $`c=\frac{kz}{1+kz}`$.
Lemma Let $`\phi :\overline{𝔼}_3\overline{𝔼}_3^r`$ be relief perspective of the space $`\overline{𝔼}_3`$. The point $`[x,y,z]`$ is the relief of the point $`[\frac{kx}{1+kz}`$,$`\frac{ky}{1+kz}`$,$`\frac{kz}{1+kz}]`$.
By given planar points $`V_i`$ and their weights $`w_i`$, where $`1+k>w_ii`$, we can define in $`\overline{𝔼}_3`$ points
$$W_i=[\frac{kx_iw_i}{1+kw_i},\frac{ky_iw_i}{1+kw_i},\frac{kw_i}{1+kw_i}]i\{0,1,\mathrm{},n\}$$
(7)
With respect to the lemma above it is obvious, that the reliefs of the points $`W_i`$ are represented by the points (5). Using the points $`V_i`$, the space nonrational Bezier curve $`\overline{𝐏}(t)`$ of the form (4) is defined and we notice, that $`\overline{𝐏}(t)`$ under mapping $`\phi `$ is an image of the curve
$$𝐐(t)=[\frac{\underset{i=0}{\overset{n}{}}kx_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(1+kw_i)B_i^n(t)},\frac{\underset{i=0}{\overset{n}{}}ky_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(1+kw_i)B_i^n(t)},\frac{\underset{i=0}{\overset{n}{}}kw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(1+kw_i)B_i^n(t)}]$$
(8)
This curve Q(t) is the space rational Bezier curve defined by control points of the form (7) and their weights $`\mathrm{\Omega }_i`$, that are computed according to $`\mathrm{\Omega }_i=1+kw_i`$ because of
$$\frac{kx_iw_i}{1+kw_i}\mathrm{\Omega }_i=kx_iw_i,\frac{ky_iw_i}{1+kw_i}\mathrm{\Omega }_i=ky_iw_i,\frac{kw_i}{1+kw_i}\mathrm{\Omega }_i=kw_i$$
Applying all these results we obtain the following theorem
Theorem The curve $`\overline{𝐏}`$(t) is a relief of the curve Q(t).
In order to get more information about Bezier curves and the relief perspective, let us assume having a central projection with the centre $`O[0,0,0]`$ and the plane $`z1=0`$. In this case the planar rational curve $`𝐏(t)`$ of the form (3) is the image of the space rational Bezier curve $`𝐐(t)`$ defined by (8).
Let the space nonrational Bezier curve $`\overline{𝐏}(t)`$ expressed by (4) be given. What is the relief of this curve? We know $`z+k=\underset{i=0}{\overset{n}{}}(w_i+k)B_i^n(t)`$ and now the relief $`\overline{𝐏}^𝐫(t)`$ of the given curve $`𝐏^𝐫(t)`$ can be written as
$$\overline{𝐏}^𝐫(t)=[\frac{\underset{i=0}{\overset{n}{}}(1+k)x_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(w_i+k)B_i^n(t)},\frac{\underset{i=0}{\overset{n}{}}(1+k)y_iw_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(w_i+k)B_i^n(t)},\frac{\underset{i=0}{\overset{n}{}}(1+k)w_iB_i^n(t)}{\underset{i=0}{\overset{n}{}}(w_i+k)B_i^n(t)}]$$
$`t0,1,`$ where control vertices are points expressed as follow
$$\overline{V}_i^r=[\frac{(1+k)x_iw_i}{w_i+k},\frac{(1+k)y_iw_i}{w_i+k},\frac{(1+k)w_i}{w_i+k}]$$
and weights $`\overline{\mathrm{\Omega }}_i`$ are computed by $`\overline{\mathrm{\Omega }}_i=w_i+k`$.
## 6 Intersection of space rational Bezier curves
Let $`𝐏(t)`$, $`𝐐(u)`$ be space rational Bezier curves of the form (8). The aim is to find their intersection.
The solution to this problem can be found using already known facts. According to the theorem, which was formed in the previous section, reliefs of space rational Bezier curves of the form (8) are space nonrational Bezier curves of the form (4) (in Fig. 6 Intersection of space rational Bezier curves Bezier curves of the form (8) and their reliefs for $`k=2`$ are shown).
The central projections of these nonrational curves from the point $`[0,0,0]`$ onto the plane $`z1=0`$ are planar rational Bezier curves defined by (3). Their intersection is possible to find specifically by using the method of Bezier clipping. This method specifies values of the parameter $`t0,1`$ of the curve $`𝐏^𝐬(t)`$, respectively the parameter $`u0,1`$ of the curve $`𝐐^𝐬(u)`$, that correspond to the common point $`𝐑^𝐬`$ (the upper indexes "r" and "s" denote the reliefs and central projections of curves or points). This point is the intersection of both curves . The preimage of the point $`𝐑^𝐬`$ in the central projection is the space point, which is the relief of the intersection of the space rational Bezier curves $`𝐏(t)`$ and $`𝐐(u)`$.
The following scheme shows a whole proces of finding the intersection of the $`𝐏(t)`$ and $`𝐐(u)`$ curves:
$$\begin{array}{ccc}𝐏(t)\stackrel{\phi }{}𝐏^𝐫(t)\stackrel{\psi }{}𝐏^𝐬(t)\hfill & & \\ 𝐐(u)\stackrel{\phi }{}𝐐^𝐫(u)\stackrel{\psi }{}𝐐^𝐬(u)\hfill & & \end{array}\}\stackrel{\text{B. clipping}}{}𝐑^𝐬$$
$$𝐑^𝐬=𝐏^𝐬(t)𝐐^𝐬(u)\stackrel{\psi ^1}{}𝐑^𝐫\stackrel{\phi ^1}{}𝐑$$
$$𝐑=𝐏(t)𝐐(u)$$
$`\phi `$ – relief perspektive $`\phi :\overline{𝔼}^3\overline{𝔼}^3`$
$`\psi `$ – central projection $`\psi :\overline{𝔼}^3\overline{𝔼}^2`$.
The central projection $`\psi `$ is not a bijective mapping and in case that the intersection of curves $`𝐏^𝐬(t)`$ and $`𝐐^𝐬(u)`$ exists (in opposite case the given curves $`𝐏(t)`$ and $`𝐐(u)`$ certainly do not intersect) the intersection of curves $`𝐏(t)`$ and $`𝐐(u)`$ does not have to exist. This situation occurs when preimages of point $`𝐑^𝐬`$ on the curves $`𝐏^𝐫(t)`$ and $`𝐐^𝐫(u)`$ in the central projection are different.
## 7 Conclusions and future work
The relations between Bezier curves and the relief perspective have been described. The necessary and sufficient conditions for expressing the space nonrational Bezier curve as the relief of the space rational Bezier curve have been formed. We have shown that to the planar rational curve of the form (3), defined by planar points $`V_i\overline{𝔼}_2`$ and their weights $`w_i`$, it is possible to assign a class of the space curves by the relief perspective. One of them is nonrational curve defined by (4) and two are rational curves defined by (8) and (5 Rational Bezier curves and relief perspective).
The described method can be used as a direction for application how to find the intersection of the space rational curves of the form (8). It is possible to express every polynomial curve in Bezier’s representation and due to this representation the method can be applied to all polynomial curves after modifications (e.g. spline curves which are considered as curves consist of Bezier segments) for solution to the curve/curve or curve/line intersection problems.
Despite the author’s attempt he did not succeed in finding similar comparable published methods on website. In addition to this he is not able to compare his results with any other. Obviously author’s knowledge is limited and he would appreciate to get information about any other similar method.
In the future work we want to extend possibilities of the relief perspective in geometric modeling. Our aim is to find an answer whether any polynomial 3D curve can be converted to the space curve of the form (8). |
warning/0001/nucl-th0001058.html | ar5iv | text | # Electroproduction of the 𝑑^∗ dibaryon
## I Introduction
Given the availability of excellent electron beams, the electroproduction of dibaryon resonances represents a promising way to look for these long-sought objects . In assessing the practical prospects of such experiments, one needs to begin with rough estimates of the electroproduction cross sections. This has to be done separately for each dibaryon candidate of interest, since each candidate may involve unique theoretical issues.
One candidate dibaryon that has been discussed recently is the di-delta $`d^{}`$ of quantum numbers $`J^\pi T=3^+0`$ . This dibaryon is a six-quark state which at large inter-baryon separations may be visualized as a pair of $`\mathrm{\Delta }`$’s. Preliminary studies of its electroproduction from deuteron targets have recently been made by Qing and by Sun , both of Nanjing University. They find that the Kroll-Ruderman $`\gamma d\mathrm{\Delta }\mathrm{\Delta }`$ production process does not contribute because of the special isospin structure of $`d`$ and $`d^{}`$. They concentrate instead on another process important in the electroproduction of two pions from proton and deuteron based on the electroexcitation of a nucleon isobar $`N^{}(1520)`$ in the intermediate state, namely $`\gamma dNN(1520)\mathrm{\Delta }\mathrm{\Delta }`$. This mechanism can be interpreted as the production of $`d^{}`$ through its $`NN^{}`$ component.
However, these preliminary results for the calculated differential cross sections turn out to be very small at small momentum transfers (say $`<500`$ MeV/$`c`$), almost two orders of magnitude smaller than a simple estimate made by me using a simple picture of inelastic scattering from the $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ of the deuteron, as described in diagram a of Fig. 1. These results seem to suggest that $`d^{}`$ is more easily produced from the deuteron at these small momentum transfers through elastic electron-baryon (eB) scatterings rather than inelastic excitations of the struck baryon.
It is worth noting in connection with diagram 1a that if the $`d^{}`$ is visualized as a $`{}_{}{}^{7}S_{3}^{}`$ didelta, there is no production from the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^3S_1)`$ component in the lowest order. This is because the electromagnetic operators are of rank only 0 or 1 in intrinsic spin. Hence electroproduction in this simple lowest-order picture is possible only from the $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component of the deuteron.
In addition to diagram 1a, or equivalently its perturbative part diagram 1b, $`d^{}`$ can also be electroproduced through the $`NN(^3D_3^3G_3)`$ component of $`d^{}`$, as described in diagram 1c. Lomon has looked for a $`d^{}`$ resonance in these channels by studying the energy dependence of their $`NN`$ phase parameters . These $`NN(^3D_3^3G_3)`$ channels are in fact the dominant decay channels of $`d^{}`$ . He has found that these phase parameters are consistent with the absence of any dibaryon resonance with a width exceeding 1 (2) MeV near 2.1 (2.25) GeV.
A much stricter upper limit on the width of an np dibaryon resonance has been obtained experimentally some time ago by Lisowski et al. . They measured the total $`np`$ cross section at c.m. energies 2.00 to 2.23 GeV at high energy resolution, namely about 1.4 MeV at 2.11 GeV. No evidence was seen for narrow resonances with areas greater than 5 mb-MeV. At 2.1 GeV where the $`np`$ $`I=0`$ total cross section is 33 mb, a totally elastic $`{}_{}{}^{3}D_{3}^{}`$ resonance would have a maximum unitarity-limited cross section of $`\sigma _{\mathrm{max}}=39`$ mb. Its energy-integrated area is $`(\pi /2)\sigma _{\mathrm{max}}\mathrm{\Gamma }`$ for a pure Breit-Wigner shape, where $`\mathrm{\Gamma }`$ is its total decay width. The assumption of a pure Breit-Wigner resonance should be a very good one, because the $`NN`$ $`{}_{}{}^{3}D_{3}^{}`$ phase shift at the total energy of 2.1 GeV is only 4 . The experimental bound of 5 mb-MeV on the integrated area translates to an upper bound of only 0.08 MeV for the width of any resonance that has escaped detection in this experiment.
The theoretical decay width of $`d^{}`$ has been estimated in . If the $`d^{}`$ is taken to be a bound system of two finite-sized $`\mathrm{\Delta }`$’s, describable by a nonrelativistic quark model, the decay has been estimated to be about 10 MeV when its mass is 2.1 GeV. The result is, however, sensitive to both dynamical input and rescattering corrections; it could be as large as 40 MeV for a realistic potential treated perturbatively without any rescatterinfg correction, as I shall show in more detail in this paper.
It has been argued in that if the dibaryon is made up of six delocalized quarks instead of two separated $`\mathrm{\Delta }`$’s, its decay width could well be smaller by an order of magnitude, say 1 MeV. This suggestions remains to be confirmed by a detailed calculation. A reduction to 0.1 MeV would be much harder to realize. Furthermore, if the estimated decay width has to be reduced for any reason, its estimated production cross sections, including the electroproduction cross section estimated here using a didelta model, must also be reduced correspondingly.
Thus the case for a $`d^{}`$ resonance near 2.1 GeV does not appear promising. However, the $`d^{}`$ could have a mass higher than expected if the dynamnics is different from that described by the delocalization and color-screening model. Hence it is still of interest to study the electroproduction cross section for $`d^{}`$, should such a resonance exist.
It is obvious that in addition to diagram 1a or b, one should also included the process shown in diagram 1c, where the $`d^{}`$ is produced through the $`NN(^3D_3)`$ channel, the $`NN(^3G_3)`$ contribution having been ignored in this lowet-order picture.
Although these two processes alone do not add up to a quantitative description of the electroproduction, they are of sufficient interest to justify the detailed report given here, including the contributions of convective and magnetization currents. The picture is necessarily very rough, because of the neglect of many components and processes. However, uncertainties about the mass and structure of $`d^{}`$ and about short-distance nuclear dynamics have discouraged me from undertaking a more ambitious calculation at this time.
The paper is organized as follows: The notation used is defined in Sec. II where brief comments relevant to the present calculation are made. Sec. III shows how the calculation is done for diagram 1a, when the needed $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component of the deuteron is already available. The contributions from different reduced matrix elements (RME’s) are briefly discussed.
Sec. IV shows that the inclusion of the $`d^{}(NN^3D_3)`$ contribution via the perturbative process Fig. 1c can be broken down into three steps: (a) a perturbative evaluation of the $`d^{}(NN^3D_3)`$ wave function, (b) a calculation of the $`d^{}`$ decay width for the same input dynamics (taken to be the 28-channel Argonne potential and the Bonn B potential ), and finally (c) the evaluation of the electroproduction cross section itself.
Calculational results are presented in Sec. V, where the production amplitude from diagram 1c is found to be greater than that from diagram 1a by a factor of 2-4 at certain angles or momentum transfers. At a lab angle of $`30^{}`$ and an electron lab energy of 1 GeV, I find a differential production cross section of about 0.05 nb/sr when integrated over the energy loss.
My calculated cross section is smaller by two orders of magnitude from the preliminary value of about 10 nb/sr obtained recently by the Nanjing group (to be called QSW) using a different method of calculation. Part of the discrepancy, accounting for a factor of 5, comes from the fact that QSW has included only $`\pi `$ exchange but not $`\rho `$ exchange. The remaining discrepancy, of more than an order of magnitude, must be due to differences in calculational methods used. For example, I use a principal-value Green function for two nucleons in the $`d^{}(NN^3D_3)`$ wave function, whereas QSW use an outgoing-wave boundary condition. To help in disentangling the discrepancy in the future, I have included many details in the present paper.
The decay width calculated here for the $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ model of $`d^{}`$ is about 40 MeV at $`m^{}=2.1`$ GeV. When used with the experimental upper bound of only 0.08 MeV of any $`I=0,J=3`$ np resonance , the calculated width makes it very unlikely that any undetected $`d^{}`$ has a significant probablity of the $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ configuration.
## II Electroproduction cross section for $`eded^{}`$
Consider the scattering of a relativistic electron beam of lab energy $`ϵ`$ to the lab angle $`\theta `$ by a target that goes from an initial state $`i`$ to a final state $`f`$ as the result of the scattering. The differential cross section in the lab (after integrating over the energy transfer) is known to have the form
$$\frac{\mathrm{d}\sigma _{\mathrm{fi}}}{\mathrm{d}\mathrm{\Omega }}=\sigma _Mf_{\mathrm{rec}}^1[v_LT_{\mathrm{fi}}^L+v_T(T_{\mathrm{fi}}^{\mathrm{el}}+T_{\mathrm{fi}}^{\mathrm{mag}})],$$
(1)
where
$`\sigma _M=\left[{\displaystyle \frac{\alpha \mathrm{cos}(\theta /2)}{2ϵ\mathrm{sin}^2(\theta /2)}}\right]^2`$ (2)
is the Mott differential cross section, $`\alpha `$ is the fine structure constant,
$`f_{\mathrm{rec}}=1+{\displaystyle \frac{2ϵ\mathrm{sin}^2(\theta /2)}{M_{\mathrm{target}}}}`$ (3)
is a target recoil factor,
$`v_L=\left({\displaystyle \frac{Q^2}{q^2}}\right)^2,v_T={\displaystyle \frac{1}{2}}\left({\displaystyle \frac{Q^2}{q^2}}\right)+\mathrm{tan}^2\left({\displaystyle \frac{\theta }{2}}\right)`$ (4)
are the electron kinematical factors that depend on its four-momentum transfer $`Q=KK^{}=(\omega ,𝐪)`$. Here $`K=(ϵ,𝐤)`$ is the four-momentum of the electron in the lab before the scattering, and $`K^{}`$ its four-momentum after the scattering. The notation is that of , with $`Q^20`$.
The target factors are
$`T_{\mathrm{fi}}^L={\displaystyle \frac{4\pi }{2J_i+1}}{\displaystyle \underset{J=0}{\overset{\mathrm{}}{}}}|J_f\widehat{M}_J(q)J_i|^2;`$ (5)
$`T_{\mathrm{fi}}^\alpha ={\displaystyle \frac{4\pi }{2J_i+1}}{\displaystyle \underset{J=1}{\overset{\mathrm{}}{}}}|J_f\widehat{T}_J^\alpha (q)J_i|^2,`$ (6)
where $`\alpha =`$ el or mag. The nuclear reduced matrix elements (RME) $`J_f𝒪J_i`$ that appear are all dimensionless quantities. The operators $`𝒪`$ are the longitudinal Coulomb, and the transverse electric and magnetic, multipole operators that have the following simple forms in momentum space
$`\widehat{M}_J(q)={\displaystyle \frac{(i)^J}{4\pi }}{\displaystyle }`$ $`d\mathrm{\Omega }_𝐪Y_J(\mathrm{\Omega }_𝐪)\rho (𝐪),`$ (7)
$`\widehat{T}_J^{\mathrm{el}}(q)=\sqrt{8\pi }{\displaystyle \underset{\kappa }{}}`$ $`\kappa Y_J1\left\{\begin{array}{ccc}1& 1& 1\\ J& J& \kappa \end{array}\right\}`$ (10)
$`\times {\displaystyle \frac{(i)^J}{4\pi }}{\displaystyle }`$ $`d\mathrm{\Omega }_𝐪[Y_\kappa (\mathrm{\Omega }_𝐪)\widehat{𝐉}(𝐪)]^J,`$ (11)
$`\widehat{T}_J^{\mathrm{mag}}(q)={\displaystyle \frac{(i)^J}{4\pi }}{\displaystyle }`$ $`d\mathrm{\Omega }_𝐪[Y_J(\mathrm{\Omega }_𝐪)\widehat{𝐉}(𝐪)]^J.`$ (12)
Here $`Y_J`$ is a spherical harmonic, $`\rho (𝐪)`$ and $`\widehat{𝐉}(𝐪)`$ are the baryon charge and current density operators.
The current density operator can be separated into convective and magnetization terms:
$`\widehat{𝐉}(𝐪)=\widehat{𝐉}_\mathrm{c}(𝐪)+\widehat{𝐉}_\mathrm{S}(𝐪).`$ (13)
The spin magnetization term $`\widehat{𝐉}_\mathrm{S}=\mathbf{}\times \widehat{𝝁}_\mathrm{S}`$ originates from the dibaryon isoscalar magnetic moment operator $`\widehat{𝝁}_\mathrm{S}=\mu _0\mu _\mathrm{s}\widehat{𝐒}`$. Here $`\mu _0`$ is the nuclear magneton, $`\widehat{𝐒}`$ is the total spin angular momentum operator of the nucleus. In this notation, the nucleon magnetic-moment operator is written as $`\widehat{𝝁}(N)=\mu _0[\mu _\mathrm{s}(N)\widehat{𝐒}+\mu _\mathrm{v}(N)\widehat{𝝉}_3\widehat{𝐒}]`$, while the isoscalar part of the operator for $`\mathrm{\Delta }`$ is $`\mu _0\mu _\mathrm{s}(\mathrm{\Delta })\widehat{𝐒}`$. The isoscalar baryon magnetic-moment parameters used are the experimental value
$`\mu _\mathrm{s}(N)=\mu _\mathrm{p}+\mu _\mathrm{n}=0.880`$ (14)
for the nucleon, and the theoretical value from the non-relativistic quark model
$`\mu _\mathrm{s}(\mathrm{\Delta })=\mu _\mathrm{v}(N)/5=0.94`$ (15)
for the $`\mathrm{\Delta }`$. This quark model actually gives the same value for both baryon isoscalar magnetic moments. I prefer to use the slightly larger value shown for $`\mu _\mathrm{s}(\mathrm{\Delta })`$ obtained from the experimental nucleon isovector magnetic moment of $`\mu _\mathrm{v}(N)=\mu _\mathrm{p}\mu _{rmn}=4.71`$.
As is known, the orbital magnetization term is already contained in $`\widehat{𝐉}_\mathrm{c}`$, and does not have to appear explicitly.
For specific components in the initital and final nuclear states, only a few terms are allowed in the multipole sum shown in Eq. (6) by the triangle rule for angular momenta. For the dibaryon components included in our calculations, the relative BB orbital angular momenta of initial and final nuclear states $`(L_i,L_f)`$ are either (0,2) or (2,0). Then only the $`J=2`$ multipole term appears for the Coulomb target factor.
The situation for the transverse RME’s is slightly more complicated. It is controlled by the spatial RME
$`L^{}=0{\displaystyle 𝑑\mathrm{\Omega }_𝐪[Y_\kappa (\mathrm{\Omega }_𝐪)\overline{𝐩}]^\lambda }L=2`$ (16)
$`=\delta _{\lambda 2}{\displaystyle \underset{\kappa =1,3}{}}f_\kappa (q),`$ (17)
where $`\overline{𝐩}𝐩+(𝐪/4)`$ is the mean of the initial relative BB momentum $`𝐩`$ and its final value $`𝐩+(𝐪/2)`$ after the absorption of a virtual photon of momentum $`𝐪`$. Part of the photon momentum, $`𝐪/2`$, goes into the recoil of the dibaryon. The functions $`f_\kappa (q)`$ are not needed at this point, and will not be given. The important features are that the operator must be a spatial quadrupole operator, and that $`\kappa `$ can only be 1 or 3.
Eq. (17) has the consequence that the only operators contributing to the transvers RME for the dibaryon wave functions used in the present study are (a) the convective and orbital magnetization current terms in $`\widehat{T}_{J=2}^{\mathrm{el}}`$, (b) the spin magnetization current term in $`\widehat{T}_J^{\mathrm{el}}`$, with $`J=`$2 and 3, and (c) the spin magnetization current term in $`\widehat{T}_{J=3}^{\mathrm{mag}}`$.
## III Contributions of the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component
The $`d^{}`$ dibaryon will be treated in this paper as a pure $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ state, using the two-centered Gaussian wave function parametrized in as a sum of Gaussians. This wave function will be specified with other wave functions used in this paper in the Appendix.
The inelastic production of $`d^{}`$ from the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component in the initial state by elastic $`e\mathrm{\Delta }`$ scattering is described by diagram 1a. The Argonne-28 (A28) deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ wave functions are used, expanded in harmonic-oscillator wave functions (HOWF).
However, it is sufficient to give explicit expressions for the RME’s appearing in the differential production cross section only for single-term wave functions such as the single Gaussian
$`\psi _d^{}(𝐩)=N_0^{}e^{p^2/2\beta ^2}`$ (18)
for $`d^{}`$. Here
$`N_0^{}=(\pi \beta ^2)^{3/4}.`$ (19)
For the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component, a one-term form of the HOWF is
$`\psi _d(𝐩,\mathrm{\Delta }\mathrm{\Delta },^7D_1)=\sqrt{P_{\mathrm{\Delta }\mathrm{\Delta }7}}N_2\mathrm{e}^{p^2/2\beta ^2}𝒴_{2\mathrm{m}}(𝐩),`$ (20)
where $`P_{\mathrm{\Delta }\mathrm{\Delta }7}`$ is the $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$-state probability of the deuteron,
$`N_2=\left({\displaystyle \frac{16\pi }{15\beta ^4}}\right)^{1/2}(\pi \beta ^2)^{3/4},`$ (21)
and $`𝒴_{2\mathrm{m}}(𝐩)=p^2Y_{2\mathrm{m}}(\widehat{𝐩})`$ is a solid spherical harmonic. Then the Coulomb RME is
$`d^{}\widehat{M}_2(q)d_7=\sqrt{3/5}d^{},S\widehat{M}_2(q)d_7,D`$ (22)
$`={\displaystyle \frac{\sqrt{3}}{4\pi }}(\lambda q)^2f(q),`$ (23)
where $`d_7`$ stands for the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component;
$`\lambda ={\displaystyle \frac{a^{}}{2(a^{}+b)}};`$ (24)
$`f(q)=\sqrt{P_{\mathrm{\Delta }\mathrm{\Delta }}}N_0^{}N_2\left({\displaystyle \frac{\pi }{a^{}+b}}\right)^{3/2}\mathrm{e}^{cq^2/2};`$ (25)
$`a^{}=(2\beta ^2)^1,b=(2\beta ^2)^1;`$ (26)
$`c=\lambda b+\alpha _0;\alpha _0=r_p^2/3=0.12fm^2.`$ (27)
The $`\alpha _0`$ term takes care of the baryon form factor at the $`\gamma `$-baryon vertex. If the $`\mathrm{\Delta }`$ is assumed for simplicity to have the same size as the nucleon, the same Gaussian form factor appears in all terms of the production amplitude in the nonrelativistic quark model. Additional baryon form factors that should appear are already included in the BB interactions themselves.
The transverse RMS’s can be separated into convective, and spin magnetization terms:
$`d^{}\widehat{T}_{\mathrm{c2}}^{\mathrm{el}}(q)d_7`$ $`=`$ $`{\displaystyle \frac{\sqrt{2}}{4\pi }}{\displaystyle \frac{1}{M_\mathrm{B}}}g_\lambda (q)f(q),`$ (28)
$`d^{}\widehat{T}_{\mu 2}^{\mathrm{el}}(q)d_7`$ $`=`$ $`(\mu _\mathrm{s}(B)\mu _0/\pi )f(q)`$ (30)
$`\times [g_\lambda (q)g_\nu (q)],`$
where $`B=\mathrm{\Delta }`$ in $`M_\mathrm{B}`$ and $`\mu _\mathrm{s}(B)`$, and
$`g_\lambda (q)={\displaystyle \frac{3\lambda q}{2(a^{}+b)}},g_\nu (q)=\lambda ^2\nu q^3,\nu ={\displaystyle \frac{1}{4}}\lambda ;`$ (31)
$`d^{}\widehat{T}_{c3}^{\mathrm{mag}}(q)d_7=d^{}\widehat{T}_{\mathrm{L3}}^{\mathrm{mag}}(q)d_7=0,`$ (32)
$`d^{}\widehat{T}_{\mathrm{S3}}^{\mathrm{mag}}(q)d_7=\sqrt{8/7}{\displaystyle \frac{\mu _0}{4\pi }}\mu _\mathrm{s}(B)g_\nu (q)f(q).`$ (33)
I use a 3-term approximation to the deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ wave function of the Argonne 28-channel potential , denoted below as potential A28. The approximate wave function used is given in the Appendix.
It is convenient to present the calculated results not as angular distributions which contain strong dependence on the electron energy, but as the effective $`T`$ factor
$$T_{\mathrm{eff}}=T_{\mathrm{fi}}^L+(v_T/v_L)(T_{\mathrm{fi}}^{\mathrm{el}}+T_{\mathrm{fi}}^{\mathrm{mag}}).$$
(34)
The Coulomb $`T`$ factor $`T^L`$ is a function only of the three-momentum transfer $`q`$, and is energy-independent in our model. The transverse terms contain the kinetmatical factor $`v_T/v_L`$, but the energy dependence from the angle-dependent term is small except at low electron energies.
The weak energy dependence of these effective $`T`$ factors is explicitly illustrated in Fig. 2 for a $`d^{}`$ mass $`m^{}`$ of 2100 MeV and for three different electron energies. The A28 $`d(\mathrm{\Delta }\mathrm{\Delta }^7D_1)`$ wave function is used. The three-momentum transfer has the energy-independent minimal value of $`q_{\mathrm{min}}238`$ MeV/$`c`$ at this value of $`m^{}`$. The momentum range shown corresponds to an angular range of about $`90^{}`$ ($`15^{}`$, $`2^{}`$) for 1 GeV (4 GeV, 27.5 GeV) electrons.
In the transverse electric $`T`$ factor, the amplitude from the magnetization current is a factor of 2.5 or 3 larger than that from the convective current. This means that the electric $`T`$ factor is larger by roughly an order of magnitude when the contribution from the magnetization current is included.
On the other hand, the transverse magnetic $`T`$ factor is very small at small angles, but has become 1/3 as large as the transverse electric $`T`$ factor at $`q1`$ GeV/$`c`$.
The calculated cross section for diagram 1a is proportional to the probability $`P_{\mathrm{\Delta }\mathrm{\Delta }7}`$ of the $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ component of the deuteron, other things being equal. Besides the A28 model, the Argonne group has constructed a weaker model with $`P_{\mathrm{\Delta }\mathrm{\Delta }7}=0.23\%`$. The coupled-channel models E and F constructed by Dymarz and Khanna have $`P_{\mathrm{\Delta }\mathrm{\Delta }7}0.10.4\%`$, and a total $`P_{\mathrm{\Delta }\mathrm{\Delta }}0.40.5\%`$. The relativistic field-theory model studied by Ivanov et al. gives a total $`P_{\mathrm{\Delta }\mathrm{\Delta }}`$ of only 0.08%.
These theoretical estimates are consistent with the best experimental information on the total $`P_{\mathrm{\Delta }\mathrm{\Delta }}`$, namely that it does not exceed 0.4% at 90% CL from the null result of a bubble chamber search for the spectator $`\mathrm{\Delta }^{++}`$ in a $`\nu d`$ knockout reaction .
## IV Inclusion of the $`NN(^3D_3)`$ component of $`d^{}`$
Although diagram 1c has a structure rather similar to that for diagram 1b, its production amplitudes are much harder to calculate because of three complications: (1) The wave function in the minor component $`d^{}(NN^3D_3)`$ needed in the calculation is not to our best knowledge readily available in the literature. It has to be evaluated ab initio. (2) This wave function is very sensitive to the spreading width $`\mathrm{\Gamma }`$ of $`d^{}(\mathrm{\Delta }\mathrm{\Delta }^3S_3)`$ into $`NN`$ channels. Although the production cross section turns out to be only mildly dependent on $`\mathrm{\Gamma }`$, it is desirable to use the width consistent with the assumed input dynamics in the calculation. This too has to be evaluated. (3) The wave function has sharp kinks near $`p_0`$ where the nucleon energy $`E_N(p_0)`$ has the value $`m^{}/2`$. We find it simpler to perform a one-dimensional integration for the production amplitudes instead of using the harmonic-oscillator expansion described in the last section.
The calculation is thus broken up into three major steps in order to handle the three complications listed above.
### A Perturbative treatment of wave-function components
Our first concern is to estimate the accuracy of a perturbative treatment of $`d^{}(NN^3D_3)`$. This is done by first examining a similar perturbative generation of the $`d(\mathrm{\Delta }\mathrm{\Delta }^7D_1)`$ from $`d(NN^3S_1)`$ using the Argonne-28 potential from :
$`\psi (\mathrm{\Delta }\mathrm{\Delta }^7D_1,p)\mathrm{\Delta }\mathrm{\Delta }^7D_1,p|d`$ (35)
$`\mathrm{\Delta }\mathrm{\Delta }^7D_1,p|{\displaystyle \frac{1}{\mathrm{\Delta }H}}V|d(NN^3S_1)`$ (36)
$`={\displaystyle \frac{1}{m_d2E_\mathrm{\Delta }(p)}}{\displaystyle \frac{1}{(2\pi )^3}}{\displaystyle \mathrm{\Delta }\mathrm{\Delta }^7D_1|𝒪_{18}(\widehat{𝐪})|NN^3S_1}`$ (37)
$`\times v_{18}(q)𝐩+𝐪|d(NN^3S_1)d^3𝐪.`$ (38)
Here $`\mathrm{\Delta }H=m_dH_0`$, with $`H_0`$ the unperturbed Hamiltonian and $`m_d`$ the deuteron mass, $`E_\mathrm{\Delta }(p)`$ is the nonrelativistic $`\mathrm{\Delta }`$ energy, $`𝒪_{18}(\widehat{𝐪})`$ is the $`NN\mathrm{\Delta }\mathrm{\Delta }`$ tensor-force operator $`S_{12}^{\mathrm{III}}(\widehat{𝐪})𝐓_1𝐓_2`$ in the A28 notation with $`𝐓_i`$ the $`N\mathrm{\Delta }`$ isospin operator. The subscript 18 refers to channel 18 of A28. The momentum-space potential is
$`v_{18}(q)=v_{\pi 0}^{\mathrm{III}}{\displaystyle \frac{4\pi }{\mu ^3}}{\displaystyle \frac{q^2}{q^2+\mu ^2}}f(q),`$ (39)
where $`\mu =138`$ MeV is the pion mass, $`v_{\pi 0}^{\mathrm{III}}=14.91`$ MeV, and $`f(q)`$ is a numerically generated cutoff function that decreases from 1 at $`q=0`$ to 0 at $`q=\mathrm{}`$.
Equation (38) can be simplified to a one-dimensional integral if the deuteron $`NN(^3S_1)`$ wave function is expanded as a sum of Gaussians. It is sufficient to give the final expression for a single Gaussian wave function of the form of Eq. (18), but with falloff parameter $`\beta `$ and an associated normalization constant $`N_0`$ if the S-state probability were 100%. Then
$`\psi _7(p)\psi (\mathrm{\Delta }\mathrm{\Delta }^7D_1,p){\displaystyle \frac{h(p)}{m_d2E_\mathrm{\Delta }(p)}},`$ (40)
where the subscript 7 refers to the spin degeneracy,
$`h(p)=\sqrt{{\displaystyle \frac{28}{5\pi ^3}}}N_0\mathrm{e}^{p^2/2\beta ^2}{\displaystyle v_{18}(q)\mathrm{e}^{q^2/2\beta ^2}}`$ (41)
$`\times j_2(ix)q^2dq,`$ (42)
and $`x=pq/\beta ^2`$.
The actual calculation is made with a three-term fit to the A28 deuteron $`NN(^3S_1)`$ wave function given in the Appendix. The resulting deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ wave function calculated from Eq. (40) (long dashed curve) is compared in Fig. 3 with the actual A28 wave function (solid curve) in the fitted form given by Eqs. (A1)-(A3). The wave function is so defined that $`\psi _7^2(p)p^2𝑑p`$ gives the fractional normalization $`P_7`$ in this $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ state. The calculated value in the perturbative approximation is $`P_7=0.65\%`$ compared to the exact value of 0.419%.
One can see from Fig. 3 that the inaccuracy in the perturbative treatment comes from the overestimate of the wave function at all momenta larger than $`1.3\mu `$, where $`\mu =138.0`$ MeV is the average pion mass used in these calculations. The error seems to arise primarily from the neglect of a short-range repulsive central potential in the $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ channel. Such a repulsive potential would have reduced the wave function at small inter-baryon distances. The size of the error in our perturbative treatment in this $`{}_{}{}^{7}D_{1}^{}`$ state seems to be much greater than the results found by . We shall see that there are much greater uncertainties elsewhere in the present calculation. Consequently, I shall consider the accuracy of the perturbative treatment adequate for the present qualitative study.
The same perturbative method is now used to obtain the $`NN(^3D_3)`$ component of $`d^{}`$. An added complication appears because the $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ component of $`d^{}`$ is a bound state embedded in the $`NN`$ continuum. The calculation will require the use of a spreading width into the $`NN`$ continuum (which is essentially just the total width $`\mathrm{\Gamma }`$ of $`d^{}`$) and a principal-value Green’s function. The matrix element of the tensor-force operator is also different from Eq. (38). The final result is rather similar to Eq. (40):
$`\psi _3(p)NN^3D_3,p|d^{}{\displaystyle \frac{\mathrm{\Delta }E}{(\mathrm{\Delta }E)^2+(\mathrm{\Gamma }/2)^2}}h^{}(p),`$ (43)
where $`\mathrm{\Delta }E(p)=m^{}2E_N(p)`$,
$`h^{}(p)=\sqrt{{\displaystyle \frac{12}{5\pi ^3}}}N_0^{}\mathrm{e}^{p^2/2\beta ^2}{\displaystyle v_{18}(q)\mathrm{e}^{q^2/2\beta ^2}}`$ (44)
$`\times j_2(ix^{})q^2dq,`$ (45)
and $`x^{}=pq/\beta ^2`$.
In Eq. (43), a rapidly changing factor $`\mathrm{\Delta }E/((\mathrm{\Delta }E)^2+(\mathrm{\Gamma }/2)^2)`$ has been separated from a function $`h^{}(p)`$ that does not depend explicitly on $`m^{}`$ or $`\mathrm{\Gamma }`$. $`\mathrm{\Delta }E`$ changes sign as the nucleon momentum $`p`$ increases above that value $`p_0`$ at which $`\mathrm{\Delta }E(p)`$ vanishes. Above $`p_0`$, $`\mathrm{\Delta }E`$ is negative so that the perturbative wave function $`\psi _3(p)`$ tends to have the same phase relation to the driving $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ wave function as $`\psi _7(p)`$ is to its own driving wave function. This means that diagrams 1b and c tend to interfere constructively at large momentum transfers, but destructively at small momentum transfers.
The falloff parameter $`\beta ^{}`$ and normalization constant $`N_0^{}`$ appearing in Eq. (45) are those for the single-Gaussian approximation to the $`d^{}(\mathrm{\Delta }\mathrm{\Delta }^7S_3)`$ wave function.
In the actual calculation, a three-term Gaussian fit to the two-centered $`d^{}`$ wave function is used. The results for potential A28 are shown as long dashed curves in Fig. 4 for $`h^{}(p)`$ and the radial part $`\psi _3(p)`$ of Eq. (43) calculated with $`m^{}=2100`$ MeV and $`\mathrm{\Gamma }=100`$ MeV. This spin triplet wave-function component has the perturbative normalization of $`P_3=0.038`$, or 3.8%. This normalization is very sensitive to the choice of the decay width, increasing to 0.55 if the width is decreased to 10 MeV. The rapid increase comes from the roughly $`1/\mathrm{\Gamma }`$ behavior of the amplitude of the kink where the wave function oscillates rapidly from a positive value to a negative value as the momentum $`p`$ goes through the zero of $`\mathrm{\Delta }E`$.
The great sensitivity of the $`{}_{}{}^{3}D_{3}^{}`$ wave function to the width $`\mathrm{\Gamma }`$ does not mean that the inelastic production amplitude increases just as dramatically. The wave function changes sign in the kink, leading to much cancellation in its contribution to the inelastic RME. Consequently, most RME’s increase by only 25% when the width decreases from 100 MeV to 10 MeV, and approach stable values for smaller widths.
Although the dependence on $`\mathrm{\Gamma }`$ as described above is technically correct in a narrow sense, it is also counter-intuitive in that a smaller width with weaker coupling to the $`NN`$ channel somehow leads to a stronger $`{}_{}{}^{3}D_{3}^{}`$ component. The truth is that in the discussion just given, the width is treated as if it were a free parameter when it actually is not. Rather once the dynamics is chosen, a certain width is implied, and must be used in Eqs. (40) and (43).
Hence it is desirable to have a more consistent treatment. I shall first calculate the width for each input potential at each dibaryon mass $`m^{}`$, and then use this calculated width to calculate admixed components.
### B Decay width of $`d^{}`$
The decay width of $`d^{}`$ into the $`NN`$ ($`\pi NN`$, $`d\pi \pi `$) channel has been estimated in (, ). It is clear that the $`NN`$ channel is by far its dominant decay channel; it is the only channel that has to be included in the present study. The $`d^{}`$ decay widths into $`NN`$ for the A28 potential and five different Bonn potentials are easily calculated by the same method. The results are shown in Table 1 for $`m^{}=2100`$ MeV and a single Gaussian wave function for $`d^{}`$ of radius $`r^{}=0.7`$ fm.
The potentials are ordered in Table 1 in decreasing values of their deuteron $`D`$-state probability $`P_\mathrm{D}`$. The differences in $`P_\mathrm{D}`$ have been obtained by adjusting the amount of short-range repulsion in the tensor potential by changes in the coupling constants and baryon form factors.
These potentials differ from one another in other interesting ways. Potential A28 has a $`NN\mathrm{\Delta }\mathrm{\Delta }`$ tensor potential from $`\pi `$ exchange with an overall strength $`f_{\pi \mathrm{\Delta }N}^2/f_{\pi NN}^2`$ relative to the $`NN`$ tensor potential given the Chew-Low value of 4 . The potential function has the Yukawa form corresponding to a meson propagator $`1/\omega ^2(q)`$ in momentum space, where $`\omega ^2(q)=q^2+m_{\mathrm{meson}}^2`$. Its short-distance cutoff is specified in coordinate space using a function that bends the potential back to the origin to simulate the contribution of $`\rho `$ exchange .
The Bonn potentials use nucleon form factors for the cutoff, with a monopole form for the $`\pi `$ vertex and usually a dipole form for the $`\rho `$ vertex. The cutoff masses $`\mathrm{\Lambda }_i`$ used are shown in the table. The meson ($`m`$) coupling constants used here for the Bonn-A, B, and C potentials , denoted here as potentials BA, BB and BC, and for the “Full” Bonn potential FB , are the quark-model values based on the ratio $`f_{m\mathrm{\Delta }N}^2/f_{mNN}^2=72/25`$ . Potentials BA, BB and BC are relativistic momentum-space potnetials. FB is a relativistic model with energy-dependent meson propagators. The coupled-channel III, denoted here as CC3, is nonrelativistic and uses a strong-coupling ratio of about 4.4.
The second numerical entry given in the first line for each potential is the width calculated with the $`\pi `$-exchange only of the potential. The result is clearly correlated with cutoff mass $`\mathrm{\Lambda }_\pi `$ in the Bonn potentials. It decreases roughly monotonically from 130 MeV to only 30 MeV as we go from potential BC to BB, BA, Full Bonn (FB) and CC3 potentials as $`\mathrm{\Lambda }_\pi `$ decreases from 3.0 to 0.8 GeV.
When the $`\rho `$-exchange contribution is added, the resulting width, shown as the third numerical entry in the first line, decreases dramatically to the range of 57 to 20 MeV. The trend appears to be correlated with the decreasing value of the deuteron D-state probability $`P_\mathrm{D}`$, except for the special case of potential FB discussed separately. It is of course the differences in $`P_\mathrm{D}`$ caused by different partial cancallations by the $`\rho `$-exchange potential that distinguish between these potentials in the first place.
The result for the FB potential is different because the $`NN\mathrm{\Delta }\mathrm{\Delta }`$ potential generated by a simple quark-model prescription is not that shown in the first line of the results given in the table for the FB potential, but on the fourth line, where the width of 17 MeV is in rough agreement with the value for potential BA and with the trend dictated by $`P_\mathrm{D}`$. Note that the $`\rho \mathrm{\Delta }N`$ form factor is a monopole, a difference noted in the table by the entry $`[(n_\pi ,n_\rho )=]`$ $`(1,1)`$ under the column “Specials”. The results of line 4 for the FB potential are also the results reported previously in .
In the actual FB potential, the meson-$`\mathrm{\Delta }N`$ vertices are further modified: First, the $`\rho \mathrm{\Delta }N`$ form factor is changed from a monopole to a dipole form with no change in the cutoff mass $`\mathrm{\Lambda }_\rho `$. This change of the form factor causes the $`\rho `$ contribtuion to be reduced significantly, thus causing the “unexpected” increase in the decay width.
A second change made in the actual FB potential is to decrease the cutoff mass $`\mathrm{\Lambda }_\pi `$ from 1.3 to 1.2 GeV. This change is dictated by the need to prevent the uncorrelated 2$`\pi `$ contributions to the potential in the lower partial waves from becoming unmanageably large . a situation that is not yet encountered in our lowest-order calculation. As far as our lowest-order result is concerned, Table 1 shows that the effect of this change is quite minor.
The widths for potential CC3 also stand out from the trend. Comparing them with those for potential BB, which has a comparable $`P_\mathrm{D}`$, we can see that particularly the $`\pi `$ contribution has been reduced significantly by the use of the much smaller cutoff mass, thus giving an abnormally small width in perturbation theory. Presumably, the channel-coupling potentials in CC3 will give rise to larger contributions from the long-range $`\pi `$-exchange potentials in higher orders. Hence it is likely that the large difference seen in the lowest-order calculation reported here will not persist so noticeably in higher orders.
To summarize, the $`d^{}`$ decay width appears to be controlled to a good extent by $`P_\mathrm{D}`$, and has a value in the range 15-60 MeV. In comparison, the value of 100 MeV for potential A28 is more appropriate to $`\pi `$-exchange only. In other words, the simulation of $`\rho `$ exchange contained in it has not been very effectively in the long-distance part of the tensor potential that controls the decay width.
I now turn to the many remaining entries in Table 1. The second line for each potential, or for each group within a potential, contains results obtained by replacing each cutoff factor $`1/[1+(q^2/\mathrm{\Lambda }^2)]`$ by the Gaussain $`\mathrm{exp}[(q^2/\mathrm{\Lambda }^2)]`$. This change, marked in the column “Specials” by the symbol G, improves the high-momentum behavior of various momentum integrals and also reduces the overestimate of the high-momentum wave function in our perturbative treatment. Its influence on our decay widths is minimal, except in potential CC3 where the small cutoff mass $`\mathrm{\Lambda }_\pi `$ allows a relatively large effect.
In using the potential models discussed here, we either use the given coordinate-space potential $`v(r)`$ or the standard meson propagator $`1/\omega ^2(q)`$, plus corrections from vertex form factors. The exception is the coupled-channel model CC3 of , for which we use the off-shell propagator
$`𝒫(q)=\omega ^1(q)[\omega (q)+\mathrm{\Delta }M]^1,`$ (46)
defined in the potential. Here $`\mathrm{\Delta }M=M_\mathrm{\Delta }M_N`$.
The question could be raised as to whether one should use the off-shell propagator in these cross-channel potential as a general policy. Taken by itself, the answer is probably no. The reason is that the standard propagator leads to energy-independent Yukawa potentials that are actually superior to energy-dependent potentials arising from the use of off-shell propagators. It is not obvious, but the use of an energy-independent potential actually corresponds to the inclusion of that component of the two-baryon state having a virtual meson “in the air” .
Although the use of the off-shell propagators is not recommended, it is clear that their use will reduce the decay width calculated with the standard propagators. To study the size of the reduction, I now repeat the calculation with off-shell propagators. The results are shown on the third line of each group, and marked by the symbol $`\mathrm{\Delta }M`$ under the “Specials” column. We see that this off-shell effect is quite large, especially on the $`\pi `$-exchange contributions. As a result, there is much closer cancellation than before between the $`\pi `$ and $`\rho `$ contributions, thus giving rise to a much smaller net decay width.
In subsequent calculations in this paper, I shall use the A28 potentials to give an extreme case, and the BB/G potential for a more representative example, of the results expected for the electroproduction $`T`$ factors. The decay widths calculated for these two potentials are shown in Fig. 5 as functions of the dibaryon mass $`m^{}`$. Gaussian vertex form factors are used in potential BB/G.
With these calculated widths, the $`{}_{}{}^{3}D_{3}^{}`$ component can now be estimated more reliably by perturbation theory. The results are given in Fig. 4 as long dashed (solid) curves for potential A28 (BB/G) using $`m^{}=2100`$ MeV and the calculated value of $`\mathrm{\Gamma }=100`$ MeV (38 MeV). Both the smooth part $`h^{}(p)`$ and the complete radial wave function $`\psi (NN^3D_3,p)`$ of Eq. (43) are shown. The perturbative normalization in this state for potential BB/G is $`P_3=4.5\%`$.
The perturbative result for $`\psi (\mathrm{\Delta }\mathrm{\Delta }^7D_1,p)`$ of Eq. (LABEL:DD7d1final) calculated for potential BB/G with $`m^{}=2100`$ MeV and $`\mathrm{\Gamma }=38`$ MeV are also given in Fig. 4 as a thick dashed curve when only the $`\pi `$-exchange part of the potential is included in the calculation, and as a thick solid curve when the $`\rho `$-exchange contribution is also included. The large reduction in the wave function caused by the inclusion of $`\rho `$-exchange is worthy of note. The perturbative normalization in this state for the complete BB/G potential is $`P_7=0.38\%`$.
### C Electroproduction $`T`$ factors
The inelastic production amplitudes from the dominant deuteron $`NN(^3S_1)`$ component to the $`NN(^3D_3)`$ component of $`d^{}`$, as described by diagram 1c can finally be calculated. The major difference from the procedure described in Sec. III is that the radial wave function $`\psi _3d^{}(NN^3D_3)`$ is now too complicated to be expanded readily in terms of harmonic-oscillator wave functions. It is kept in numerical form so that the final result, instead of being entirely analytic, now requires a one-dimensional numerical integration.
To simplify the calculation, the deuteron $`d(^3S_1)`$ wave function is expressed as a sum of three Gaussians. However, results need to be given only for a single Gaussian. For this purpose, I use the notation of Eq. (18), but with un-starred parameters. All production amplitudes then have the generic form
$`d(^3S_1)\widehat{𝒪}(q)d_3^{}=c_0N_0\mathrm{e}^{(q/2)^2/2\beta ^2}\mathrm{e}^{\alpha _0q^2/2}`$ (47)
$`\times {\displaystyle }\mathrm{e}^{p^2/2\beta ^2}\psi _3(p)f_𝒪(p,q)p^2dp,`$ (48)
but different integrand functions
$`f_𝒪(p,q)`$ $`=`$ $`j_2,\mathrm{for}\widehat{M}_2;`$ (49)
$`=`$ $`{\displaystyle \frac{\sqrt{6}}{M_N}}{\displaystyle \frac{2\beta ^2}{q}}j_2,\mathrm{for}\widehat{T}_{\mathrm{c2}}^{\mathrm{el}};`$ (50)
$`=`$ $`{\displaystyle \frac{\mu _\mathrm{s}(N)}{M_N}}{\displaystyle \frac{c_1}{\sqrt{6}}}\left[\left({\displaystyle \frac{q}{4}}+{\displaystyle \frac{4\beta ^2}{q}}\right)j_2ipj_1\right],`$ (52)
$`\mathrm{for}\widehat{T}_{\mu 2}^{\mathrm{el}};`$
$`=`$ $`{\displaystyle \frac{\mu _\mathrm{s}(N)}{M_N}}{\displaystyle \frac{2}{\sqrt{21}}}\left[\left({\displaystyle \frac{q}{4}}+{\displaystyle \frac{10\beta ^2}{q}}\right)j_2ipj_1\right],`$ (54)
$`\mathrm{for}\widehat{T}_{\mu 3}^{\mathrm{mag}}.`$
Here $`j_ij_i(ipq/\beta ^2)`$ is a spherical Bessel function, and $`c_0=\sqrt{7}`$, $`c_1=1`$.
The production amplitudes from diagram 1a have a similar form, but with $`c_0=\sqrt{3}`$, $`c_1=4`$, and of course different radial wave functions. The baryon magnetic-moment parameter that appears is now $`\mu _\mathrm{s}(\mathrm{\Delta })`$ instead of $`\mu _\mathrm{s}(N)`$. I have verified explicitly that they give the same numerical values as the oscillator expressions given in Sec. III. The difference in the numerical coefficients $`c_\alpha `$ comes from the recoupling of angular momentum. These alone favor the amplitudes form diagram 1c by a factor of roughly $`\sqrt{7/3}`$, which is a ratio of two 6j-symbols.
To perform the calculation for potential BB/G, I use a three-term approximation to the BB deuteron $`NN(^3S_1)`$ wave function . The resulting perturbative deuteron $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ wave function calculated from Eq. (40) is fitted to a 4-term harmonic-oscillator form. Both fitted wave functions are given in the Appendix. This fitted $`\mathrm{\Delta }\mathrm{\Delta }(^7D_1)`$ wave function is the one shown as a thick solid curve in Fig. 3.
The calculated effective $`T`$ factors are shown in Fig. 6 with the curves defined in the legion.
we can see that electroproduction RMS’s from diagram 1a (in the case of A28) or b (BB/G) and diagram 1c interfere destructively at small $`q`$ values. At $`q`$ = 300 MeV/$`c`$, for example, the production amplitude that contributes to the Coulomb factor $`T^L`$ from diagram 1c for potential A28 is opposite in sign and about 3.6 times larger than that from diagram 1a. Of the increase, a factor $`\sqrt{7/3}`$ comes from a 6j-symbol, while the remaining factor of 2.4 comes from the wave functions in a radial integral. Hence $`T^L`$ is about 7 times that for diagram 1a alone.
The production amplitude from diagram 1c varies more strongly with the momentum transfer because of the much stronger momentum dependence of the $`d^{}(NN^3D_3)`$ wave function it contains. This production amplitude changes sign just below $`q=1000`$ MeV/$`c`$ where the curve for diagrams 1a and c crosses that for diagram 1a alone the second time. This is above the interference zero that can be seen in Fig. 6. After the sign change, the amplitudes from the two diagrams add constructively.
In the effective transverse $`T`$-factors for potential A28, the production amplitudes from the two diagrams are of comparable magnitudes but opposite signs at low momentum transfers. The resulting destructive interference is very severe, causing the total contribution to $`T_{\mathrm{eff}}`$ to be some two orders of magnitude smaller than that from diagram 1a alone. The interference eventually becomes constructive at roughly the same momentum transfer as for the Coulomb $`T^L`$ factor.
The results for the BB/G potential is qualitatively similar, but the Coulomb $`T^L`$ factor is smaller at the smaller momentum transfers.
## V Results and discussions
For all electron energies greater than a few GeV, the effective $`T`$ factors are essentially energy independent. The remaining factors in the integrated differential production cross section $`d\sigma /d\mathrm{\Omega }`$ of Eq. (1) do depend on the energy, however. The differential cross section then increases roughly as the square of the scattered electron energy at comparable four-momentum transfers and small lab angles. This means that the cross section tends to increase quadratically with the incident electron energy as the latter increases.
The resulting angular distributions are shown in Fig. 7 for 1 GeV electrons. The result for potential A28 is for diagrams 1 a and c, while those for potential BB/G are for diagrams 1b and c. The interference between the two diagrams can be visualized more readily by also examining Fig. 8 which show additional results for potential BB/G from the diagram 1c alone.
For the dibaryon mass $`m^{}=2.1`$ GeV, the sharp local minima near $`60^{}`$ comes from the amplitude zero in the Coulomb $`T^L`$ factor. When diagram 1c appears alone, this zero comes from the sign change in the $`d^{}(NN^3D_3)`$ wave function. When both diagrams 1b and s are included, it comes from an interference zero in the sum of their amplitudes. The two diagrams interefere destructively below the sharp minimum, and constructively above it. For example, the integrated cross section at the lab angle of $`30^{}`$ is 0.05 nb/sr for potential BB/G, but 0.11 nb/sr for diagram 1c alone.
The over-estimate of the cross section by potential A28 can be seen in Fig. 7. At the lab angle of $`30^{}`$, the cross section of 0.09 nb/sr for potential A28 is about twice as large as the value for potential BB/G. Other than this, both cross sections show the same decreasing trend with increasing lab angle.
The Mott cross section $`\sigma _\mathrm{M}`$ and the target recoil factor $`f_{\mathrm{rec}}`$ do not depend on the dibaryon mass $`m^{}`$. However, the magnitude of four-momentum transfer decreases as $`m^{}`$ increases, while that of the three-momentum transfer increases. This leads to a rapid decrease in the kinetimatical factor $`v_\mathrm{L}`$.
The dynamical factor $`T_{\mathrm{eff}}`$ also changes with $`m^{}`$. This is partly because of the three-momentum transfer appearing in it and partly because the $`d^{}`$ wave function $`d^{}(NN^3D_3)`$ itself also changes. The interference between the two diagrams becomes rather complex. The Coulomb $`T^L`$ factor decreases rapidly with increasing $`m^{}`$. At $`m^{}=2.3`$ GeV, the minimum in the angular distribution at $`135^{}`$ is caused by the interference zero in both $`T^L`$ and $`T^T`$. At $`m^{}=2.4`$ GeV, the shallow minimum near $`90^{}`$, comes from the combined effects of an interference zero in $`T^T`$ at $`100^{}`$ and an interference zero in $`T^L`$ at $`50^{}`$. The destructive interference between the two production amplitudes at small momentum transfers tends to become more severe as $`m^{}`$ increases.
Fig. 7 shows how rapidly the cross section decreases with increasing $`m^{}`$. At the lab angle of $`30^{}`$, the integrated cross section with both diagrams present, which at 0.05 nb/sr is already small for $`m^{}=2.1`$ GeV, now falls down to only 0.0005 (0.00006) nb/sr for $`m^{}=2.3`$ (2.4) GeV, as shown in Fig. 7.
The very small calculated cross sections obtained for both larger angles and larger $`m^{}`$ suggest that the present lowest-order picture might not be adequate under these circumstances. A much more elaborate calculation that includes higher-order diagrams as well as additional production processes important at these larger three-momentum transfers might have to be considered.
The result reported here can be compared with a recent calculation by QSW , who include diagram 1c and a production amplitude going to the $`NN^{}(1520)`$ component of $`d^{}`$. They find that diagram 1c dominates the calculated cross section for 1 GeV electrons scattered to a lab angle of $`30^{}`$, and obtain a preliminary value for the integrated cross section greater than 10 nb/sr there.
In order to compare with QSW, we show in Fig. 8 our result for diagram 1c alone at the same $`m^{}=2.1`$ GeV but calculated with potential BB/G. Our cross section is only 0.11 nb/sr at 30, two orders of magnitude smaller.
One difference between the two calculations is that the baryon-baryon interaction used by QSW comes from and contains only the $`\pi `$-exchange contribution. In potential A28 and certainly in the Bonn potentials used here, the baryon-baryon tensor potentials are significantly reduced by cancellation against additional $`\rho `$-exchange contributions.
To isolate the contribution of $`\pi `$-exchange alone, we repeat the calculation for diagram 1c alone with only one modification — using the interaction of (called here potential GO) to generate the perturbative $`d^{}(NN^3D_3)`$ wave function. The potential GO has the coupling constant $`f_{\pi \mathrm{\Delta }N}^2=0.36`$ ($`=f^2/4\pi `$ in the notation of ), roughly consistent with those shown in Table I. A monopole form factor for the $`\pi \mathrm{\Delta }N`$ vertex is used with the cutoff parameter $`\mathrm{\Lambda }_\pi =1250`$ MeV. The $`{}_{}{}^{3}S_{1}^{}`$ wave function used in this calculation remains that of the Bonn B deuteron. This is not very different from the Bonn C wave function used by QSW.
The resulting cross section, for $`m^{}=2.1`$ GeV, is shown in Fig. 8 as a dotted curve. The calculated cross section at 30 is 0.41 nb/sr, almost four times bigger than the BB/G result. This shows that the neglected $`\rho `$-exchange contributions do reduce the cross section significantly.
The situation remains qualitatively the same when diagram 1a is also included: The cross section for potential GO at 30 is 0.29 nb/sr, a factor of 6 greater than the value of 0.05 nb/sr for potential BB/G.
The cross section for diagram 1c alone for potential BB/G calculated with $`m^{}=`$ 2.3 and 2.4 GeV are also given in the figure and compared with the results for $`m^{}=`$ 2.1 GeV with only diagram 1c and with both diagrams present.
After accounting for the difference in dynamical inputs, I find a remaining discrepancy of more than an order of magnitude between my calculation and that of QSW. This remaining difference must be due to differences in the calculational methods used. The most important difference seems to be my use of a principal-value Green’s function in my $`d^{}(NN^3D_3)`$ wave function compared to the use of an outgoing-wave boundary condition in QSW. The QSW calculation thus gives an additional term in the production amplitude from diagram 1c that contains an energy-conserving $`\delta `$-function. This additional term has a part that is the decay amplitude of $`d^{}(\mathrm{\Delta }\mathrm{\Delta }`$ into the $`NN`$ channel. I do not include this part as the physical electroproduction production amplitude. Additional studies must be made to understand if this could account for all the remaining disagreement.
Returning to the general problem of calculating the electroproduction cross section of $`d^{}`$, there are of course the additional uncertainties in the predicted mass and structure of $`d^{}`$ itself. Given the experimental non-observation of $`I`$=0 dibaryons in the mass range 2.00-2.23 GeV , future experimental searches and theoretical studies should probably move to higher masses. For electroproduction, the rapid decrease of the cross section with increasing $`d^{}`$ mass is a cause of concern if it persists in higher-order calculations.
The theoretical uncertainty in the $`d^{}`$ structure remains a major obstacle. The didelta model of $`d^{}`$ used here is very crude, and may not be adequate at short distances where the baryons overlap. Other exotic components in the deuteron wave function must also be taken into consideration, especially at the larger momentum transfers. The possibility of quark delocalization and color screening remains an open question, but the failure to see any $`d^{}`$ dibaryon in the experiment of is discouraging.
All these issues make it clear that much more work remains to be done before a quantitative description of the $`d^{}`$ production cross section can be achieved, given any theoretical model for $`d^{}`$. The preliminary study given in this paper does suggest that the electroproduction cross sections to $`d^{}`$ are likely to be very small.
Perhaps more interestingly, the existence of $`d^{}`$ at low masses can now be viewed from a rather elementary perspective. Consider a very naive model of $`d^{}`$ containing a probability $`P_{\mathrm{\Delta }\mathrm{\Delta }}(d^{})`$ of the $`\mathrm{\Delta }\mathrm{\Delta }(^7S_3)`$ configuration and the remaining probability, $`1P_{\mathrm{\Delta }\mathrm{\Delta }}(d^{})`$, of a closed-channel configuration that does not decay at all into any channel. The two pieces of information:
1. the best available experimental upper bound of 0.08 MeV obtained by for its decay width into the $`NN`$ channel near 2.1 GeV, and
2. the best theoretical estimate of its width of about 40 MeV given in this paper if $`d^{}`$ is a didelta,
if taken literally, would require that $`P_{\mathrm{\Delta }\mathrm{\Delta }}(d^{})`$ cannot exceed something of the order of $`0.2\%`$. This seems to suggest that any model of $`d^{}`$ containing a much larger $`P_{\mathrm{\Delta }\mathrm{\Delta }}(d^{})`$ might already have been excluded by the measurements of .
I wish to thank Fan Wang, Terry Goldman, Stan Yen, Earle Lomon, Bob Wiringa, Faqir Khanna and R. Dymarz for many stimulating conversations and correspondence. I am indebted to Stan Yen for asking about the resonance contribution to the total cross section measurement of Lisowski et al. that led to the experimental upper bound given in the Introduction.
## A Wave functions used in the calculation
Many of the wave functions used have been fitted to the form
$$\psi (p)\underset{i=1}{\overset{3}{}}c_i\psi _i(p),$$
(A1)
where $`\psi _i(p)`$ is a normalized harmonic-oscillator wave functions (HOWF) like that shown in Eq. (20), with the falloff parameter
$$𝜷^2=(\beta _1^2,\beta _2^2,\mathrm{}).$$
(A2)
The dimensionless expansion coefficients are
$$𝐜=(c_1,c_2,\mathrm{}).$$
(A3)
In order to emphasize the stronger high-momentum components in these short-distance wave functions, the range parameters are obtained by minimizing the percentage mean-square deviation.
The fitted parameters are given in Table II. Each fitted wave function has been renormalized in order to change the raw percent probability to the value the original wave function has, as indicated in the third line of the table for each state. |
warning/0001/hep-th0001138.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In this paper we examine in detail the constraints imposed by superconformal invariance and analyticity on four-point correlation functions of analytic operators in four-dimensional superconformal field theories with $`N=2`$ supersymmetry. The analytic operators under consideration are gauge-invariant products of hypermultiplets which are represented by analytic superfields in $`N=2`$ harmonic superspace. Analytic superfields obey a generalised chirality-type constraint and depend holomorphically on the coordinates of the two-sphere which is adjoined to Minkowski superspace to form harmonic superspace. The definition of analyticity is given in more detail below.
The sphere can be thought of as the homogeneous space $`U(1)\backslash SU(2)`$ and fields carry a charge with respect to the $`U(1)`$ isotropy group of this internal space. We shall be particularly interested in the case where each of the operators has charge 2. These operators are hypermultiplet bilinears and also have dimension 2. For the particular case of $`N=4`$ super Yang-Mills theory (SYM) there is a single hypermultiplet transforming under the adjoint representation of the gauge group and from this and its conjugate one can construct three charge 2 analytic bilinears. They can be viewed as $`N=2`$ components of the $`N=4`$ supercurrent, so that the corresponding four-point correlation functions give $`N=2`$ projections of the $`N=4`$ four-point correlator of four supercurrents. This correlator, or particular spacetime components of it, has been much studied in the context of the Maldacena conjecture on both the AdS and field theory sides .
The study of correlation functions of the above type in the harmonic superspace setting has been advocated in a series of papers <sup>2</sup><sup>2</sup>2For an analysis of superconformal field theories in Minkowski superspace see, for example, . The exact functional forms for two- and three-point functions were given (see also ) and it was later argued that the coefficients of such correlators in $`N=4`$ should be non-renormalised using analytic superspace techniques, the reduction formula and the notion of $`U(1)_Y`$ symmetry introduced in . (For the supercurrent correlators this non-renormalisation can also be seen as a consequence of anomaly considerations .) It was also conjectured that four-point correlators of operators with sufficiently low charges might be soluble . However, in spite of the claims made for future work in these references, this turns out not to be the case and one purpose of the present work is to give the precise result that one finds for four charge 2 operators, which are known to be non-trivial . As we shall show, the requirement of analyticity leads to additional constraints beyond those that one might expect on grounds of superconformal symmetry alone, but these are not enough to determine the correlation function completely. Elsewhere the result for charge 2 operators has been used to simplify the computation of the four-point function at two loops in perturbation theory .
The work described here has been carried out over a period of several years and has focused on four-point functions of operators with equal charges. As we have just mentioned, these results are not as strong as had been conjectured. More recently, however, it has become apparent that more striking results can be obtained by considering asymmetric sets of charges. In the methods of the present paper were used to show that four-point extremal correlators ($`p_4=p_1+p_2+p_3,p_i=`$ charges) are simply given by products of free two-point functions. A discussion in $`N=1`$ perturbation theory is given in . Since the analogous result had already been found in AdS supergravity (see also ), this also established a part of the Maldacena conjecture. So, although the initial conjectures of for equal charges have turned out to be incorrect, perhaps it is not unreasonable to claim some partial vindication for the initial optimism in the light of the fact that the very same strategy gives a good way of proving the recent results for extremal correlators in a field-theoretic setting.
The analysis of analytic operators and correlation functions can be carried out in two equivalent frameworks each having their own distinctive features. One can either work in analytic superspace (with complexified spacetime) using explicit coordinates, or one can work in harmonic superspace (with real spacetime) using an equivariant formalism with respect to the internal $`SU(2)`$ symmetry group. In the former approach, all the coordinates appear on an equal footing. This makes the action of the superconformal group is more transparent and facilitates the construction of invariants. Further, analyticity (or the lack of it) is manifest. In the latter approach the internal $`SU(2)`$ is treated covariantly, so that explicit coordinates for the sphere do not have to be introduced. In addition, harmonic analyticity can be interpreted as an irreducibility condition under $`SU(2)`$. This allows one to limit the analysis to the first two non-trivial levels in the $`\theta `$ expansion of the correlator. We shall employ both methods in this paper, using the analytic formalism in section 2 and the covariant formulation in section 3.
## 2 Coordinate approach
### 2.1 Analytic superspace
$`N=2`$ harmonic superspace, first introduced in , is the product of $`N=2`$ Minkowski superspace and the two-sphere $`S^2=\text{}P^1`$. A field on this space can be expanded in harmonics on the sphere with coefficients which are ordinary $`N=2`$ superfields. A general superfield on $`N=2`$ harmonic superspace is therefore equivalent to an infinite number of ordinary superfields, but, if a superfield depends holomorphically on the coordinates of the sphere, it will have a finite expansion in ordinary superfields due to the fact that spaces of holomorphic tensors on the sphere are finite-dimensional. Such a field is called harmonic analytic (H-analytic). In addition one can define a generalised notion of chirality, called Grassmann analyticity (G-analyticity) such that a G-analytic field depends on only half the number of odd coordinates of the full superspace. A field which is both H-analytic and G-analytic will be called analytic.
An alternative description of analytic superfields on complexified Minkowski superspace is in terms of (holomorphic) fields on analytic superspace, a space with half the odd dimensionality of harmonic superspace. It bears a similar relation to harmonic superspace as chiral superspace does to Minkowski superspace.
Analytic superspace is a homogeneous space of the complexified $`N=2`$ superconformal group, $`SL(4|2)`$ (for a review of various homogeneous superspaces in this context see ). This group acts naturally (to the left) on $`N=2`$ supertwistor space $`\text{}^{4|2}`$, and the coset we are interested in has the geometrical interpretation as the Grassmannian of $`(2|1)`$-planes in twistor space. The body of the whole of this Grassmannian is compact, whereas we are interested only in the usual, non-compact Minkowski spacetime. The body of the actual space we shall work with will therefore be restricted in this sense, although its internal part, $`\text{}P^1`$, remains compact. This is similar in spirit to regarding as an open subset of $`\text{}P^1`$ obtained by omitting the point at infinity. We shall be a bit more precise about this below, after we have introduced appropriate local coordinates.
In the usual basis for $`N=2`$ supertwistor space the first four elements correspond to the even part and the second two to the odd part. If we instead make a choice of basis ordered in the sequence two even, one odd, two even, one odd, the isotropy group, $`H`$, of analytic superspace will consist of supermatrices of the simple form
$$\left(\begin{array}{cc}& 0\\ & \end{array}\right).$$
(1)
Here each entry represents a $`(2|1)\times (2|1)`$ supermatrix and the bullets denote non-singular such matrices. The superdeterminant is constrained to be 1. In this basis it is reasonably clear that the coset space $`H\backslash SL(4|2)`$ is indeed the Grassmannian of $`(2|1)`$-planes in $`\text{}^{4|2}`$.
Using standard homogenous space techniques we may choose a local coset representative $`s(X),XM_A`$ (analytic superspace) as follows:
$$M_AXs(X)=\left(\begin{array}{cc}1& X\\ 0& 1\end{array}\right)SL(4|2)$$
(2)
and again each entry represents a $`(2|1)\times (2|1)`$ supermatrix. The components of $`X`$ are given by
$$X=\left(\begin{array}{cc}x^{\alpha \dot{\alpha }}& \lambda ^\alpha \\ \pi ^{\dot{\alpha }}& y\end{array}\right)$$
(3)
where $`\alpha ,\dot{\alpha }`$ are two-component spinor indices, $`x`$ is the spacetime coordinate, $`\lambda `$ and $`\pi `$ are the odd coordinates and $`y`$ is the standard coordinate on $`\text{}P^1`$. (We shall occasionally use index notation $`X^{AA^{}}`$ for the coordinate matrix $`X`$.) As we mentioned above, we want the body of our superspace to consist of non-compact spacetime together with a compact internal space, $`\text{}P^1`$. The full space we are interested in can be covered by two open sets corresponding to the two standard open sets of the sphere. If we denote these two sets by $`U`$ and $`U^{}`$, and put primes on the coordinates for $`U^{}`$ we find that the two sets are related as follows on the overlap:
$`x^{}`$ $`=`$ $`x{\displaystyle \frac{\lambda \pi }{y}},`$
$`\lambda ^{}`$ $`=`$ $`{\displaystyle \frac{1}{y}}\lambda ,`$
$`\pi ^{}`$ $`=`$ $`{\displaystyle \frac{1}{y}}\pi ,`$
$`y^{}`$ $`=`$ $`{\displaystyle \frac{1}{y}}.`$ (4)
We see that the odd coordinates and the coordinates of the internal space together parametrise $`\text{}^{(1|4)}`$, so that the whole space has the form of an affine bundle of rank $`(4|0)`$ over $`\text{}P^{(1|4)}`$.
Superconformal transformations can be discussed straightforwardly using standard homogeneous space methods. Under a superconformal transformation $`XXg,gSL(4|2)`$. The transformed coordinates $`Xg`$ are determined using the formula
$$s(Xg)=h(X,g)s(X)g.$$
(5)
The rôle of the compensating transform $`h(X,g)`$ (an element of $`H`$) is to restore the form of the coset representative $`s(X)`$ after right-multiplication by $`g`$. For an infinitesimal transformation specified by $`𝒜\text{s}\text{l}(4|2)`$, the Lie superalgebra of $`SL(4|2)`$, we have
$$\delta X=B+AX+XD+XCX$$
(6)
with
$$𝒜=\left(\begin{array}{cc}A& B\\ C& D\end{array}\right).$$
(7)
An analytic field of charge $`p`$ is a field $`\varphi `$ on analytic superspace which transforms under an infinitesimal superconformal transformation according to the rule
$$\delta \varphi =V\varphi +p\mathrm{\Delta }\varphi $$
(8)
where $`V`$ is the vector field generating the transformation, $`V=\delta X\frac{}{X}`$, and $`\mathrm{\Delta }:=\mathrm{str}(A+XC)`$. The free, on-shell hypermultiplet is such a field with charge $`1`$ and will be discussed in more detail below. In an interacting theory, the on-shell hypermultiplet is covariantly analytic, but gauge-invariant products of hypermultiplets are analytic in the above sense with charges equal to the number of hypermultiplets in the product.
For completeness we reproduce here the explicit expressions for the vector fields corresponding to the different types of superconformal transformation. From (6) one can read off the vector fields for each of the parameters. They divide into translational ($`B`$), linear ($`A,D`$) and quadratic ($`C`$) types. The translations are ordinary spacetime translations, half of the $`Q`$-supersymmetry transformations and translations in the internal $`y`$ space, $`\text{}P^1`$. The corresponding vector fields are
$$V_{AA^{}}=\frac{}{X^{AA^{}}}.$$
(9)
The linearly realised symmetries are Lorentz transformations ($`SL(2)\times SL(2)`$) in complex spacetime) and dilations, internal dilations, $`R`$-symmetry transformations, the other half of the $`Q`$-supersymmetries and half of the $`S`$-supersymmetries. The Lorentz transformations are handled in the usual way so that we do not need to write them down. The vector fields generating dilations ($`D`$), internal dilations ($`D^{}`$) and $`R`$-symmetry transformations are
$`V(D)`$ $`=`$ $`x^{\alpha \dot{\alpha }}_{\alpha \dot{\alpha }}+{\displaystyle \frac{1}{2}}(\lambda ^\alpha ^{}_\alpha +\pi ^{\dot{\alpha }}_{\dot{\alpha }}),`$ (10)
$`V(D^{})`$ $`=`$ $`y_y+{\displaystyle \frac{1}{2}}(\lambda ^\alpha _\alpha +\pi ^{\dot{\alpha }}_{\dot{\alpha }}),`$ (11)
$`V(R)`$ $`=`$ $`\lambda ^\alpha _\alpha \pi ^{\dot{\alpha }}_{\dot{\alpha }}.`$ (12)
The vector fields generating linearly realised $`Q`$-supersymmetry are
$`V(Q)_\alpha `$ $`=`$ $`\pi ^{\dot{\alpha }}_{\alpha \dot{\alpha }}+y_\alpha ,`$ (13)
$`V(Q)_{\dot{\alpha }}`$ $`=`$ $`\lambda ^\alpha _{\alpha \dot{\alpha }}y_{\dot{\alpha }},`$ (14)
while those generating linearly realised $`S`$-supersymmetry are
$`V(S)^\alpha `$ $`=`$ $`x^{\alpha \dot{\alpha }}_{\dot{\alpha }}+\lambda ^\alpha _y,`$ (15)
$`V(S)^{\dot{\alpha }}`$ $`=`$ $`x^{\alpha \dot{\alpha }}_\alpha \pi ^{\dot{\alpha }}_y.`$ (16)
The remaining supersymmetry transformations are the non-linearly realised $`S`$-supersymmetries generated by
$`V(S)^{\dot{\alpha }}`$ $`=`$ $`x^{\beta \dot{\alpha }}\pi ^{\dot{\beta }}_{\beta \dot{\beta }}+x^{\beta \dot{\alpha }}y_\beta \pi ^{\dot{\alpha }}\pi ^{\dot{\beta }}_{\dot{\beta }}\pi ^{\dot{\alpha }}y_y,`$ (17)
$`V(S)^\alpha `$ $`=`$ $`\lambda ^\beta x^{\alpha \dot{\beta }}_{\beta \dot{\beta }}\lambda ^\beta \lambda ^\alpha _\beta +yx^{\alpha \dot{\beta }}_{\dot{\beta }}+y\lambda ^\alpha _y.`$ (18)
Finally, we have conformal boosts ($`K`$) and internal conformal boosts ($`K^{}`$) generated by
$`V(K)^{\alpha \dot{\alpha }}`$ $`=`$ $`x^{\beta \dot{\alpha }}x^{\alpha \dot{\beta }}_{\beta \dot{\beta }}+x^{\beta \dot{\alpha }}\lambda ^\alpha _\beta +\pi ^{\dot{\alpha }}x^{\alpha \dot{\beta }}_{\dot{\beta }}+\pi ^{\dot{\alpha }}\lambda ^\alpha _y,`$ (19)
$`V(K^{})`$ $`=`$ $`\lambda ^\beta \pi ^{\dot{\beta }}_{\beta \dot{\beta }}+\lambda ^\beta y_\beta +y\pi ^{\dot{\beta }}_{\dot{\beta }}+y^2_y.`$ (20)
The function $`\mathrm{\Delta }`$ is non-zero in the following cases:
$`\mathrm{\Delta }(D)`$ $`=`$ $`1,`$
$`\mathrm{\Delta }(D^{})`$ $`=`$ $`{\displaystyle \frac{1}{2}},`$
$`\mathrm{\Delta }(S)^{\dot{\alpha }}`$ $`=`$ $`\pi ^{\dot{\alpha }},`$
$`\mathrm{\Delta }(S)^\alpha `$ $`=`$ $`\lambda ^\alpha ,`$
$`\mathrm{\Delta }(K)^{\alpha \dot{\alpha }}`$ $`=`$ $`x^{\alpha \dot{\alpha }},`$
$`\mathrm{\Delta }(K^{})`$ $`=`$ $`y.`$ (21)
In these lists for $`V`$ and $`\mathrm{\Delta }`$ we have omitted the parameters which of course have the opposite indices to those displayed. The parameters can easily be restored, but this should be done from the left in order to get the right signs.
### 2.2 Analytic fields
We recall that a holomorphic tensor field of charge $`p`$ on $`\text{}P^1`$ is given by two local functions $`a,a^{}`$ on $`U,U^{}`$ respectively, such that, in the overlap
$$a(y)=y^pa^{}(y^{})$$
(22)
Expanding both sides in power series in their respective variables, equating powers of $`y(=\frac{1}{y^{}})`$ and demanding the absence of poles we find
$$a(y)=\underset{n=0}{\overset{n=p}{}}a_ny^n.$$
(23)
and similarly for $`a^{}(y^{})`$. The two expansions are related by
$$a_n=a_{pn}^{}.$$
(24)
Hence the space of tensor fields of charge $`p`$ is a finite-dimensional space with dimension $`p+1`$ which can be identified with the space of $`p`$th rank totally symmetric tensors under $`SL(2)`$.
In a similar fashion we define an analytic superfield of charge $`p`$ on $`M_A`$ to be specified by two local holomorphic functions $`\varphi (x,\lambda ,\pi ,y)`$ and $`\varphi ^{}(x^{},\lambda ^{},\pi ^{},y^{})`$ defined on the two standard coordinate patches $`U,U^{}`$ respectively, such that, in the overlap
$$\varphi (x,\lambda ,\pi ,y)=y^p\varphi ^{}(x^{},\lambda ^{},\pi ^{},y^{}).$$
(25)
If we now expand both sides in the odd variables and in $`y`$ or $`y^{}`$, we obtain restrictions on the component functions. For example, the zeroth order term in $`\lambda \pi `$ is an $`x`$-dependent charge $`p`$ holomorphic tensor on $`\text{}P^1`$, the component of $`\lambda `$ behaves like a tensor of charge $`p1`$ and so on. In addition, since the relation between $`x`$ and $`x^{}`$ involves a shift, spacetime derivatives will appear in the constraints.
The basic superfield we shall consider is the hypermultiplet. In this language this multiplet is represented by an analytic superfield $`\varphi `$ of charge 1. Using the above method one can easily seen that it has only a short expansion:
$$\varphi (x,\lambda ,\pi ,y)=\phi (x,y)+\lambda ^\alpha \psi _\alpha (x)+\pi ^{\dot{\alpha }}\chi _{\dot{\alpha }}(x)+\lambda ^\alpha \pi ^{\dot{\alpha }}\stackrel{ˇ}{\phi }_{\alpha \dot{\alpha }}(x)$$
(26)
Furthermore
$$\phi (x,y)=\phi _o(x)+y\phi _1(x)$$
(27)
and similarly for $`\phi ^{}(x^{},y^{})`$, with
$$\phi _o=\phi _1^{};\phi _1=\phi _o^{}.$$
(28)
We also find
$$\stackrel{ˇ}{\phi }_{\alpha \dot{\alpha }}=_{\alpha \dot{\alpha }}\phi _1;\stackrel{ˇ}{\phi }_{\alpha \dot{\alpha }}^{}=_{\alpha \dot{\alpha }}\phi _o.$$
(29)
In addition, the fields $`\phi _o,\phi _1,\psi ,\chi `$ must all satisfy their equations of motion
$`\mathrm{}\phi _o=\mathrm{}\phi _1`$ $`=`$ $`0,`$
$`^{\dot{\alpha }\alpha }\psi _\alpha `$ $`=`$ $`0,`$
$`^{\alpha \dot{\alpha }}\chi _{\dot{\alpha }}`$ $`=`$ $`0.`$ (30)
This is the usual hypermultiplet with two complex scalar fields and two complex Weyl fermions, all of which are physical and on-shell. In the interacting theory, the on-shell hypermultiplet will be covariantly analytic, but gauge-invariant products of hypermultiplets will be analytic tensor fields of the type we have just described with charges $`2,3,\mathrm{}`$. For example, a field of charge two contains an independent vector field $`v_{\alpha \dot{\alpha }}`$, which is conserved, but there are no equations of motion. In other words a charge two field is a linear multiplet.
### 2.3 Correlation functions and Ward identities
In this section we consider the superconformal Ward identities for four-point correlation functions for analytic operators with charges $`p_1,p_2,p_3,p_4`$. We shall denote such correlators by $`<p_1p_2p_3p_4>`$. Note, however, that the operators are not assumed to be the same even if they have the same charges, in particular, we shall not impose any symmetry requirements on such correlators.
If we assume that analyticity holds in the quantum theory, the superconformal Ward identity for such a correlator reads
$$\underset{i=1}{\overset{4}{}}(V_i+p_i\mathrm{\Delta }_i)<p_1p_2p_3p_4>=0$$
(31)
The assumption of analyticity is tantamount to using the field equations of the underlying hypermultiplet at operator level. This should be reasonable provided that we keep the points separated. In the case of charge two operators, which is the main focus of this paper, the H-analyticity condition implies, as we noted above, that the superfield describes a linear multiplet with a conserved spacetime current. Analyticity can be examined directly in perturbation theory using the harmonic superspace formalism and has been verified in all the examples that have been looked at so far.
We shall consider correlators of the above type which have non-vanishing leading terms. Further, we will specialise to correlators with four equal charges $`p`$. In this case we can write
$$<pppp>=(g_{12})^p(g_{34})^pF$$
(32)
where $`g_{ij}`$ is the free two-point function for charge one operators at points $`i`$ and $`j`$,
$$g_{ij}=\mathrm{sdet}X_{ij}^1=\frac{\widehat{y}_{ij}}{x_{ij}^2}$$
(33)
and $`F`$ is an arbitrary function of invariants. Here $`X_{ij}=X_iX_j`$ denotes the coordinate difference matrix for points $`i`$ and $`j`$ and
$$\widehat{y}_{ij}=y_{ij}\pi _{ij}x_{ij}^1\lambda _{ij}$$
(34)
with the index convention that $`x^1`$ has a pair of subscript indices $`\dot{\alpha }\alpha `$.
It is straightforward to check that $`g_{ij}`$ satisfies the equation
$$(V_i+V_j+\mathrm{\Delta }_i+\mathrm{\Delta }_j)g_{ij}=0$$
(35)
so that the Ward identity for $`<pppp>`$ will indeed be satisfied for any invariant $`F`$.
It is not difficult to see that there must be more $`N=2`$ analytic superconformal invariants than there are spacetime conformal invariants for four points. If we expand $`F`$ in the odd variables $`\lambda `$ and $`\pi `$ the leading term must be invariant under spacetime conformal transformations and also under conformal ($`SL(2)`$) transformations of $`\text{}P^1`$. At four points there are two independent spacetime ($`x`$) cross-ratios and one independent internal ($`y`$) cross-ratio, so that there should be at least three independent analytic superconformal invariants at four points. In fact, there are no more. Any additional independent invariant would have to be nilpotent, but one can easily show that there are no such invariants at four points. The reason is essentially due to counting; there are $`4\times 4=16`$ odd coordinates which is precisely equal to the number of supersymmetries in $`SL(4|2)`$. On a putative nilpotent invariant these supersymmetries behave essentially like translational symmetr! ies so that any possible leading term must vanish. In more detail, suppose that $`F`$ is a nilpotent four-point invariant. It can be written $`F_o+\mathrm{}`$ where $`F_o`$ is the term with the lowest power of $`\lambda \pi `$. By examining the supersymmetry transformations directly one can see that the first term in the variation of $`F`$ involves only $`F_o`$. Furthermore, again by looking at each transformation in turn, one finds that setting this first term in the variation equal to zero (because $`F`$ is an invariant) leaves no possible solutions for separated points. Consequently one concludes that there can be no nilpotent invariants. This argument is presented in detail in .
It was shown in that non-nilpotent analytic superspace superconformal invariants can be expressed in terms of superdeterminants and supertraces of the coordinate differences $`X_{ij}`$. For the case in hand a possible choice of three independent invariants is given as follows: we take two super cross-ratios
$$S=\frac{\mathrm{sdet}X_{14}\mathrm{sdet}X_{23}}{\mathrm{sdet}X_{12}\mathrm{sdet}X_{34}},T=\frac{\mathrm{sdet}X_{13}\mathrm{sdet}X_{24}}{\mathrm{sdet}X_{12}\mathrm{sdet}X_{34}}$$
(36)
and one supertrace invariant
$$U=\mathrm{str}(X_{12}^1X_{23}X_{34}^1X_{41}).$$
(37)
The invariants $`S`$ and $`T`$ may be expressed in terms of the spacetime cross-ratios
$$s=\frac{x_{14}^2x_{23}^2}{x_{12}^2x_{34}^2},t=\frac{x_{13}^2x_{24}^2}{x_{12}^2x_{34}^2}$$
(38)
and the internal cross-ratio
$$v=\frac{y_{14}y_{23}}{y_{12}y_{34}}$$
(39)
in the form
$$S=\frac{s}{\widehat{v}},T=\frac{t}{\widehat{w}}.$$
(40)
Here $`w=\frac{y_{13}y_{24}}{y_{12}y_{34}}=1+v`$, and the hats on $`v`$ and $`w`$ are defined by hatting each of the $`y`$ variables in their definitions,
$$\widehat{v}=\frac{\widehat{y}_{14}\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}},\widehat{w}=\frac{\widehat{y}_{13}\widehat{y}_{24}}{\widehat{y}_{12}\widehat{y}_{34}}.$$
(41)
We may write
$$\widehat{w}=1+\widehat{v}+\mathrm{\Delta }w$$
(42)
and express the third invariant $`U`$ in the form
$$U=1t+s+\widehat{v}+\mathrm{\Delta }U.$$
(43)
Both $`\mathrm{\Delta }W`$ and $`\mathrm{\Delta }U`$ are nilpotent quantities. The explicit expressions are as follows:
$`\mathrm{\Delta }U`$ $`=`$ $`{\displaystyle \frac{1}{\widehat{y}_{12}\widehat{y}_{34}}}(\widehat{y}_{12}\mathrm{\Pi }_2x_{32}^1\mathrm{\Lambda }_2+\widehat{y}_{34}\mathrm{\Pi }_1x_{23}^4\mathrm{\Lambda }_1`$ (44)
$`+\widehat{y}_{23}\left(\mathrm{\Pi }_1x_{21}^4\mathrm{\Lambda }_1\mathrm{\Pi }_2x_{34}^1\mathrm{\Lambda }_2\mathrm{\Pi }_1\overline{x}_{21}^4\mathrm{\Lambda }_2+\mathrm{\Pi }_2\overline{x}_{34}^1\mathrm{\Lambda }_1\right)`$
$`+(\mathrm{\Pi }_1x_{23}\mathrm{\Lambda }_2)(\mathrm{\Pi }_2x_{23}\mathrm{\Lambda }_1))`$
and
$`\mathrm{\Delta }w`$ $`=`$ $`{\displaystyle \frac{1}{\widehat{y}_{12}\widehat{y}_{34}}}(\widehat{y}_{12}\mathrm{\Pi }_2x_{32}^4\mathrm{\Lambda }_2+\widehat{y}_{34}\mathrm{\Pi }_1x_{23}^1\mathrm{\Lambda }_1`$ (45)
$`+\widehat{y}_{23}\left(\mathrm{\Pi }_1[x_{23}^1x_{24}^1]\mathrm{\Lambda }_1+\mathrm{\Pi }_2[x_{31}^4x_{32}^4]\mathrm{\Lambda }_2+\mathrm{\Pi }_1\overline{x}_{24}^1\mathrm{\Lambda }_2\mathrm{\Pi }_2\overline{x}_{31}^4\mathrm{\Lambda }_1\right)`$
$`(\mathrm{\Pi }_1x_{21}^3\mathrm{\Lambda }_1)(\mathrm{\Pi }_2x_{34}^2\mathrm{\Lambda }_2))`$
where we have used the convenient shorthand
$`x_{ij}^k`$ $`=`$ $`x_{ik}x_{jk}^1x_{ij},`$ (46)
$`\overline{x}_{ij}^k`$ $`=`$ $`x_{ik}x_{jk}^1x_{jl}l,i,j,k.`$ (47)
The odd variables are defined by
$`\mathrm{\Lambda }_{1\dot{\alpha }}`$ $`=`$ $`(x_{12})_{\dot{\alpha }\alpha }^1\lambda _{12}^\alpha (x_{23})_{\dot{\alpha }\alpha }^1\lambda _{23}^\alpha ,`$
$`\mathrm{\Lambda }_{2\dot{\alpha }}`$ $`=`$ $`(x_{23})_{\dot{\alpha }\alpha }^1\lambda _{23}^\alpha (x_{34})_{\dot{\alpha }\alpha }^1\lambda _{34}^\alpha `$ (48)
and, similarly,
$`\mathrm{\Pi }_{1\alpha }`$ $`=`$ $`\pi _{12}^{\dot{\alpha }}(x_{12})_{\dot{\alpha }\alpha }^1\pi _{23}^{\dot{\alpha }}(x_{23})_{\dot{\alpha }\alpha }^1,`$
$`\mathrm{\Pi }_{2\alpha }`$ $`=`$ $`\pi _{23}^{\dot{\alpha }}(x_{23})_{\dot{\alpha }\alpha }^1\pi _{34}^{\dot{\alpha }}(x_{34})_{\dot{\alpha }\alpha }^1.`$ (49)
Now the crucial point is the following: each of the operators in the correlator can be expanded as a polynomial in $`y`$, so that the correlator is manifestly analytic in the $`y`$ variables. On the other hand, each of the invariants depend on the $`y`$’s in a rational manner so that one might expect the absence of singularities to impose further constraints on $`F`$. These constraints will clearly depend on the charges involved. For the correlator $`<pppp>`$ the lowest term in $`F`$ must be of the form
$$F|_{\lambda =\pi =0}=a_0+a_1v+\mathrm{}a_{p+1}v^{p+1}$$
(50)
where each of the $`a`$’s depends on the cross-ratios $`s,t`$. The question then arises whether there are further constraints on the coefficient functions at higher orders. It is clear that the lower the charge the more constrained $`F`$ must be. For charge one, there are no gauge-invariant operators, so the simplest interesting case to examine is charge 2 to which we turn in the next section.
### 2.4 Analyticity analysis
In this section we analyse the constraints imposed by H-analyticity on the four-point function of four charge two operators (not necessarily the same). It can be written
$$<2222>=\frac{\widehat{y}_{12}^2\widehat{y}_{34}^2}{x_{12}^4x_{34}^4}F(S,T,U)$$
(51)
The invariants $`S,T,U`$ are convenient from some points of view - it is easy to see that they are invariants, and they have concise explicit forms. Requiring H-analyticity implies that $`F`$ should have no singularities in $`(\widehat{y}_{13},\widehat{y}_{14},\widehat{y}_{23},\widehat{y}_{24})`$ and that it can have poles up to order two in $`(y_{12},y_{34})`$ (if we had charge $`p`$ operators this would become order $`p`$). Since $`(\widehat{y}_{12}\widehat{y}_{34})`$ occurs as a denominator in $`S^1,T^1`$ and $`U`$ the regularity of the correlator will lead to constraints for $`F`$.
Note that regularity in the hatted variables is equivalent to regularity in the unhatted ones: if we demand that the whole correlator be a polynomial in the hatted $`y`$’s and write, for example, $`\widehat{y}_{12}=y_{12}+\delta _{12}`$, then a Taylor expansion will produce a polynomial of the same degree in the unhatted $`y`$’s, because the $`\delta `$’s are non-singular and $`y`$-independent.
Clearly, H-analyticity should hold at each order in the odd variables separately. We shall carry out the expansion in two steps: we first expand in the nilpotent quantities $`\mathrm{\Delta }U,\mathrm{\Delta }w`$ and then express the latter in terms of products of spinors. We will here refer to the order in $`\mathrm{\Delta }U,,\mathrm{\Delta }w`$ as “level” in order to distinguish the first from the second step. The main technical problem one faces in this approach is to compute which of the various products of the form $`(\mathrm{\Delta }u)^p(\mathrm{\Delta }w)^q`$ are independent for each fixed value of $`p+q`$, i.e. at a given “level”.
From the point of view of expanding in odd variables it is helpful to think about an equivalent set of invariants, $`S^{},T^{},V`$, which are defined to have leading terms $`s,t`$ and $`v`$ respectively. These invariants are expressed in terms of $`S,T,U`$ by
$$S^{}=SV,T^{}=T(1+V),V=\frac{T+U1}{1+ST}$$
(52)
Since we have
$`S^{}`$ $`=`$ $`s+\mathrm{}`$ (53)
$`T^{}`$ $`=`$ $`t+\mathrm{}`$ (54)
$`V`$ $`=`$ $`v+\mathrm{}`$ (55)
it follows from (51) that we may write $`F`$ in the form
$$F=a_1(S^{},T^{})+a_2(S^{},T^{})V+a_3(S^{},T^{})V^2.$$
(56)
This expression clearly meets the lowest level requirements of analyticity for charge 2 and shows that the dependence on the third invariant $`V`$ is thereby fixed. The objective now is to Taylor expand F about $`S^{}=s,T^{}=t`$ and $`V=\widehat{v}`$. However, since the original set of invariants is easier to evaluate explicitly, we shall convert the Taylor expansion back into these variables as we go. In this way we arrive at a power series in the nilpotent quantities $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }w`$. We will then express both of these and the coefficients that accompany them in terms of the set of coordinates $`(x_{12},x_{23},x_{34},\widehat{y}_{12},\widehat{y}_{23},\widehat{y}_{34},\mathrm{\Lambda }_1,\mathrm{\Lambda }_2,\mathrm{\Pi }_1,\mathrm{\Pi }_2)`$. These variables are convenient in the sense that they are invariant under translational $`Q`$-supersymmetry and linear $`S`$-supersymmetry, and in addition they involve no $`y`$-singularities. Furthermore, as noted above, the difference between $`y`$ and $`\widehat{y}`$ is non-singular so it is permissible to study analyticity in the latter rather than the former.
The Taylor expansion of $`F`$ is then
$`F(S,T,U)`$ $`=F_o+\mathrm{\Delta }U(_UF)+\mathrm{\Delta }w({\displaystyle \frac{t}{\overline{w}^2}}_TF)`$ (57)
$`{\displaystyle \frac{1}{2}}\mathrm{\Delta }U^2(_U^2F)+\mathrm{\Delta }U\mathrm{\Delta }w({\displaystyle \frac{t}{\overline{w}^2}}_U_TF)+{\displaystyle \frac{1}{2}}\mathrm{\Delta }w^2({\displaystyle \frac{t^2}{\overline{w}^4}}_T^2F+2{\displaystyle \frac{t}{\overline{w}^3}}_TF)+\mathrm{}`$
where $`F_o`$ is $`F(S^{},T^{},V)`$ evaluated at $`(s,t,\widehat{v})`$. Clearly
$$F_o=a_1(s,t)+a_2(s,t)\widehat{v}+a_3(s,t)\widehat{v}^2.$$
(58)
In the new variables the partial derivatives, evaluated at $`(S^{},T^{},\widehat{v})`$, are
$`_U`$ $`=`$ $`{\displaystyle \frac{1}{R}}D,`$
$`_T`$ $`=`$ $`{\displaystyle \frac{(1+\widehat{v})}{R}}(D+R_t)`$ (59)
with
$`R`$ $`=`$ $`s+\widehat{v}{\displaystyle \frac{\widehat{v}t}{1+\widehat{v}}},`$
$`D`$ $`=`$ $`s_s+\widehat{v}_{\widehat{v}}+{\displaystyle \frac{\widehat{v}t}{1+\widehat{v}}}_t.`$ (60)
Here $`v`$ has been replaced by $`\widehat{v}`$ because we wish to use the Taylor expansion about the point $`(s,t,\widehat{v})`$. We will almost always multiply $`R`$ and $`D`$ by $`\overline{w}=1+\widehat{v}`$ in order to avoid the singularity in their denominators. Note that $`DR=R`$.
We shall refer to the expressions in round brackets multiplying a certain power $`\mathrm{\Delta }U^p\mathrm{\Delta }w^q`$ in (57) as “component functions”. The expansion ends at fourth order, because $`\mathrm{\Delta }U,\mathrm{\Delta }w`$ are $`R`$-symmetric and are therefore power series in $`(\mathrm{\Pi }\mathrm{\Lambda })`$. Since the spinors are two-component objects and since there are two $`\mathrm{\Pi }`$’s and two $`\mathrm{\Lambda }`$’s it follows that the highest possible power is $`(\mathrm{\Pi }\mathrm{\Lambda })^4`$.
Let us investigate the linear level in the Taylor expansion. We ask whether singularities in $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }W`$ can conspire to cancel or whether the latter are independent objects. The component functions depend on $`s,t,\widehat{v}`$, which yields a linear dependence problem with coefficients in the ring of functions of $`s,t,\widehat{v}`$.
The functional form of these coefficients can be made much more explicit. Given the form of $`F`$ (58), by commuting all $`R`$’s to the left and all $`_t`$’s to the right we can show that the general component functions at $`k`$-th level have the form
$$\frac{1}{(\overline{w}(\overline{w}R))^k}\underset{n=0}{\overset{2+2k}{}}c_n(s,t)\widehat{v}^n.$$
(61)
The extra factors $`\overline{w}`$ in the denominator are introduced by the coefficients of the $`_T`$ derivatives and obviously factor out in the pure $`(_U)^k`$ component functions. A direct computer calculation shows them to cancel in the other components, too. Without enhanced factorisation abilities this point is hard to show and therefore we keep the $`\overline{w}`$’s in the scheme. Incidentally, by introducing $`\tau =1/T`$ the component functions can be more easily computed: They are simply
$$\frac{1}{t^n}_\tau ^n_U^mF.$$
(62)
We now examine the first order independence problem in detail. We work to lowest order, so the hat on $`v`$ is left out in the following and $`R_0`$ denotes the body of $`R`$. For charge two operators analyticity requires that there be functions of $`(s,t)`$ for $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }w`$ such that
$$\mathrm{\Delta }U\frac{1}{w(wR_0)}(\underset{n=0}{\overset{4}{}}c_{0n}(s,t)v^n)+\mathrm{\Delta }w\frac{1}{w(wR_0)}(\underset{n=0}{\overset{4}{}}c_{1n}(s,t)v^n)=O(\frac{1}{(\widehat{y}_{12}\widehat{y}_{34})^2})$$
(63)
with $`w=1+v`$.
Spinors occur in $`\mathrm{\Delta }U,\mathrm{\Delta }w`$ in combinations of the form $`(\mathrm{\Pi }xx^1x\mathrm{\Lambda })`$. Minkowski space is four-dimensional; a basis consists of four independent elements. In order to express the various $`x`$-triples in a given basis, we use
$$x_ix_j^{}+x_jx_i^{}=2(x_i.x_j)\delta $$
(64)
to commute $`x_{12}`$ left of $`x_{23}`$ left of $`x_{34}`$. ($`(x_i.x_j)`$ means the dot product of the associated four-vectors.) In this way the spinors are seen to be contracted on elements of the basis $`\{x_{12},x_{23},x_{34},x_{12}x_{23}^{}x_{34}\}`$.
There are two $`\mathrm{\Pi }_i`$ and two $`\mathrm{\Lambda }_i`$ so that we get a total of sixteen independent structures at first order in $`(\mathrm{\Pi }\mathrm{\Lambda })`$. A dependence relation between $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }w`$ is a linear combination with scalar coefficients which is of a certain order in $`\widehat{y}`$-singularities in all sixteen components separately. We restrict the analysis to the $`(\mathrm{\Pi }_1\mathrm{\Lambda }_2)`$ part of the odd expansion:
$`\mathrm{\Delta }U|_{\mathrm{\Pi }_1\mathrm{\Lambda }_2}=`$ $`{\displaystyle \frac{\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}}}\mathrm{\Pi }_1(x_{34}^2x_{12}+x_{14}^2x_{23}+x_{12}^2x_{34}+x_{12}x_{23}^{}x_{34})\mathrm{\Lambda }_2,`$ (65)
$`\mathrm{\Delta }w|_{\mathrm{\Pi }_1\mathrm{\Lambda }_2}=`$ $`{\displaystyle \frac{\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}}}\mathrm{\Pi }_1(x_{34}^2x_{12}+x_{12}^2x_{34}+x_{12}x_{23}^{}x_{34})\mathrm{\Lambda }_2`$ (66)
which we write in short as
$`\mathrm{\Delta }U|_{\mathrm{\Pi }_1\mathrm{\Lambda }_2}=`$ $`{\displaystyle \frac{\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}}}\mathrm{\Pi }_1X_U\mathrm{\Lambda }_2,`$ (67)
$`\mathrm{\Delta }w|_{\mathrm{\Pi }_1\mathrm{\Lambda }_2}=`$ $`{\displaystyle \frac{\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}}}\mathrm{\Pi }_1X_w\mathrm{\Lambda }_2.`$ (68)
On Minkowski space we change basis to $`\{X_U,X_w,x_{12},x_{34}\}`$ upon which the equation (63) breaks into two separate parts. We can omit the spinors and the $`x`$-vectors from the discussion as they have to be equal on both sides of the equations. This leads to the scalar equation
$$\frac{\widehat{y}_{23}}{\widehat{y}_{12}\widehat{y}_{34}}(\underset{i=0}{\overset{4}{}}c_n(s,t)v^n=(1+v)(s+v(1+st)+v^2)O(\frac{1}{(\widehat{y}_{12}\widehat{y}_{34})^2})$$
(69)
for both the $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }W`$ parts.
Instead of the variables $`\{\widehat{y}_{12},\widehat{y}_{23},\widehat{y}_{34}\}`$ we may choose the set $`\{v,Y_p=\widehat{y}_{23}/(\widehat{y}_{12}\widehat{y}_{34}),\widehat{y}_{23}\}`$. The Jacobian of the transform is
$$J=\frac{\widehat{y}_{23}^2}{(\widehat{y}_{12}\widehat{y}_{34})^3}(\widehat{y}_{12}\widehat{y}_{34})$$
(70)
and is regular at a generic point. Given this choice, the L.H.S. of (69) is independent of $`\widehat{y}_{23}`$ and factors out a single power of $`Y_p`$. We can conclude that the same is true for the R.H.S. and hence the as yet unspecified term is a function of $`s,t,v`$ of maximum order $`1/(\widehat{y}_{12}\widehat{y}_{34})`$. It must be a polynomial of degree $`1`$ in $`v`$.
It follows that the ansatz polynomial on the L.H.S. factors in the same way and $`w(wR_0)`$ cancels from the equation. The only solution of the regularity problem at first order is therefore the trivial one:
$$\mathrm{\Delta }U(\underset{n=0}{\overset{1}{}}g_n(s,t)v^n)+\mathrm{\Delta }w(\underset{n=0}{\overset{1}{}}h_n(s,t)v^n)=O(\frac{1}{(\widehat{y}_{12}\widehat{y}_{34})^2})$$
(71)
This is a general solution because $`\mathrm{\Delta }U`$ and $`\mathrm{\Delta }w`$ are not more singular than $`1/(\widehat{y}_{12}\widehat{y}_{34})`$ in any of the components w.r.t. the sixteen independent structures at first order. By inspection, the higher order terms arising from the second order in $`\mathrm{\Delta }U,\mathrm{\Delta }W`$ and from the soul of $`\widehat{v}`$ will also be regular.
Let us state the result again: The explicit forms for the level one component functions in the Taylor expansion (57) must have the dependence on $`\widehat{v}`$ indicated by the above equation. Hence
$$\frac{DF_o}{R}=g_1+\widehat{v}g_2$$
(72)
and
$$\frac{(D+R_t)F_o}{\overline{w}R}=h_1+\widehat{v}h_2.$$
(73)
We ought to stress that the proof makes use of certain regularity assumptions. Throughout the calculation we have assumed the $`x`$-scalars and in particular $`s`$ and $`t`$ to be regular. Hence none of the four points can be light-like separated in Minkowski space. This can possibly be relaxed. A necessary assumption is certainly that the vector basis is non-degenerate, so in particular the points do not coincide. For the $`\widehat{y}`$-coordinates we must also demand that the points are distinct. For the Jacobian (70) to be non-vanishing we additionally need $`\widehat{y}_{12}\widehat{y}_{34}`$.
If $`\widehat{y}_{34}=\alpha \widehat{y}_{12}`$ we can change from $`\{\widehat{y}_{12},\widehat{y}_{23}\}`$ to $`\{\widehat{y}_{23}/(\alpha \widehat{y}_{12}^2),v\}`$. In this case (69) still has the same consequences: The prefactor must occur on both sides and drops out. The unknown part on the R.H.S. is a function of $`s,t,v`$ as before. The Jacobian of the transformation is
$$J=\frac{\widehat{y}_{23}}{\alpha ^2\widehat{y}_{12}^5}((1+\alpha )\widehat{y}_{12}+2\widehat{y}_{23})$$
(74)
and hence is regular if the points do not coincide and $`\widehat{y}_{12}`$ is not proportional to $`\widehat{y}_{23}`$, so if there is more than one difference variable. The argument holds in fact as long as two of the $`\widehat{y}_{ij}`$ are independent. Otherwise $`v`$ is a constant and there is nothing to discuss.
The generalisation of the independence problem (63) to the higher levels like $`\mathrm{\Delta }U^2,`$ $`\mathrm{\Delta }U\mathrm{\Delta }w,\mathrm{}`$ is obvious. We have investigated this in the same manner to lowest order and we find that the spinor combinations are not independent. Additionally, the simple argument does not apply, which allowed us to ignore higher orders stemming from the lower level terms in the Taylor expansion (57). These terms define a non-trivial right hand side in the equivalent of (63) in the levels above. Due to linearity it suffices to find one special solution to this inhomogeneous problem which is to be added to the general solution of the homogeneous one.
To analyse equation (72) we begin by multiplying it by $`(1+\widehat{v})R`$ which gives a polynomial equation in $`\widehat{v}`$. This allows us to express the unknown functions $`g_1,g_2`$ in terms of $`a_1,a_2,a_3`$,
$$g_1=a_{1s},g_2=a_{1s}+a_{1t}a_{2t}+a_{3t}$$
(75)
where the literal subscripts denote partial derivatives. This leaves two first-order partial differential equations for $`a_1,a_2,a_3`$,
$`(ts1)a_{1s}+(ts)a_{1t}+s(a_{2s}+a_{2t})+a_2sa_{3t}`$ $`=`$ $`0,`$
$`a_{1s}a_{1t}+a_{2t}+sa_{3s}+(t1)a_{3t}+2a_3`$ $`=`$ $`0.`$
If we make the change of variables
$`a_1`$ $`=`$ $`\alpha +\gamma +sa_3,`$
$`a_2`$ $`=`$ $`\gamma +(st+1)a_3`$ (77)
with $`a_3`$ unchanged, we find that $`a_3`$ drops out of the above equations altogether - it is completely undetermined. The equations are satisfied if
$`\alpha _s+\alpha _t+\gamma _s`$ $`=`$ $`0,`$
$`t\gamma _t+\gamma +(1s)\alpha _ts\alpha _s`$ $`=`$ $`0.`$ (78)
This pair of first-order coupled differential equations can be equivalently rewritten as a set of second-order independent ones:
$`s\alpha _{ss}+t\alpha _{tt}+(s+t1)\alpha _{st}+2(\alpha _s+\alpha _t)`$ $`=`$ $`0,`$
$`s\gamma _{ss}+t\gamma _{tt}+(s+t1)\gamma _{st}+2(\gamma _s+\gamma _t)`$ $`=`$ $`0.`$ (79)
Note that in the correlator there is still one arbitrary coefficient function ($`a_3(s,t)`$ for this choice of variables) in addition to the solutions to these equations.
When carrying out this type of analysis for the higher level regularity problems, we find that the dependence relations between $`\mathrm{\Delta }U^k`$ etc. introduce so many unknowns into the equations that no new constraints are found. We have done these calculations for operator weight one through three with the result that the only constraints arise from the first level.
## 3 $`SU(2)`$-covariant approach
In this section we explain how the four-point function results obtained in the first part of this paper can be found in an independent way in a harmonic superspace formulation which maintains the explicit $`SU(2)`$ covariance. The technique is quite different, the main point being that here we shall keep Q supersymmetry (including its $`SU(2)`$ automorphism) manifest at each step. However, S supersymmetry, as well as harmonic analyticity will have to be checked level by level in the $`\theta `$ expansion of the four-point function. The advantage of this approach is the possibility to reformulate the H-analyticity condition in an equivalent way which will allow us to essentially eliminate the dependence on the G-analytic Grassmann variables. After this the constraints at levels 1 and 2 become sufficiently easy to work out. Moreover, it becomes obvious that there are no constraints beyond level 2.
### 3.1 $`SU(2)`$ harmonics and Grassmann analyticity
As discussed in the previous section $`N=2`$ harmonic superspace is the product of super Minkowski space (coordinates $`(x^\mu x^{\alpha \dot{\alpha }},\theta _i^\alpha ,\overline{\theta }^{\dot{\alpha }i})`$) and the two-sphere. However, in this section, rather than using explicit coordinates $`(y,\overline{y})`$ for the sphere, we shall use an alternative approach first proposed in in which the sphere is described by harmonic variables $`u_i^\pm `$ defined as the two columns of an $`SU(2)`$ matrix; the index $`i`$ transforms under the (left) $`SU(2)`$ and $`\pm `$ under an independent (right) $`U(1)`$ group. The components of $`u`$ have the defining properties:
$$\left(\begin{array}{cc}u_1^+& u_1^{}\\ u_2^+& u_2^{}\end{array}\right)SU(2)u_i^{}=(u^{+i})^{},u^{+i}u_i^{}=1$$
(80)
where the $`SU(2)`$ indices are raised and lowered in the following way, $`f^i=ϵ^{ij}f_j,f_i=ϵ_{ij}f^j`$ with $`ϵ_{ij}`$ defined by $`ϵ_{12}=ϵ^{12}=1`$.
The harmonic functions $`f^q(u_i^\pm )`$ are defined as singlets of the left $`SU(2)`$ but they are homogeneous of degree $`q`$ under the right $`U(1)`$, i.e. carry a charge $`q`$. Effectively, such functions live on the coset $`SU(2)/U(1)S^2`$ and are assumed to have a harmonic expansion on the sphere. A powerful feature of the coordinateless parametrisation of $`S^2`$ in terms of harmonics $`u_i^\pm `$ is the possibility to write down such expansions in a manifestly $`SU(2)`$ covariant way, e.g., for $`q0`$:
$$f^q(u)=\underset{n=0}{\overset{\mathrm{}}{}}f^{(i_1\mathrm{}i_{n+q}j_1\mathrm{}j_n)}u_{i_1}^+\mathrm{}u_{i_{n+q}}^+u_{j_1}^{}\mathrm{}u_{j_n}^{}.$$
(81)
The coefficients in this expansion are totally symmetric multispinors, i.e. irreps of $`SU(2)`$ of isospins $`q/2+n,n=0,1,\mathrm{}`$. Thus, using harmonic variables allows one to deal with $`U(1)`$ covariant objects without loosing the $`SU(2)`$ symmetry.
With the aid of $`u`$ we can define Grassmann analyticity in an $`SU(2)`$-covariant way. We split the Grassmann variables $`\theta _i^\alpha ,\overline{\theta }^{\dot{\alpha }i}`$ into two $`U(1)`$ projections,
$$\theta ^{\pm \alpha }=u_i^\pm \theta ^{i\alpha },\overline{\theta }^{\pm \dot{\alpha }}=u_i^\pm \overline{\theta }^{i\dot{\alpha }},$$
(82)
still maintaining the $`SU(2)`$ invariance. We then define a G-analytic function $`\varphi `$ on harmonic superspace to be one which satisfies
$$D_\alpha ^+\varphi =\overline{D}_{\dot{\alpha }}^+\varphi =0$$
(83)
where the $`U(1)`$ projections of the supercovariant derivatives are defined in a similar way. These constraints are solved by
$$\varphi =\varphi (x_A,\theta ^{+\alpha },\overline{\theta }^{=\dot{\alpha }})$$
(84)
where
$$x_A^{\alpha \dot{\alpha }}=x^{\alpha \dot{\alpha }}4i\theta ^{(i\alpha }\overline{\theta }^{j)\dot{\alpha }}u_i^+u_j^{}$$
(85)
with $`x^{\alpha \dot{\alpha }}=x^\mu \sigma _\mu ^{\alpha \dot{\alpha }}`$. Clearly G-analyticity is a Q-supersymmetric notion because it involves the supercovariant derivatives.
As explained in the previous section, complexified analytic superspace can be defined as a coset space of the complex superconformal group. Although this is not possible in real spacetime we can nevertheless define a representation of the Lie superalgebra of $`SU(2,2|2)`$ on $`G`$-analytic fields. For all practical purposes this amounts to taking infinitesimal $`SU(2)`$ transformations (parameters $`\lambda ^{jk}`$) to have the form $`u_i^\pm `$:
$$\delta u_i^+=(\lambda ^{jk}u_j^+u_k^+)u_i^{},\delta u_i^{}=0.$$
(86)
More generally, it is sufficient to treat the transformations of the harmonic variables $`u_i^\pm `$ with generators $`K,D,R,S,I`$ as active ones. For instance, a harmonic function $`f^{(p)}(u)`$ of weight $`p`$ will transform as follows:
$`\delta f^{(p)}(u)`$ $`=`$ $`f^{}{}_{}{}^{(p)}(u)f^{(p)}(u)`$ (87)
$`=`$ $`(\lambda ^{ij}u_i^+u_j^+)u_k^{}{\displaystyle \frac{}{u_k^+}}f^{(p)}(u)+p(\lambda ^{ij}u_i^{}u_j^+)f^{(p)}(u),`$
so that the non-unitary transformation appears in the form of a derivative of a function of unitary harmonics.
Henceforth we shall be dealing exclusively with G-analytic fields so that we can replace $`x_A`$ by $`x`$ without loss of clarity. The actions of the supersymmetry transformations on the coordinates are given by
$`\delta _Qx^{\alpha \dot{\alpha }}`$ $`=`$ $`4iu_i^{}(ϵ^{i\alpha }\overline{\theta }^{+\dot{\alpha }}+\theta ^{+\alpha }\overline{ϵ}^{i\dot{\alpha }})`$
$`\delta _Q\theta ^{+\alpha ,\dot{\alpha }}`$ $`=`$ $`u_i^+ϵ^{i\alpha ,\dot{\alpha }}`$
$`\delta _Qu_i^\pm `$ $`=`$ $`0`$ (88)
and
$`\delta _Sx^{\alpha \dot{\alpha }}`$ $`=`$ $`4i(x^{\alpha \dot{\beta }}\overline{\theta }^{+\dot{\alpha }}\overline{\eta }_{\dot{\beta }}^ix^{\dot{\alpha }\beta }\theta ^{+\alpha }\eta _\beta ^i)u_i^{}`$
$`\delta _S\theta ^{+\alpha }`$ $`=`$ $`2i(\theta ^+)^2\eta ^{\alpha i}u_i^{}+x^{\alpha \dot{\beta }}\overline{\eta }_{\dot{\beta }}^iu_i^+`$
$`\delta _S\theta ^\alpha `$ $`=`$ $`4i\eta _\beta ^i\theta ^\beta (\theta ^\alpha u_i^+\theta ^{+\alpha }u_i^{})+\overline{\eta }_{\dot{\beta }}^i(x^{\alpha \dot{\beta }}+4i\theta ^\alpha \overline{\theta }^{+\dot{\beta }})u_i^{}`$
$`\delta _Su_i^+`$ $`=`$ $`[4i(\theta ^{+\alpha }\eta _\alpha ^i+\overline{\eta }_{\dot{\alpha }}^i\overline{\theta }^{+\dot{\alpha }})u_i^+]u_i^{}`$
$`\delta _Su_i^{}`$ $`=`$ $`0`$ (89)
($`\delta _S\overline{\theta }^\pm `$ are obtained by conjugation). From this one can compute the action of the rest of the superconformal algebra by commuting Q and S supersymmetry transformations.
### 3.2 The hypermultiplet
In this subsection we give the harmonic formulation of the hypermultiplet describing an $`SU(2)`$ doublet of scalars $`f^i(x)`$ and a pair of Weyl (complex) spinors $`\psi _\alpha (x),\overline{\xi }^{\dot{\alpha }}(x)`$ on shell. Off shell it can only exist with an infinite set of auxiliary fields . Such a set is naturally provided by the G-analytic superfield of $`U(1)`$ charge $`+1`$:
$$q^+(x,\theta ^+,\overline{\theta }^+,u)=F^+(x,u)+\theta ^{+\alpha }\mathrm{\Psi }_\alpha (x,u)+\overline{\theta }_{\dot{\alpha }}^+\overline{\mathrm{\Xi }}^{\dot{\alpha }}(x,u)+\mathrm{}+(\theta ^+)^2(\overline{\theta }^+)^2P^3(x,u).$$
(90)
The components of this $`\theta ^+`$ expansion are harmonic functions with infinite expansions on $`S^2`$ (see (81)):
$`F^+(x,u)`$ $`=`$ $`f^i(x)u_i^++f^{(ijk)}(x)u_i^+u_j^+u_k^{}+\mathrm{}`$
$`\mathrm{\Psi }_\alpha (x,u)`$ $`=`$ $`\psi _\alpha (x)+\psi _\alpha ^{(ij)}(x))u^+_iu^{}_j+\mathrm{}`$
$`\mathrm{}`$
$`P^3(x,u)`$ $`=`$ $`p^{(ijk)}(x)u_i^{}u_j^{}u_k^{}+\mathrm{}`$ (91)
The coefficients in these expansions are ordinary fields belonging to different $`SU(2)`$ representations. All of them, with the exception of the physical fields $`f^i(x),\psi _\alpha (x),\overline{\xi }^{\dot{\alpha }}(x)`$ are auxiliary, so they should vanish on shell. We need a way to write down a supersymmetric on-shell constraint on the G-analytic superfield $`q^+(x,\theta ^+,\overline{\theta }^+,u)`$.
The key to this on-shell constraint is provided by the notion of harmonic (H-)analyticity. The harmonic coset $`SU(2)/U(1)`$ has two real (or one complex) dimensions, to which correspond the following harmonic derivatives:
$`^{++}=u_i^+{\displaystyle \frac{}{u_i^{}}}`$ $``$ $`^{++}u_i^+=0,^{++}u_i^{}=u_i^+,`$
$`^{}=u_i^{}{\displaystyle \frac{}{u_i^+}}`$ $``$ $`^{}u_i^+=u_i^{},^{}u_i^{}=0.`$ (92)
These are Cartan’s covariant derivatives on the coset. In our context this simply means that they preserve the defining condition $`u^{+i}u_i^{}=1`$. To them one may add the charge-counting operator
$$^0=u_i^+\frac{}{u_i^+}u_i^{}\frac{}{u_i^{}}^0u_i^\pm =\pm u_i^\pm .$$
(93)
By definition all harmonic functions are eigenfunctions of $`^0`$, $`^0f^q(u)=qf^q(u)`$.
An important point is that the three covariant derivatives above form an $`SU(2)`$ algebra:
$$[^{++},^{}]=^0,[^0,^{\pm \pm }]=\pm 2^{\pm \pm }.$$
(94)
They can be regarded as the generators of right $`SU(2)_R`$ rotations acting on the indices $`\pm `$ of the harmonics $`u_i^\pm `$. Thus, $`^{++}`$ is the raising and $`^{}`$ the lowering operator of $`SU(2)_R`$ (see (92)). This observation suggests the way to define short harmonic functions as highest weights of irreps of $`SU(2)_R`$. Thus, depending on the value of the $`U(1)`$ charge, the condition
$$^{++}f^q(u)=0\{\begin{array}{cc}f^q(u)=0,q<0\hfill & \\ f^q(u)=u_{i_1}^+\mathrm{}u_{i_q}^+f^{(i_1\mathrm{}i_q)},q0\hfill & \end{array}$$
(95)
has either a trivial solution or defines an irrep of isospin $`q/2`$. This property is a direct consequence of the general form (81) of the harmonic expansion on $`S^2`$ and of the action of $`^{++}`$ on the harmonics (92).
An alternative interpretation of the condition (95) is that of harmonic (H-)analyticity. Let us introduce stereographic coordinates on the sphere (see for more detail):
$$\left(\begin{array}{cc}u_1^+& u_1^{}\\ u_2^+& u_2^{}\end{array}\right)=\frac{1}{\sqrt{1+y\overline{y}}}\left(\begin{array}{cc}1& \overline{y}\\ y& 1\end{array}\right).$$
(96)
Our harmonic functions $`f^q(u)`$ are by definition eigenfunctions of the charge “operator” $`^0`$:
$$^0f^q(y,\overline{y})=qf^q(y,\overline{y}).$$
(97)
In this parametrisation the covariant derivative $`^{++}`$ becomes
$$^{++}=(1+y\overline{y})\frac{}{\overline{y}}\frac{y}{2}^0.$$
(98)
In these terms eq. (95) takes the form of a (covariant) harmonic analyticity condition:
$$\frac{f^q}{\overline{y}}+\frac{qy}{2(1+y\overline{y})}f^q=0.$$
(99)
It admits the general solution $`f^q(y,\overline{y})=(1+y\overline{y})^{\frac{q}{2}}f_0(y)`$ where $`f_0(y)`$ is an arbitrary holomorphic function. Remembering that we are looking for solutions globally defined on the sphere it is not hard to show that for $`q<0`$ the only solution is $`f_0=0`$ and for $`q0`$ $`f_0(y)`$ must be a polynomial of degree $`q`$ whose $`SU(2)`$ covariant form is given in (95).
It should be stressed that the above H-analytic harmonic functions are regular, i.e. well-defined on the whole of $`S^2`$. In practice one has also to deal with singular harmonic functions. A typical example we shall encounter in what follows is the harmonic distribution
$$\frac{1}{(12)}$$
(100)
where
$$(12)u_1^{+i}u_{2i}^+=\frac{y_2y_1}{\sqrt{(1+y_1\overline{y}_1)(1+y_2\overline{y}_2)}}.$$
(101)
At first sight, it is a function of $`u^+`$ only, therefore one would expect $`_1^{++}(12)^1=0`$. However, this distribution is singular at the point $`u_1=u_2`$, so it should be differentiated with care:
$`_1^{++}{\displaystyle \frac{1}{(12)}}`$ $`=`$ $`(1+y_1\overline{y}_1)^{3/2}(1+y_2\overline{y}_2)^{1/2}{\displaystyle \frac{}{\overline{y}_1}}{\displaystyle \frac{1}{y_1y_2}}`$ (102)
$`=`$ $`(1+y_1\overline{y}_1)^2i\pi \delta (y_1y_2)`$
$`=`$ $`(12^{})\delta (u_1,u_2),`$
where we have used the well-known relation
$$\frac{}{\overline{y}}\frac{1}{y}=i\pi \delta (y).$$
(103)
Note that the factor $`(12^{})u_1^{+i}u_{2i}^{}`$ is needed for keeping the balance of charges on both sides of eq. (102) in the $`S(2)`$ covariant notation.
The conclusion from the above discussion is that the condition (95) is, on the one hand, the definition of a highest weight of an $`SU(2)`$ irrep and, on the other hand, a harmonic analyticity condition on the sphere. This type of condition can easily be supersymmetrised in order to be applied to superfields such as the hypermultiplet $`q^+`$ (90) and we obtain the following operator invariant under Q supersymmetry:
$$D^{++}=^{++}2i\theta ^{+\alpha }\overline{\theta }^{+\dot{\alpha }}_{\alpha \dot{\alpha }}$$
(104)
where $`_{\alpha \dot{\alpha }}\sigma _{\alpha \dot{\alpha }}^\mu /x^\mu `$.
Now, let us impose the (supercovariant) H-analyticity condition
$$D^{++}q^+(x,\theta ^+,\overline{\theta }^+,u)=0.$$
(105)
Inserting the expansion (90) into eq. (105), we obtain a set of harmonic differential equations which are solved just like eq. (95). The result is the short (on-shell) hypermultiplet
$$q^+=f^i(x)u_i^++\theta ^{+\alpha }\psi _\alpha (x)+\overline{\theta }_{\dot{\alpha }}^+\overline{\xi }^{\dot{\alpha }}(x)+2i\theta ^{+\alpha }\overline{\theta }^{+\dot{\alpha }}_{\alpha \dot{\alpha }}f^i(x)u_i^{}$$
(106)
where all the auxiliary fields have been eliminated and the remaining physical ones put on shell,
$$\mathrm{}f^i(x)=/\psi =/\overline{\xi }=0.$$
So, in the case of the hypermultiplet the combination of G- and H-analyticities results in an on-shell superfield. Note that this result crucially depends on the $`U(1)`$ charge of the G-analytic superfield. For example, a superfield $`L^{++}(x,\theta ^+,\overline{\theta }^+,u)`$ of charge $`+2`$ subject to the same H-analyticity condition
$$D^{++}L^{++}=0$$
(107)
describes an off-shell multiplet (the linear or tensor multiplet consisting of a triplet of real scalars, a divergenceless real vector, a Majorana spinor and a complex auxiliary field). For charges $`+3`$ the H-analyticity condition simply cuts off the tail of auxiliary fields without imposing any constraints on the remaining physical fields. On the contrary, for charges $`0`$ the condition is too strong and only admits a trivial solution.
A very important observation is that the H-analyticity conditions (105) or (107) admit an equivalent form in terms of the harmonic derivative $`^{}`$. Remembering that $`^{++}`$ and $`^{}`$ are the raising and lowering operators of $`SU(2)_R`$ and that the H-analyticity condition $`^{++}f^q(u)=0`$ defines the highest weight of an $`SU(2)_R`$ irrep of isospin $`q/2`$ (dimension $`q+1`$), we immediately see the equivalence relation
$$^{++}f^q(u)=0(^{})^{q+1}f^q(u)=0$$
(108)
(alternatively, it can be derived by inspecting the harmonic expansion (81)). The supersymmetric version of the new form of the H-analyticity condition involves the operator
$$D^{}=^{}2i\theta ^\alpha \overline{\theta }^{\dot{\alpha }}_{\alpha \dot{\alpha }}+\theta ^\alpha \frac{}{\theta ^{+\alpha }}+\overline{\theta }^{\dot{\alpha }}\frac{}{\overline{\theta }^{+\dot{\alpha }}}.$$
(109)
There is a crucial difference between these two conditions, which we shall heavily exploit in what follows. The point is that singular harmonic functions of the type (100) give rise to delta-type singularities under $`^{++}`$ (see (102)), whereas they can be differentiated as ordinary functions by $`^{}`$, e.g.
$$^{}\frac{1}{(12)}=\frac{(1^{}2)}{(12)^2}.$$
(110)
The explanation is that in the former case we deal with a derivative of the type $`/\overline{y}y^1=i\pi \delta (y)`$ and in the latter $`/yy^1=y^2`$.
In the rest of this section we shall examine the non-trivial implications of H-analyticity combined with the requirement of superconformal covariance for correlation functions of charge $`+2`$. For that purpose we shall need the superconformal transformation properties of the superfields and operators we have introduced. The transformation law of the harmonic derivatives $`D^{++}`$ and $`D^{}`$ can be found using Cartan’s coset scheme (or checked directly ):
$`\delta D^{++}`$ $`=`$ $`\mathrm{\Lambda }^{++}D^0,`$
$`\delta D^{}`$ $`=`$ $`(D^{}\mathrm{\Lambda }^{++})D^{}`$ (111)
where
$$D^{++}\mathrm{\Lambda }=\mathrm{\Lambda }^{++},D^{++}\mathrm{\Lambda }^{++}=0$$
(112)
and
$$\mathrm{\Lambda }=a+k_{\alpha \dot{\alpha }}x^{\alpha \dot{\alpha }}+\lambda ^{ij}u_i^+u_j^{}+4i(\theta ^{+\alpha }\eta _\alpha ^i+\overline{\eta }_{\dot{\alpha }}^i\overline{\theta }^{+\dot{\alpha }})u_i^{}$$
(113)
is the superconformal weight factor. For completeness, besides the S supersymmetry parameter $`\eta `$ we have also included those of dilation $`a`$, conformal boosts $`k^\mu `$ and $`SU(2)_C`$ $`\lambda ^{ij}`$. Then it is not hard to check that the H-analyticity condition
$$D^{++}q^+=0(D^{})^2q^+=0$$
(114)
is covariant if the hypermultiplet transforms with superconformal weight $`+1`$:
$$\delta q^+=\lambda q^++\mathrm{\Lambda }q^+$$
(115)
(here $`\lambda `$ denotes the coordinate transformations). Similarly, the linear multiplet subject to the H-analyticity condition
$$D^{++}L^{++}=0(D^{})^3L^{++}=0$$
(116)
should have weight $`+2`$:
$$\delta L^{++}=\lambda L^{++}+2\mathrm{\Lambda }L^{++}.$$
(117)
### 3.3 Two- and three-point functions
The simplest example of a two-point function is the hypermultiplet propagator
$$G^{(1,1)}(1|2)=\stackrel{~}{q}^+(1)q^+(2)$$
(118)
where the superscript $`(1,1)`$ indicates the $`U(1)`$ charges at the two points and $`\stackrel{~}{}`$ is a special conjugation on $`S^2`$ preserving G-analyticity . It is defined as the Green’s function of the field equation (105):
$$D_1^{++}G^{(1,1)}(x_1,\theta _1^+,u_1|x_2,\theta _2^+,u_2)=\delta ^4(x_1x_2)(\theta _1^+(u_1^+u_2^{})\theta _2^+)^4(u_1^{}u_2^+)\delta (u_1,u_2).$$
(119)
Naturally, like the hypermultiplet superfield $`q^+`$ itself, the Green’s function should be G-analytic. The right-hand side of eq. (119) is the complete delta-function of the G-analytic superspace (as in eq. (102), the factors $`(u_1^+u_2^{})`$ and $`(u_1^+u_2^{})`$ maintain the balance of $`U(1)`$ charges). Throughout this paper we assume that all the correlation functions are considered away from the coincident points where they usually have singularities. In this case the right-hand side of eq. (119) just vanishes:
$$D_1^{++}G^{(1,1)}(1|2)=0\text{for points }12\text{.}$$
(120)
The same is of course true if we replace $`D_1^{++}`$ by $`D_2^{++}`$. In other words, this two-point function is H-analytic away from the singular point.
Another basic property of the hypermultiplet propagator is superconformal covariance. According to the transformation law (115) of the hypermultiplet itself, the propagator transforms as follows:
$$\delta G^{(1,1)}(1|2)=\lambda G^{(1,1)}(1|2)+(\mathrm{\Lambda }(1)+\mathrm{\Lambda }(2))G^{(1,1)}(1|2).$$
(121)
The combination of H-analyticity and the conformal properties of the propagator allow us to find the explicit expression for $`G^{(1,1)}`$. We start by examining the leading component of this superfield
$$g^{(1,1)}(x_{12}^2,u_1,u_2)=G^{(1,1)}(\theta _1^+=\theta _2^+=0).$$
(122)
Here we have taken into account translation and Lorentz invariance which tell us that the function must depend on the space-time invariant $`x_{12}^2(x_1x_2)^2`$. In the absence of $`\theta ^+`$ the harmonic derivative $`D^{++}^{++}`$, so the H-analyticity condition (120) simply tells us that $`g^{(1,1)}`$ must be linear in $`u_1^+`$ (recall (95)). At the same time it is an $`SU(2)`$ invariant, so the index $`i`$ of $`u_{1i}^+`$ must be contracted with the other harmonic variable $`u_2^\pm `$. Given the charges $`+1`$ at both points, we conclude that the only such invariant combination of harmonics is $`(12)u_i^{+i}u_{2i}^+`$. So, $`g^{(1,1)}`$ is reduced to
$$g^{(1,1)}=(12)g(x_{12}^2).$$
(123)
The remaining function $`g(x_{12}^2)`$ can be most easily determined by making use of the dilation part of the conformal group. The first component of the superfield $`q^+`$ is the physical scalar $`f^i(x)`$ which has conformal weight 1, and so does the leading term in the hypermultiplet propagator. The harmonic factor in (123) is weightless, so we conclude that $`g=C/x_{12}^2`$. The constant $`C`$ can be fixed by comparing with the standard scalar propagator and the result is
$$g^{(1,1)}=\frac{1}{4i\pi ^2}\frac{(12)}{x_{12}^2}.$$
(124)
Now, the less trivial part of the determination of the propagator $`G^{(1,1)}`$ is completing it to a full superfield, i.e. restoring the dependence on $`\theta _{1,2}^+`$. Here we shall use a trick which will prove very useful in the study of the four-point correlator in the next subsection. The two-point function is supposed invariant under Q supersymmetry (88), which acts as a shift of the Grassmann variables:
$$(\theta _{1,2}^{+\alpha ,\dot{\alpha }})^{}=\theta _{1,2}^{+\alpha ,\dot{\alpha }}+u_{1,2i}^+ϵ^{i\alpha ,\dot{\alpha }}.$$
(125)
It is then clear that by making a finite Q supersymmetry transformation with parameter
$$ϵ^{i\alpha ,\dot{\alpha }}=\frac{u_2^{+i}}{(12)}\theta _1^{+\alpha ,\dot{\alpha }}\frac{u_1^{+i}}{(12)}\theta _2^{+\alpha ,\dot{\alpha }}$$
(126)
we can eliminate both $`\theta _1^+`$ and $`\theta _2^+`$:
$$\text{Q frame:}(\theta _1^{+\alpha ,\dot{\alpha }})^{}=(\theta _2^{+\alpha ,\dot{\alpha }})^{}=0.$$
(127)
In this frame the two-point function becomes independent of the Grassmann variables. In other words, it coincides with its leading component (124), $`G^{(1,1)}|_Qg^{(1,1)}`$. Then we can go back to the original frame by performing the same finite Q supersymmetry transformations on the remaining coordinates.<sup>3</sup><sup>3</sup>3As a simpler example of this trick, consider translation invariance for a set of two space-time points $`x_1,x_2`$. By means of the finite translation $`P:x_2^{}=x_2+a=0`$ we can go to a $`P`$ frame in which only $`x_1`$ survives. Then, to restore manifest invariance, we make the same shift on $`x_1`$: $`x_1^{}=x_1+a=x_1x_2x_{12}`$. Such a transformation only affects the difference $`x_{12}`$ and gives
$$\widehat{x}_{12}^{\alpha \dot{\alpha }}=x_{12}^{\alpha \dot{\alpha }}+\frac{4i}{(12)}[(1^{}2)\theta _1^+\overline{\theta }_1^++(2^{}1)\theta _2^+\overline{\theta }_2^++\theta _1^+\overline{\theta }_2^++\theta _2^+\overline{\theta }_1^+]^{\alpha \dot{\alpha }}.$$
(128)
By construction, this modified coordinate difference is invariant under Q supersymmetry, which can be easily verified using (88). So, to find out the $`\theta _{1,2}^+`$ dependence of the hypermultiplet propagator, it is enough to replace $`x_{12}`$ by $`\widehat{x}_{12}`$:
$$G^{(1,1)}(1|2)=\frac{1}{4i\pi ^2}\frac{(12)}{\widehat{x}_{12}^2}.$$
(129)
In deriving this two-point function we have only used the dilation part of the superconformal group. In fact, since the result is unique, it is guaranteed to have the right superconformal properties (121) of the propagator (this can also be checked directly). Further, so far we have only solved the H-analyticity constraint (120) at the lowest (leading) order of the $`\theta ^+`$ expansion. One might try to argue that since the left-hand side of eq. (120) is itself an invariant of Q supersymmetry, it is sufficient to check H-analyticity in the Q frame (i.e., in the absence of $`\theta ^+`$). However, this argument is not safe here. Indeed, in the expansion of the two-point function there are harmonic singularities of pole type (e.g., $`(12)^1`$), on which the operator $`^{++}`$ creates a delta-type singularity (recall (102)). In such a situation we will not be allowed to use the supersymmetry parameter (126) which itself contains harmonic poles. A safe way to extend H-analyticity to all orders in the $`\theta ^+`$ expansion by means of the transformation (126) is to use the alternative form of the H-analyticity constraint involving $`D^{}`$ (see (114)). We shall come back to this important point in the next subsection.
Knowing the hypermultiplet propagator, we can easily predict the general form of correlators of two or three composite operators made out of hypermultiplets. Take, for instance, the two-point function
$$G^{(2,2)}(1|2)=\text{Tr}\left(\stackrel{~}{q}^+(1)\right)^2\text{Tr}\left(q^+(2)\right)^2.$$
(130)
Note that it has charges $`+2`$ at each point matching the number of elementary hypermultiplets in each composite operator. One can imagine this correlator in the context of $`N=4`$ super-Yang-Mills theory, where a hypermultiplet in the adjoint representation of the gauge group interacts with the $`N=2`$ super-Yang-Mills gauge potential. Since the $`N=4`$ theory is finite (conformally invariant), we can demand that the correlator (130) be superconformally covariant,
$$\delta G^{(2,2)}(1|2)=\lambda G^{(2,2)}(1|2)+2(\mathrm{\Lambda }(1)+\mathrm{\Lambda }(2))G^{(2,2)}(1|2).$$
(131)
In addition to this, the correlator should be H-analytic. Indeed, let us differentiate it with the harmonic derivative $`D^{++}`$.<sup>4</sup><sup>4</sup>4The traces in (130) make the composite operators gauge invariant, so we can use a flat $`D^{++}`$ (no gauge connection). Since $`D^{++}`$ is the operator of the free field equation (90) for the hypermultiplet, one can argue that such a differentiation will give rise to a Schwinger-Dyson equation for the correlator:
$$D_1^{++}G^{(2,2)}(1|2)=\text{contact terms}.$$
(132)
Since the composite operators are bilinear in this case (charges $`+2`$), equation (132) can also be interpreted as a Ward identity. Indeed, the bilinears $`\stackrel{~}{q}^+\stackrel{~}{q}^+`$, $`\stackrel{~}{q}^+q^+`$ and $`q^+q^+`$ are the currents of an extra $`SU(2)`$ symmetry of the $`N=4`$ theory realised in terms of $`N=2`$ superfields<sup>5</sup><sup>5</sup>5In fact, this symmetry is the visible part of the full $`SU(4)`$ R symmetry of the $`N=4`$ theory.. So, in this case eq. (132) corresponds to the current conservation law. In the context of this paper we treat contact terms as zeros, so eq. (132) takes the form of an H-analyticity condition:
$$D_1^{++}G^{(2,2)}(1|2)=0\text{for points }12$$
(133)
(and similarly at point 2). Since the product of two H-analytic functions is H-analytic as well, we immediately find an obvious solution to this constraint as the square of the hypermultiplet propagator,
$$G^{(2,2)}(1|2)=C\frac{(12)^2}{\widehat{x}_{12}^4}$$
(134)
where $`C`$ is a constant. In fact, this is the general solution. The argument is as in the case of the propagator. One first examines the leading component $`G^{(2,2)}(\theta _1^+=\theta _2^+=0)`$. The constraint (133) fixes the harmonic dependence since the combination $`(12)^2`$ is the only $`SU(2)`$ invariant of charges $`(2,2)`$ annihilated by $`^{++}`$. The dependence on $`x_{12}^2`$ is determined by simple dilation covariance. Finally, with two G-analytic Grassmann variables $`\theta _{1,2}^+`$ we already know that the complete $`\theta ^+`$ dependence is fixed by Q supersymmetry alone, by just putting a hat on $`x_{12}^2`$. It should be mentioned that the above considerations cannot predict the value of the constant in (134). In principle, it might receive quantum corrections at each level of perturbation theory, but it can be shown that this type of correlator is protected by a non-renormalisation theorem .
Next we turn to three-point correlators. As an example, take the correlator of three currents, i.e. bilinears made out of hypermultiplets:
$$G^{(2,2,2)}(1|2|3)=\text{Tr}\left(\stackrel{~}{q}^+(1)\right)^2\text{Tr}\left(\stackrel{~}{q}^+(2)q^+(2)\right)\text{Tr}\left(q^+(3)\right)^2$$
(135)
and subject to the requirements of H-analyticity
$$D_1^{++}G^{(2,2,2)}(1|2|3)=0\text{for points }123$$
(136)
and of superconformal covariance
$$\delta G^{(2,2,2)}(1|2|3)=\lambda G^{(2,2,2)}(1|2|3)+2(\mathrm{\Lambda }(1)+\mathrm{\Lambda }(2)+\mathrm{\Lambda }(3))G^{(2,2,2)}(1|2|3).$$
(137)
Just as for $`G^{(2,2)}(1|2)`$ above, it is obvious that the product of three propagators
$$G^{(2,2,2)}(1|2|3)=C\frac{(12)}{\widehat{x}_{12}^2}\frac{(23)}{\widehat{x}_{23}^2}\frac{(31)}{\widehat{x}_{31}^2}$$
(138)
satisfies both requirements. To prove its uniqueness, we argue as follows. Firstly, at the lowest level in the $`\theta ^+`$ expansion there is a single $`SU(2)`$ invariant combination of the three harmonics with the right charges and vanishing under $`D_1^{++}`$, namely $`(12)(23)(31)`$. Secondly, the space-time dependence is now determined by the full conformal group (and not just dilations, as for two points). It is well-known that there exists no conformal invariant made out of three space-time variables, therefore the product $`x_{12}^2x_{23}^2x_{31}^2`$ is the only function with the required conformal properties. Finally, we have to show that putting hats on the $`x`$’s gives the unique completion of the leading component to a full superfield. Before we did this by using the Q frame (127) in which the two Grassmann variables had been eliminated. Now we have three $`\theta ^+`$’s, and the Q supersymmetry parameter $`ϵ^i`$ alone is not enough to shift away all of them. This time we have to invoke S supersymmetry as well. Looking at the transformation law of $`\theta ^+`$ in (89) we see that S supersymmetry acts essentially as a shift (although non-linear), provided that the matrix $`x^{\alpha \dot{\alpha }}`$ is invertible. Then the combination of Q and S supersymmetry makes it possible to find a
$$\text{Q\&S frame:}(\theta _1^{+\alpha ,\dot{\alpha }})^{}=(\theta _2^{+\alpha ,\dot{\alpha }})^{}=(\theta _3^{+\alpha ,\dot{\alpha }})^{}=0$$
(139)
in which there are no $`\theta ^+`$’s left.<sup>6</sup><sup>6</sup>6In fact, the S supersymmetry parameter $`\eta ^i`$ is an $`SU(2)`$ doublet, just as $`ϵ^i`$. Using both of them we can shift away up to four $`\theta ^+`$, as we shall do in the four-point case. So, if there existed another superfield completion of the leading component above, their difference would be a nilpotent (i.e., proportional to $`\theta ^+`$) superconformal covariant. But such an object would vanish in the Q&S frame, therefore it must vanish in any frame. So, H-analyticity and superconformal covariance can predict the form of the three-point correlator up to a constant factor. Once again, it turns out protected by a non-renormalisation theorem .
### 3.4 Four-point correlators
#### 3.4.1 Preliminaries
The main subject of interest in this paper are four-point correlators of hypermultiplet bilinears of the type, e.g.,
$$G^{(2,2,2,2)}(1|2|3|4)=\text{Tr}\left(\stackrel{~}{q}^+(1)\right)^2\text{Tr}\left(q^+(2)\right)^2\text{Tr}\left(\stackrel{~}{q}^+(3)\right)^2\text{Tr}\left(q^+(4)\right)^2$$
(140)
satisfying the requirements of H-analyticity
$$D_1^{++}G^{(2,2,2,2)}(1|2|3|4)=0\text{for points }1234$$
(141)
and of superconformal covariance
$$\delta G^{(2,2,2,2)}(1|2|3|4)=\lambda G^{(2,2,2,2)}(1|2|3|4)+2(\mathrm{\Lambda }(1)+\mathrm{\Lambda }(2)+\mathrm{\Lambda }(3)+\mathrm{\Lambda }(4))G^{(2,2,2,2)}(1|2|3|4).$$
(142)
Compared to the two- and three-point cases above, the structure of the four-point correlator is considerably richer, for two main reasons which can be seen at the lowest level in the $`\theta ^+`$ expansion. Firstly, now there exist three independent harmonic combinations satisfying (141):
$$(12)^2(34)^2,(14)^2(23)^2,(12)(23)(34)(41).$$
(143)
Any other combination can be reduced to these by means of the harmonic cyclic identity
$$(12)(34)+(13)(42)+(14)(23)=0$$
(144)
following from the property of the $`ϵ^{ij}`$ contraction. Secondly, given four space-time points, one can construct two independent conformal invariants,<sup>7</sup><sup>7</sup>7Here is a simple explanation, very much in the spirit of the Q&S frame argument above. The translations $`P_\mu `$ and the conformal boosts $`K_\mu `$ act on $`x^\mu `$ as linear and non-linear shifts, correspondingly. So, the combined action of both of them can define a special P&K frame in which there are only two out of the four space-time variables $`x_{1,2,3,4}^\mu `$ left. Out of them we can make three Lorentz invariants (the two squares and the scalar product). Finally, dilation invariance requires that we take the two independent ratios of those. the cross-ratios
$$s=\frac{x_{14}^2x_{23}^2}{x_{12}^2x_{34}^2},t=\frac{x_{13}^2x_{24}^2}{x_{12}^2x_{34}^2}.$$
(145)
Consequently, the most general form of the leading component of the correlator (140) consistent with H-analyticity and conformal covariance is
$$\frac{(12)^2(34)^2}{x_{12}^4x_{34}^4}a(s,t)+\frac{(14)^2(23)^2}{x_{14}^4x_{23}^4}b(s,t)+\frac{(12)(23)(34)(41)}{x_{12}^2x_{23}^2x_{34}^2x_{41}^2}c(s,t).$$
(146)
Here we see the three independent harmonic structures (143) completed to product of propagators. Such products already have the required conformal properties of the correlator, so the only freedom left are the three arbitrary coefficient functions $`a,b,c`$ of the invariant cross-ratios. Our aim will be to find constraints on these functions following from the full implementation of H-analyticity combined with superconformal covariance.
The first step is to argue, just like in the three-point case, that there exists a special frame in superspace in which there are no $`\theta ^+`$ left:
$$\text{Q\&S frame:}(\theta _1^{+\alpha ,\dot{\alpha }})^{}=(\theta _2^{+\alpha ,\dot{\alpha }})^{}=(\theta _3^{+\alpha ,\dot{\alpha }})^{}=(\theta _4^{+\alpha ,\dot{\alpha }})^{}=0.$$
(147)
As explained above, this can be achieved by fully exploiting the four spinor parameters contained in the doublets $`ϵ^i`$ of Q and $`\eta ^i`$ of S supersymmetry to shift away all four $`\theta ^+`$. The existence of such a frame implies that the completion of the leading component (146) to a full superfield is always possible and is uniquely determined by Q and S supersymmetry. To obtain this completion one could, in principle, find the finite transformation to the frame (147). However, unlike the case of the linear Q supersymmetry, S supersymmetry acts on the coordinates in a very non-linear way and the practical realisation of this step is not at all easy. Fortunately, as we shall explain below, for our purposes we shall only need to know the first non-trivial level in the $`\theta ^+`$ expansion of the correlator.
The above discussion makes it clear that no further constraints on the coefficient functions $`a,b,c`$ originate from conformal supersymmetry alone. This only takes place when we try to impose H-analyticity. There are two possible approaches in doing so. One is to use the form (141) of the constraint and try to solve it level by level in the $`\theta ^+`$ expansion. The problem here is that this expansion is very complicated (assuming that we have already found it, which is in itself not an easy task). A much more efficient approach is to use the alternative form
$$(D_4^{})^3G^{(2,2,2,2)}(1|2|3|4)=0.$$
(148)
The advantage is that we can study this constraint in the Q&S frame (147) where there are no $`\theta ^+`$’s. This results in substantial technical simplifications. The same trick is not allowed in the form (141) because of the harmonic singularities (see the discussion after eq. (129)).
#### 3.4.2 An example of H-analyticity in the Q frame
In order to better understand the idea of this approach, we are going to redo the derivation of the propagator (129), but this time starting form the alternative form of the H-analyticity condition
$$(D_1^{})^2G^{(1,1)}(1|2)=0.$$
(149)
We begin by going to the Q frame (127). There the left-hand side of eq. (149) does not depend on $`\theta ^+`$ but can still depend on $`\theta ^{}`$. In particular, since the operator $`D^{}`$ converts $`\theta ^+`$ into $`\theta ^{}`$ (see (109)), some terms in the $`\theta ^+`$ expansion of $`G^{(1,1)}`$ may survive the transformation (125), (126). Therefore we should proceed in the following order.
Step 1. Expand $`G^{(1,1)}`$ in $`\theta _1^+`$ up to the order $`\theta _1^+\overline{\theta }_1^+`$ (still in the old frame):
$$G^{(1,1)}=g^{(1,1)}(x_{12}^2,u_1,u_2)+\theta _1^{+\alpha }\overline{\theta }_1^{+\dot{\alpha }}\gamma _{\alpha \dot{\alpha }}^{(1,1)}(x_{12}^2,u_1,u_2)+\mathrm{}$$
(150)
There is no need to keep terms containing $`\theta _2^+`$ or higher orders in $`\theta _1^+`$ because they cannot be “rescued” by $`(D_1^{})^2`$ and will vanish after the transformation to the Q frame. Indeed, the function $`G^{(1,1)}`$ carries no R weight, so the Grassmann variables have to appear in its expansion in pairs $`\theta ^+\overline{\theta }^+`$. So, only the term $`\theta _1^+\overline{\theta }_1^+\theta _1^{}\overline{\theta }_1^{}`$ can survive in the Q frame.
Step 2. Differentiate the expansion (150) with $`(D_1^{})^2`$ keeping only terms without any $`\theta ^+`$. The expansion of $`(D_1^{})^2`$ is
$$(D_1^{})^2=(_1^{})^24i\theta _1^{}/_1\overline{\theta }_1^{}_1^{}+(\theta _1^{}_1^{}+\overline{\theta }_1^{}\overline{}_1^{})^22(\theta _1^{})^2(\overline{\theta }_1^{})^2\mathrm{}_1.$$
(151)
We have dropped the terms linear in $`2\theta _1^{}_1^{}+\overline{\theta }_1^{}\overline{}_1^{}`$ because they only convert one $`\theta _1^+`$ into $`\theta _1^{}`$, and the remaining $`\theta _1^+`$ in the bilinear combination will vanish in the Q frame. When applied to (150), this operator gives
$`(D_1^{})^2G^{(1,1)}`$ $`=`$ $`(_1^{})^2g^{(1,1)}`$ (152)
$`+\theta _1^\alpha \overline{\theta }_1^{\dot{\alpha }}\left[2\gamma _{\alpha \dot{\alpha }}^{(1,1)}4i_{1\alpha \dot{\alpha }}_1^{}g^{(1,1)}\right]`$
$`2(\theta _1^{})^2(\overline{\theta }_1^{})^2\mathrm{}_1g^{(1,1)}`$
$`+\theta ^+\text{ terms}`$
$`=`$ $`0.`$
Step 3. Make the transformation to the Q frame. The left-hand side of eq. (152) is superconformally covariant, so it is multiplied by the weight factor $`\mathrm{\Lambda }(1)+\mathrm{\Lambda }(2)`$. Since it is supposed to vanish, this transformation just amounts to neglecting all the $`\theta ^+`$ dependence (already taken into account at steps 1 and 2).
The resulting constraint (152) involves three levels in its $`\theta ^{}`$ expansion.
Level 0 or $`(\theta ^{}\overline{\theta }^{})^0`$:
$$(_1^{})^2g^{(1,1)}(x_{12}^2,u_1,u_2)=0g^{(1,1)}=(12)g(x_{12}^2).$$
(153)
This constraint uniquely fixes the harmonic dependence of the component $`g^{(1,1)}`$. The combination of harmonics $`(12)u_1^{+i}u_{2i}^+`$ is the only one which is $`SU(2)`$ invariant, has the right charges and is annihilated by the lowering operator $`(_1^{})^2`$.
Level 1 or $`(\theta ^{}\overline{\theta }^{})^1`$:
$$\gamma _{\alpha \dot{\alpha }}^{(1,1)}(x_{12}^2,u_1,u_2)=2i_{1\alpha \dot{\alpha }}_1^{}g^{(1,1)}(x_{12}^2,u_1,u_2)=2i(12)_{1\alpha \dot{\alpha }}g(x_{12}^2).$$
(154)
Level 2 or $`(\theta ^{}\overline{\theta }^{})^2`$:
$$\mathrm{}_1g^{(1,1)}(x_{12}^2,u_1,u_2)=0\mathrm{}_1g(x_{12}^2)=0g(x_{12}^2)=\frac{C}{x_{12}^2}$$
(155)
where $`C`$ is an arbitrary constant. We recall that we are only interested in the two-point function away from the coincident point, so we can drop the delta-function $`\delta (x_{12})`$ in (155).
We should mention that in this example we have made no use of conformal invariance or S supersymmetry. Actually, this two-point function is in a sense overdetermined. We have already seen that by imposing H-analyticity just at level 0 and then invoking dilation covariance (part of the conformal symmetry), we arrived at the same result. This, however, is an exceptional property of the propagator (the charges $`+1`$ two-point function). The typical situation is illustrated by the charges $`+2`$ two-point function (134). It is not hard to show that it remains H-analytic to all orders in the $`\theta `$ expansion even if we replace the denominator by any function of $`\widehat{x}_{12}^2`$. So, its form cannot be determined without some extra input (dilation covariance in this case). The explanation of this fact can be traced back to the different implications of H-analyticity for superfields of charges $`+1`$ and $`+2`$: for the former it is an on-shell condition and for the latter an off-shell one (see (105), (107)).
Let us summarise the above example. Using the Q frame we have been able to solve the H-analyticity constraint (149) to all relevant orders in the $`\theta `$ expansion. In the process we only used the $`\theta ^+`$ expansion of the two-point function $`G^{(1,1)}`$ to the first non-trivial order (level 1) (see (150)). At no point we encountered delta-type harmonic singularities which cannot coexist with the singular nature of the transformation to the Q frame. On the contrary, starting with the form (120), we would have to find out the $`\theta ^+`$ expansion of $`G^{(1,1)}`$ to all orders (in this case it can go up to level 4) and then solve the H-analyticity constraint order by order. The technical advantages of the use of the alternative form of the H-analyticity constraint and of the Q (or Q&S) frame result in major simplifications in the case of the four-point function.
#### 3.4.3 Superconformal covariance at level 1
Now we come back to the four-point function (140). Eq. (146) represents the solution to the H-analyticity constraint and to the conformal covariance condition at level 0. We have also argued that the possibility to go to the Q&S frame (147) guarantees the existence of a unique completion of this level 0 component to a full superfield. The way to find this completion consists of two steps. The first is to put hats on all the $`x`$’s in the denominators, thus reconstructing the products of full propagators. We already know that such products have the required superconformal properties of the correlator. The second step is to complete the conformal cross-ratios $`s`$ and $`t`$ in the coefficient functions $`a,b,c`$ to full superconformal invariants $`\widehat{s}`$ and $`\widehat{t}`$. Then the full correlator consistent with superconformal symmetry will have the form
$$G^{(2,2,2,2)}(1|2|3|4)=\frac{(12)^2(34)^2}{\widehat{x}_{12}^4\widehat{x}_{34}^4}a(\widehat{s},\widehat{t})+\frac{(14)^2(23)^2}{\widehat{x}_{14}^4\widehat{x}_{23}^4}b(\widehat{s},\widehat{t})+\frac{(12)(23)(34)(41)}{\widehat{x}_{12}^2\widehat{x}_{23}^2\widehat{x}_{34}^2\widehat{x}_{41}^2}c(\widehat{s},\widehat{t}).$$
(156)
To find $`\widehat{s}`$ and $`\widehat{t}`$ to all orders in the four $`\theta ^+`$’s is a very non-trivial task (the expansion goes up to level 8, although Q supersymmetry helps bring it down to level 4). Fortunately, the example above has taught us that we only need one level 1 term. So, we are looking for $`\widehat{s}`$ in the form
$$\widehat{s}=s+\underset{a,b=1}{\overset{4}{}}\theta _a^{+\alpha }S_{ab\alpha \dot{\alpha }}\overline{\theta }_b^{+\dot{\alpha }}+O((\theta ^+\overline{\theta }^+)^2).$$
(157)
What we really need is just the coefficient $`S_{44\alpha \dot{\alpha }}`$. Indeed, although the three derivatives $`(D_4^{})^3`$ in (148) can convert a maximum of three $`\theta _4^+`$ into $`\theta _4^{}`$, only the term $`\theta _4^+\overline{\theta }_4^+\theta _4^{}\overline{\theta }_4^{}`$ can survive in the Q&S frame.
The coefficient $`S_{44\alpha \dot{\alpha }}`$ can be solved for from a set of linear equations. It is obtained by performing a combined Q (88) and S (89) supersymmetry transformation on $`\widehat{s}`$ and demanding that it be invariant. In doing so we shall only keep the terms linear in $`\overline{\theta }_4^+`$ since only they involve the coefficients $`S_{a4}`$:
$$\delta _{Q+S}\widehat{s}=0\left[4is(ϵ_4^\alpha +x_4^{\alpha \dot{\beta }}\overline{\eta }_{4\dot{\beta }}^{})\left(\frac{x_{14}}{x_{14}^2}\frac{x_{34}}{x_{34}^2}\right)_{\alpha \dot{\alpha }}+\underset{a=1}{\overset{4}{}}(ϵ_a^{+\alpha }+x_a^{\alpha \dot{\beta }}\overline{\eta }_{a\dot{\beta }}^+)S_{a4\alpha \dot{\alpha }}\right]\overline{\theta }_4^{+\dot{\alpha }}=0.$$
(158)
Here $`(ϵ,\eta )_a^\pm u_{ai}^\pm (ϵ,\eta )^i`$. Removing $`\overline{\theta }_4^{+\dot{\alpha }}`$ and the independent parameters $`ϵ,\eta `$ from (158), we obtain four linear equations (one for each harmonic projection of the two parameters) for the four coefficients $`S_{a4}`$:
$`4isX_2{\displaystyle \underset{a=1}{\overset{3}{}}}(a4)S_{a4}=0,{\displaystyle \underset{a=1}{\overset{4}{}}}(4^{}a)S_{a4}=0,`$
$`(12)x_{24}S_{24}+(13)x_{34}S_{34}=0,(21)x_{14}S_{14}+(23)x_{34}S_{34}=0.`$ (159)
Here and in what follows we use the vectors
$$X_1=\frac{x_{14}}{x_{14}^2}\frac{x_{24}}{x_{24}^2},X_2=\frac{x_{14}}{x_{14}^2}\frac{x_{34}}{x_{34}^2},X_3=\frac{x_{34}}{x_{34}^2}\frac{x_{24}}{x_{24}^2}$$
(160)
having the useful properties
$$\frac{X_1^2}{X_3^2}=\frac{1}{s},\frac{X_2^2}{X_3^2}=\frac{t}{s},\frac{2X_1X_3}{X_3^2}=\frac{1+st}{s},\frac{2X_2X_3}{X_3^2}=\frac{1st}{s}.$$
(161)
Solving eqs. (159) is a straightforward calculation. The result for the coefficient $`S_{44}`$ is
$`{\displaystyle \frac{(12)^2(34)^2Y}{4i}}S_{44}`$ $`=`$ $`(12)(13)(23)tX_3`$ (162)
$``$ $`[(12)(13)(23)s+(12)^2(34)(34^{})s`$
$`+`$ $`(12)(23)(34)(14^{})(1+st)+(14)(14^{})(23)^2]X_2`$
where
$$Y=1+\frac{1+st}{s}U+\frac{1}{s}U^2,U=\frac{(14)(23)}{(12)(34)}.$$
(163)
Note that the vector (162) can be rewritten in the equivalent form
$$\frac{(12)^2(34)^2Y}{4i}S_{44}=\frac{s}{2}_4^{}[(12)^2(34)^2Y]X_2+(12)(13)(23)[tX_3+\frac{1}{2}(1st)X_2].$$
(164)
In in a similar manner we can find the relevant term in the expansion of $`\widehat{t}`$
$$\widehat{t}=t+\underset{a,b=1}{\overset{4}{}}\theta _a^{+\alpha }T_{ab\alpha \dot{\alpha }}\overline{\theta }_b^{+\dot{\alpha }}+\mathrm{}$$
(165)
The result for $`T_{44}`$ can be easily obtained from (162) by exchanging $`12`$ which implies $`st`$, $`X_1X_1`$, $`X_2X_3`$, $`Ys/tY`$.
The expansion of $`\widehat{s}`$, $`\widehat{t}`$ is not the only source of $`\theta _4^+\overline{\theta }_4^+`$ terms. Another contribution comes from expanding $`\widehat{x}_{a4}^2`$, $`a=1,2,3`$ in the propagators in (156):
$$\widehat{x}_{a4}^2=x_{a4}^24i\frac{(a4^{})}{(a4)}\theta _4^+x_{a4}\overline{\theta }_4^++\mathrm{}$$
(166)
(recall (128)). Collecting all of these contributions, we can write down
$`G^{(2,2,2,2)}(1|2|3|4)`$ $``$ $`\mathrm{\Pi }(\widehat{x}_{ab}^2)f(\widehat{s},\widehat{t})`$
$`=`$ $`G^{(2,2,2,2)}|_{\text{level 0}}`$
$`+`$ $`\theta _4^+\left[4i{\displaystyle \underset{a=1}{\overset{3}{}}}{\displaystyle \frac{(a4^{})}{(a4)}}x_{a4}{\displaystyle \frac{\mathrm{\Pi }}{x_{a4}^2}}f+\mathrm{\Pi }(f_sS_{44}+f_tT_{44})\right]\overline{\theta }_4^++\mathrm{}`$
Here $`\mathrm{\Pi }f`$ is a shorthand for the product of propagators ($`\mathrm{\Pi }`$) and coefficient functions ($`f`$) in (156) and $`f_{s,t}=f/(s,t)`$. This accomplishes Step 1 of our programme for imposing H-analyticity in the form (148).
#### 3.4.4 Constraints from H-analyticity
Step 2 consists of differentiating the expression (3.4.3) with the operator $`(D_4^{})^3`$
$`(D_1^{})^3`$ $`=`$ $`(_1^{})^3`$ (168)
$`+`$ $`3_1^{}(\theta _1^{}_1^{}+\overline{\theta }_1^{}\overline{}_1^{})^26i(_1^{})^2\theta _1^{}/_1\overline{\theta }_1^{}`$
$``$ $`6i\theta _1^{}/_1\overline{\theta }_1^{}(\theta _1^{}_1^{}+\overline{\theta }_1^{}\overline{}_1^{})^26_1^{}(\theta _1^{})^2(\overline{\theta }_1^{})^2\mathrm{}_1`$
$`+`$ $`\text{irrelevant terms}.`$
The irrelevant terms in (168) are those which convert an odd number of $`\theta _1^+`$ into $`\theta _1^{}`$ and thus disappear in the Q frame. We then apply (168) to (3.4.3) and collect all the terms at levels 0, 1 and 2. In fact, we have already solved the H-analyticity constraint at level 0 in (146). So, it remains to examine the constraints at levels 1 and 2. It is not hard to see that they take the form
$`\text{Level 1:}_4^{}A^\mu =0,`$ (169)
$`\text{Level 2:}_{4\mu }A^\mu =0`$ (170)
where
$$A^\mu =i_4^{}_4^\mu \left(\mathrm{\Pi }f\right)4i\underset{a=1}{\overset{3}{}}\frac{(a4^{})}{(a4)}x_{a4}^\mu \frac{\mathrm{\Pi }}{x_{a4}^2}f+\mathrm{\Pi }(f_sS_{44}^\mu +f_tT_{44}^\mu ).$$
(171)
After some simple algebra and using the relations
$$_4^\mu s=2sX_2^\mu ,_4^\mu t=2tX_3^\mu ,$$
(172)
we obtain
$$A^\mu =\frac{2ic}{s}\frac{(12)(13)(23)}{x_{12}^4x_{34}^4}X_2^\mu +[2isX_2^\mu _4^{}\mathrm{\Pi }+S_{44}^\mu \mathrm{\Pi }]f_s+[2itX_3^\mu _4^{}\mathrm{\Pi }+T_{44}^\mu \mathrm{\Pi }]f_t.$$
(173)
The terms in the brackets are computed with the help of (164), e.g.
$`[2isX_2_4^{}\mathrm{\Pi }+S_{44}\mathrm{\Pi }]f_s`$ $`=`$ $`2isX_2(12)^2(34)^2Y_4^{}\left({\displaystyle \frac{\mathrm{\Pi }f_s}{(12)^2(34)^2Y}}\right)`$
$`+`$ $`4i(12)(13)(23)[tX_3+{\displaystyle \frac{1}{2}}(1st)X_2]{\displaystyle \frac{\mathrm{\Pi }f_s}{(12)^2(34)^2Y}}.`$
Further, it is convenient to use the harmonic cross-ratio $`U`$ introduced in (163) and rewrite
$$\frac{\mathrm{\Pi }f_s}{(12)^2(34)^2Y}=\frac{1}{x_{12}^4x_{34}^4}\frac{a_s\frac{c_s}{s}U+\frac{b_s}{s^2}U^2}{1+\frac{1+st}{s}U+\frac{1}{s}U^2},$$
(175)
after which the harmonic derivative $`_4^{}`$ in (3.4.4) can be computed using the identity
$$_4^{}U=\frac{(12)(13)(23)}{(12)^2(34)^2}.$$
Repeating the same procedure for the other bracket in (173) and collecting all the terms proportional to the vector $`X_2`$, we obtain the following contribution to the vector $`A^\mu `$:
$$2i\frac{(12)(13)(23)}{x_{12}^4x_{34}^4}\left[\frac{c}{s}+\beta _0\frac{1+\frac{\beta _1}{\beta _0}U+\frac{\beta _2}{\beta _0}U^2}{1+\frac{1+st}{s}U+\frac{1}{s}U^2}\right]X_2^\mu $$
(176)
where
$`\beta _0`$ $`=`$ $`c_s+2(1t)a_s2ta_t,`$
$`\beta _1`$ $`=`$ $`2a_s{\displaystyle \frac{2}{s}}b_s+{\displaystyle \frac{2t}{s}}c_t+{\displaystyle \frac{s+t1}{s}}c_s,`$ (177)
$`\beta _2`$ $`=`$ $`{\displaystyle \frac{2}{s}}b_s{\displaystyle \frac{2t}{s^2}}b_t{\displaystyle \frac{1}{s}}c_s.`$
We still have to compute the $`X_3`$ contribution to $`A^\mu `$, but even before this we can already impose the level 1 constraint (169) on the $`X_2`$ contribution (the vectors $`X_2`$ and $`X_3`$ are linearly independent and $`_4^{}X_2=_4^{}X_3=0`$). The first term in (176) does not depend on $`u_4`$. The second term is the ratio of two polynomials of degree 2 in the cross-ratio $`U`$. It is easy to see that its derivative vanishes only if the two polynomials are equal,
$$1+\frac{\beta _1}{\beta _0}U+\frac{\beta _2}{\beta _0}U^2=1+\frac{1+st}{s}U+\frac{1}{s}U^2.$$
(178)
Comparing the coefficients in front of $`U`$ and $`U^2`$, we obtain the following constraints:
$`c_s=(t1)a_s+ta_tb_s{\displaystyle \frac{t}{s}}b_t,`$
$`c_t=sa_ssa_t+b_s+{\displaystyle \frac{t1}{s}}b_t`$ (179)
constituting our main result. This is a set of two linear first-order partial differential equations for the three coefficient functions $`a,b,c`$. These equations can only determine two of the three functions. Indeed, let us make the change of variables
$$a=\alpha +\gamma +\frac{1}{s}b,c=s\gamma +\frac{ts1}{s}b,$$
(180)
after which (179) becomes
$`\gamma _s=\alpha _s\alpha _t,`$
$`t\gamma _t+\gamma =s\alpha _s+(s1)\alpha _t`$ (181)
and we see that $`b`$ has dropped out. The set of first-order coupled differential equations (181) can be equivalently rewritten as a set of second-order independent ones:
$`s\alpha _{ss}+t\alpha _{tt}+(s+t1)\alpha _{st}+2(\alpha _s+\alpha _t)`$ $`=`$ $`0,`$
$`s\gamma _{ss}+t\gamma _{tt}+(s+t1)\gamma _{st}+2(\gamma _s+\gamma _t)`$ $`=`$ $`0.`$ (182)
These constraints are the same as eqs. (79) obtained in section 2 in the coordinate approach.
To compute the contribution to $`A^\mu `$ proportional to $`X_3`$ we go through the same steps. This time we do not find any new constraints. The final form of the vector $`A^\mu `$ is
$$A^\mu =2i(12)(13)(23)\left(A\frac{X_2^\mu }{x_{12}^4x_{34}^4s}+B\frac{X_3^\mu }{x_{12}^4x_{34}^4t}\right)$$
(183)
where
$$A=sc_s+2(1t)sa_s2sta_t+c,B=\frac{t^2}{s}(c_t+2sa_s+2sa_t).$$
(184)
The remaining step is to impose the level 2 constraint (170). This is facilitated by the useful property
$$_{4\mu }\left(\frac{X_2^\mu }{x_{12}^4x_{34}^4s}\right)=_{4\mu }\left(\frac{X_3^\mu }{x_{12}^4x_{34}^4t}\right)=0$$
of the basis vectors in (183), so we only have to differentiate the scalar coefficients $`A`$ and $`B`$. Using the identities (172) and (161), we obtain the constraint
$$_{4\mu }A^\mu =0\frac{t}{s}A_sB_t+\frac{1st}{2s}\left(\frac{t}{s}A_t+\frac{s}{t}B_s\right)=0.$$
(185)
Making the change of variables (180) and after some algebra we discover that this is a corollary of the second-order differential equations (182). So, level 2 does not give rise to any new constraints.
## 4 Conclusions
We see that using either method of analysing analyticity for the four-point charge 2 correlator leads to the same result: The requirements of H-analyticity and superconformal covariance yield constraints which fix the form of the four-point correlator (140) up to an arbitrary function of the conformal cross-ratios. An interesting solution to these constraints is provided by the explicit computation of the correlator at two loops carried out in . The result for the level 0 component is
$$\mathrm{\Phi }(s,t)\left[\frac{(12)^2(34)^2}{x_{12}^4x_{34}^4}+\frac{(14)^2(23)^2}{x_{14}^4x_{23}^4}s+\frac{(12)(23)(34)(41)}{x_{12}^2x_{23}^2x_{34}^2x_{41}^2}(ts1)\right].$$
(186)
Here $`\mathrm{\Phi }(s,t)=\mathrm{\Phi }(t,s)=\frac{1}{s}\mathrm{\Phi }(\frac{1}{s},\frac{t}{s})`$ is a function given by the one-loop scalar box integral. This solution is symmetric under the exchange $`1\mathrm{\hspace{0.33em}3}`$ and is determined by the asymptotic behaviour $`lim_{x_{14}0}c(s,t)=0`$ (see for details). It is interesting to note how the result (186) was obtained: the two-loop calculation in only provided us with the explicit form of the first two terms in (186); the third term was given as a complicated two-loop integral; the subsequent use of the differential equations (182) in conjunction with the boundary conditions following from the known asymptotic behaviour of the two-loop integral allowed us to solve for the third term as in (186).
We have shown for a number of cases that harmonic analyticity at all orders in an expansion of a correlator in the odd variables is assured by the constraints arising from the lowest and the linear order. We conjecture that this is a general feature.
Acknowledgements: This work was supported in part by the British-French scientific programme Alliance (project 98074), by the EU network on Integrability, non-perturbative effects and symmetry in quantum field theory (FMRX-CT96-0012) and by the grant INTAS-96-0308. |
warning/0001/math0001083.html | ar5iv | text | # Magic squares of Lie Algebras
## 1. Introduction
Semisimple Lie groups and Lie algebras are normally discussed in terms of their root systems, which makes possible a unified treatment and emphasises the common features of their underlying structures. However, some classical investigations depend on particularly simple matrix descriptions of Lie groups. This creates a distinction between the classical groups (naturally enough) and the exceptional ones, which is maintained in some more recent work (e.g. ). This paper is motivated by the desire to give a similar matrix description of the exceptional groups, thus assimilating them to the classical groups, with a view to extending results like the Capelli identities to the exceptional cases.
It has long been known that most exceptional Lie algebras are related to the exceptional Jordan algebra of $`3\times 3`$ hermitian matrices with entries from the octonions, $`𝕆`$. Here we show that this relation yields descriptions of certain real forms of the complex Lie algebras $`F_4,E_6`$ and $`E_7`$ which can be interpreted as octonionic versions of, respectively, the Lie algebra of antihermitian $`3\times 3`$ matrices, that of special linear $`3\times 3`$ matrices and that of symplectic $`6\times 6`$ matrices. To be precise, we define for each alternative algebra $`𝕂`$ a Lie algebra $`𝔰𝔞(3,𝕂)`$ such that $`𝔰𝔞(3,)=𝔰𝔲(3)`$ and $`𝔰𝔞(3,𝕆)`$ is the compact real form of $`F_4`$; a Lie algebra $`𝔰𝔩(3,𝕂)`$ equal to $`𝔰𝔩(3,)`$ for $`𝕂=`$ and a non-compact real form of $`E_6`$ for $`𝕂=𝕆`$; and a Lie algebra $`𝔰𝔭(6,𝕂)`$ such that $`𝔰𝔭(6,)`$ is the set of $`6\times 6`$ complex matrices $`X`$ satisfying $`X^{}J=JX`$, (where J is an antisymmetric real $`6\times 6`$ matrix and $`X^{}`$ denotes the hermitian conjugate of $`X`$), and such that $`𝔰𝔭(6,𝕆)`$ is a non-compact real form of $`E_7`$.
Our definitions can be adapted to yield Lie algebras $`𝔰𝔞(2,𝕂),𝔰𝔩(2,𝕂)`$ and $`𝔰𝔭(4,𝕂)`$ reducing to $`𝔰𝔲(2),𝔰𝔩(2,)`$ and $`𝔰𝔭(4,)`$ when $`𝕂=`$. These Lie algebras are isomorphic to various pseudo-orthogonal algebras.
These constructions are all related to Tits’s magic square of Lie algebras and based on an unpublished suggestion of Ramond . The magic square of Tits is a construction of a Lie algebra $`L(𝕁,𝕂)`$ for any Jordan algebra $`𝕁`$ and alternative algebra $`𝕂`$. If $`𝕁=H_3(𝕂_1)`$ is the Jordan algebra of $`3\times 3`$ matrices with entries from an alternative algebra $`𝕂_1`$ and if $`𝕂=𝕂_2`$ is another alternative algebra, this yields a Lie algebra $`L_3(𝕂_1,𝕂_2)`$ for any pair of alternative algebras. Taking $`𝕂_1`$ and $`𝕂_2`$ to be real division algebras, we obtain a $`4\times 4`$ square of compact Lie algebras which (magically) is symmetric and contains the compact real forms of $`F_4,E_6,E_7`$ and $`E_8`$. We will show that if the division algebra $`𝕂_2`$ is replaced by its split form $`\stackrel{~}{𝕂}_2`$, one obtains a non-symmetric square of Lie algebras whose first three rows are the sets of matrix Lie algebras described above:
$`L_3(𝕂,)`$ $`=𝔰𝔞(3,𝕂)`$
(1) $`L_3(𝕂,\stackrel{~}{})`$ $`=𝔰𝔩(3,𝕂)`$
$`L_3(𝕂,\stackrel{~}{})`$ $`=𝔰𝔭(6,𝕂).`$
We will also describe magic squares of Lie algebras based on $`2\times 2`$ matrices, which have similar properties.
The organisation of the paper is as follows. In Section 2 we establish notation and recall the definitions of the various kinds of algebra with which we will be concerned. In Section 3 we give Tits’s definition of the Lie algebras $`L_3(𝕂_1,𝕂_2)`$ and state the main properties of the magic square; we also give the definition and properties of the $`2\times 2`$ magic square $`L_2(𝕂_1,𝕂_2)`$. Section 4 is concerned with the symmetry property of the $`3\times 3`$ magic square: we reformulate the definition of $`L_3(𝕂_1,𝕂_2)`$, using Ramond’s concept of a *triality algebra*, so as to make the symmetry manifest. Section 5 contains proofs of the properties of the $`2\times 2`$ magic square which were stated in Section 3, and Section 6 contains the proofs of the corresponding properties of the $`3\times 3`$ magic square.
## 2. Notation.
We will use the notation $`\dot{+}`$ to denote the direct sum of vector spaces. This enables us to reserve the use of $``$ to denote the direct sum of Lie algebras, i.e. $`AB`$ implies that $`[A,B]=0`$.
An algebra $`𝕂`$ (over $``$) with a non-degenerate quadratic form, which we will denote by $`x\left|x\right|^2`$, satisfying
(2)
$$\left|xy\right|^2=\left|x\right|^2\left|y\right|^2x,y𝕂,$$
is known as a composition algebra. We consider $``$ to be embedded in $`𝕂`$ as the set of scalar multiples of the identity element, and denote by $`𝕂^{}`$ the subspace of $`𝕂`$ orthogonal to $``$. It can then be shown that $`𝕂=\dot{+}𝕂^{}`$ and we write $`x=\mathrm{Re}x+\mathrm{Im}x`$ with $`\mathrm{Re}x`$ and $`\mathrm{Im}x𝕂^{}`$. It can also be shown that the conjugation which fixes each element of $``$ and multiplies every element of $`𝕂^{}`$ by $`1`$, denoted $`x\overline{x}`$, satisfies
(3)
$$\overline{xy}=\overline{y}\overline{x}$$
as well as
(4)
$$x\overline{x}=\left|x\right|^2.$$
We use the notation $`[x,y,z]`$ for the associator
(5)
$$[x,y,z]=(xy)zx(yz).$$
Any composition algebra $`𝕂`$ satisfies the alternative law, i.e. the associator is an alternating function of $`x,y`$ and $`z`$. If $`\left|x\right|^2`$ is positive definite then $`𝕂`$ is a division algebra.
A division algebra is an algebra in which we have
$$xy=0x=0\text{ or }y=0.$$
The only such positive definite composition algebras are $`,,`$ and $`𝕆`$ (Hurwitz’s Theorem) which we denote in general by $`𝕂`$. We denote their dimension by $`\nu `$, thus $`\nu =1,2,4`$ or $`8`$. These algebras are obtained from the Cayley-Dickson process and, using the same process with different signs, split forms of these algebras can also be obtained. These are so called because in $`,`$ and $`𝕆`$ we have
$$i^2+1=0$$
but in the split algebras $`\stackrel{~}{},\stackrel{~}{}`$ and $`\stackrel{~}{𝕆}`$ we have
$$i^21=(i+1)(i1)=0$$
i.e. the equation can be split for at least one of the imaginary basis elements. Thus whilst the positive definite algebras $`,,`$ and $`𝕆`$ are division algebras, the split forms $`\stackrel{~}{},\stackrel{~}{}`$ and $`\stackrel{~}{𝕆}`$ are not.
Our notation for Lie algebras is that used in . We use the notation $`A^{}`$ for the hermitian conjugate of the matrix $`A`$ with entries in $`𝕂`$, defined in analogy to the complex case by
$$(X^{})_{ij}=\overline{X}_{ji}.$$
We use $`𝔰𝔲(s,t)`$ for the Lie algebra of the pseudo-unitary group,
$$𝔰𝔲(s,t)=\{A^{n\times n}:A^{}G+GA=0\}$$
where $`G=\mathrm{diag}(1,\mathrm{},1,+1,\mathrm{},+1)`$ with $`s`$ $``$ signs and $`t`$ $`+`$ signs; $`𝔰𝔮(n)`$ for the Lie algebra of antihermitian quaternionic matrices $`A`$,
$$𝔰𝔮(n)=\{A^{n\times n}:A^{}=A\};$$
and $`𝔰𝔭(2n,𝕂)`$ for the Lie algebra of the symplectic group of $`2n\times 2n`$ matrices with entries in $`𝕂`$, i.e.
$$𝔰𝔭(2n,𝕂)=\{A𝕂^{2n\times 2n}:A^{}J+JA=0\}$$
where $`J=\left(\begin{array}{cc}0& I_n\\ I_n& 0\end{array}\right)`$. We also have $`𝔰𝔬(s,t)`$, the Lie algebra of the pseudo-orthogonal group $`\mathrm{SO}(s,t)`$, given by
$$𝔰𝔬(s,t)=\{A^{n\times n}:A^TG+GA=0\}$$
where $`G`$ is defined as before. We will also write $`\mathrm{O}(V,q)`$ for the group of linear maps of the vector space $`V`$ preserving the non-degenerate quadratic form $`q`$, $`\mathrm{SO}(V,q)`$ for its unimodular (or special) subgroup, $`𝔬(V,q)`$ and $`𝔰𝔬(V,q)`$ for their Lie algebras. We omit $`q`$ if it is understood from the context. Thus for any division algebra we have $`\mathrm{SO}(𝕂)`$ and $`𝔰𝔬(𝕂)`$.
A Jordan algebra $`𝕁`$ is defined to be a commutative algebra (over a field $`𝕂`$) in which all products satisfy the Jordan identity
$$(xy)x^2=x(yx^2).$$
Let $`L_n(𝕂)`$ be the set of all $`n\times n`$ matrices with entries in $`𝕂`$, and let $`H_n(𝕂)`$ and $`A_n(𝕂)`$ be the sets of all hermitian and antihermitian matrices with entries in $`𝕂`$ respectively. We denote by $`H_n^{}(𝕂)`$, $`A_n^{}(𝕂)`$ and $`L_n^{}(𝕂)`$ the subspaces of traceless matrices of $`H_n(𝕂)`$, $`A_n(𝕂)`$ and $`L_n(𝕂)`$ respectively. We thus have $`L_n(𝕂)=H_n(𝕂)\dot{+}A_n(𝕂)`$ and $`L_n^{}(𝕂)=H_n^{}(𝕂)\dot{+}A_n^{}(𝕂)`$. We will use the fact that $`H_n(𝕂)`$ is a Jordan algebra for $`𝕂=,,`$ for all $`n`$ and for $`𝕂=𝕆`$ when $`n=2,3`$ , with the Jordan product as the anticommutator
$$XY=XY+YX.$$
This is a commutative but non-associative product.
The derivation algebra, $`\mathrm{Der}A`$, of any algebra $`A`$ is defined as
(6)
$$\mathrm{Der}A=\{DD(xy)=D(x)y+xD(y)\}$$
for $`x,yA`$. The derivation algebras of the four positive definite composition algebras are as follows:
$`\mathrm{Der}`$ $`=\mathrm{Der}=0,`$
$`\mathrm{Der}`$ $`=C(^{})=\{C_aa^{}\}\text{ where }C_a(q)=aqqa.`$
$`\mathrm{Der}𝕆`$ is an exceptional Lie algebra of type $`G_2`$.
The structure algebra $`\mathrm{Str}A`$ of any algebra $`A`$ is defined to be the Lie algebra generated by left and right multiplication maps $`L_a`$ and $`R_a`$ for $`aA`$. For Jordan algebras this can be shown to be
(7)
$$\mathrm{Str}𝕁=\mathrm{Der}𝕁\dot{+}L(𝕁)$$
where $`L(𝕁)`$ is the set of all $`L_a`$ with $`a𝕁`$. The algebra denoted by $`\mathrm{Str}^{}𝕁`$ is the structure algebra with its centre factored out. We also require another Lie algebra associated with a Jordan algebra, namely the conformal algebra as constructed by Kantor (1973) and Koecher (1967). The underlying vector space of this is
(8)
$$\mathrm{Con}𝕁=\mathrm{Str}𝕁\dot{+}𝕁^2.$$
We will specify the Lie brackets of these algebras at a later stage.
## 3. Magic Squares: Summary of Results
### 3.1. $`3\times 3`$ Matrices
Let $`𝕂`$ be a real composition algebra and $`𝕁`$ a real Jordan algebra, with $`𝕂^{}`$ and $`𝕁^{}`$ the quotients of the algebras by the subspaces of scalar multiples of the identity. Define a vector space
(9)
$$M(𝕁,𝕂)=\mathrm{Der}𝕁\dot{+}(𝕁^{}𝕂^{})\dot{+}\mathrm{Der}𝕂.$$
Then define
$$L_3(𝕂_1,𝕂_2)=M(H_3(𝕂_1),𝕂_2).$$
Explicitly this is the vector space
(10)
$$L_3(𝕂_1,𝕂_2)=\mathrm{Der}H_3(𝕂_1)\dot{+}H_3^{}(𝕂_1)𝕂_2^{}\dot{+}\mathrm{Der}𝕂_2$$
which is a Lie algebra with Lie subalgebras $`\mathrm{Der}H_3(𝕂_1)`$ and $`\mathrm{Der}𝕂_2`$ when taken with the brackets
$`[D,Ax]`$ $`=D(A)x`$
$`[E,Ax]`$ $`=AE(x)`$
(11) $`[D,E]`$ $`=0`$
$`[Ax,By]`$ $`=\frac{1}{6}A,BD_{x,y}+(AB)\frac{1}{2}[x,y]x,y[L_A,L_B]`$
with $`D\mathrm{Der}H_3(𝕂_1);A,BH_3^{}(𝕂_1);x,y𝕂_2^{}`$ and $`E\mathrm{Der}𝕂_2`$. These brackets are obtained from Schafer’s description of the Tits construction . They require some explanation. $`A,B`$ and $`(x,y)`$ denote the symmetric bilinear forms on $`H_3(𝕂_1)`$ and $`𝕂_2`$ respectively, given by
$`A,B`$ $`=\mathrm{Re}(\mathrm{tr}(AB))=2\mathrm{Re}(\mathrm{tr}(AB))`$
$`x,y`$ $`=\frac{1}{2}(\left|x+y\right|^2\left|x\right|^2\left|y\right|^2)=\mathrm{Re}(x\overline{y}).`$
The derivation $`D_{x,y}`$ is defined as
(12)
$$D_{x,y}=[L_x,L_y]+[L_x,R_y]+[R_x,R_y]\mathrm{Der}𝕂_2.$$
For future reference we note that
(13)
$$D_{x,y}z=[[x,y],z]3[x,y,z]$$
which shows that $`D_{x,y}=D_{y,x}`$. Finally $`(AB)`$ is the traceless part of the Jordan product of $`A`$ and $`B`$,
(14)
$$AB=AB\frac{1}{3}\mathrm{tr}(AB).$$
Tits (see also ) showed that this gives a unified construction leading to the so-called magic square of Lie algebras of $`3\times 3`$ matrices whose complexifications are
| | $``$ | $``$ | $``$ | $`𝕆`$ |
| --- | --- | --- | --- | --- |
| $``$ | $`A_1`$ | $`A_2`$ | $`C_3`$ | $`F_4`$ |
| $``$ | $`A_2`$ | $`A_2A_2`$ | $`A_5`$ | $`E_6`$ |
| $``$ | $`C_3`$ | $`A_5`$ | $`B_6`$ | $`E_7`$ |
| $`𝕆`$ | $`F_4`$ | $`E_6`$ | $`E_7`$ | $`E_8`$ |
The striking properties of this square are (a) its symmetry and (b) the fact that four of the five exceptional Lie algebras occur in its last row. The explanation of the symmetry property is the subject of section 4. The fifth exceptional Lie algebra, $`G_2`$, can be included by adding an extra row corresponding to the Jordan algebra $``$.
In it is asserted without proof that we can write this in a slightly different form and that we can include the isomorphisms listed in (1). This involves a different set of real forms obtained by taking the split composition algebras $`,\stackrel{~}{},\stackrel{~}{},`$ and $`\stackrel{~}{𝕆}`$ rather than $`,,,`$ and $`𝕆`$ as the second algebra. Thus the split magic square for three by three matrices looks like
| | $``$ | $``$ | $``$ | $`𝕆`$ |
| --- | --- | --- | --- | --- |
| $`\mathrm{Der}H_3(𝕂)L_3(𝕂,)`$ | $`𝔰𝔬(3)`$ | $`𝔰𝔲(3)`$ | $`𝔰𝔮(3)`$ | $`F_{4,1}`$ |
| $`\mathrm{Str}^{}H_3(𝕂)L_3(𝕂,\stackrel{~}{})`$ | $`𝔰𝔩(3,)`$ | $`𝔰𝔩(3,)`$ | $`𝔰𝔩(3,)`$ | $`E_{6,1}`$ |
| $`\mathrm{Con}H_3(𝕂)L_3(𝕂,\stackrel{~}{})`$ | $`𝔰𝔭(6,)`$ | $`𝔰𝔲(3,3)`$ | $`𝔰𝔭(6,)`$ | $`E_{7,1}`$ |
| $`L_3(𝕂,\stackrel{~}{𝕆})`$ | $`F_{4,2}`$ | $`E_{6,2}`$ | $`E_{7,2}`$ | $`E_{8,1}`$ |
where the notation <sub>,1</sub> and <sub>,2</sub> (in the style of ) is used to distinguish between different real forms of the exceptional Lie algebras in the last row and column. These are identified by their maximal compact subalgebras as follows:
| Exceptional Lie Algebra | Maximal Compact Subalgebra |
| --- | --- |
| $`E_{6,1}`$ | $`F_4`$ |
| $`E_{7,1}`$ | $`E_{6,1}𝔰𝔬(2)`$ |
| $`E_{8,1}`$ | $`E_{7,1}𝔰𝔬(3)`$ |
| $`E_{6,2}`$ | $`𝔰𝔮(3)𝔰𝔬(3)`$ |
| $`E_{7,2}`$ | $`𝔰𝔲(6)𝔰𝔬(3)`$ |
| $`E_{8,2}`$ | $`𝔰𝔬(12)𝔰𝔬(3)`$ |
.
### 3.2. $`2\times 2`$ Matrices
The Tits construction can also be adapted for $`2\times 2`$ matrix algebras. In this case we take the vector space to be
(15)
$$L_2(𝕂_1,𝕂_2)=\mathrm{Der}H_2(𝕂_1)\dot{+}H_2^{}(𝕂_1)𝕂_2^{}\dot{+}𝔰𝔬(𝕂_2^{})$$
which is again a Lie algebra when taken with the brackets
(16) $`[D,Ax]`$ $`=D(A)x`$
$`[E,Ax]`$ $`=AE(x)`$
$`[D,E]`$ $`=0`$
$`[Ax,By]`$ $`=\frac{1}{4}A,BD_{x,y}x,y[R_A,R_B]`$
where the symbols used in this set of brackets are defined in the same way as the ones used in the $`3\times 3`$ case. We note that $`D_{x,y}=2s_{x,y}`$, where $`s_{x,y}`$ is the element of $`𝔰𝔬(𝕂_2^{})`$ that maps $`x`$ to $`y`$ and $`y`$ to $`\pm x`$, depending on the metric of $`𝕂_2^{}`$ i.e.
(17)
$$S_{x,y}(z)=x,zyy,zx$$
If $`𝕂_1,𝕂_2`$ are division algebras then this gives the compact magic square for $`2\times 2`$ matrix algebras
$$L_2(𝕂_1,𝕂_2)=𝔰𝔬(\nu _1+\nu _2).$$
If $`𝕂_2`$ is one of the split composition algebras $`\stackrel{~}{},\stackrel{~}{}`$ or $`\stackrel{~}{𝕆}`$ this becomes
$$L_2(𝕂_1,𝕂_2)=𝔰𝔬(\nu _1+\frac{1}{2}\nu _2,\frac{1}{2}\nu _2).$$
giving the magic square
| | $``$ | $``$ | $``$ | $`𝕆`$ |
| --- | --- | --- | --- | --- |
| $`L_2(𝕂,)`$ | $`𝔰𝔬(2)`$ | $`𝔰𝔬(3)`$ | $`𝔰𝔬(5)`$ | $`𝔰𝔬(9)`$ |
| $`L_2(𝕂,\stackrel{~}{})`$ | $`𝔰𝔬(2,1)`$ | $`𝔰𝔬(3,1)`$ | $`𝔰𝔬(5,1)`$ | $`𝔰𝔬(9,1)`$ |
| $`L_2(𝕂,\stackrel{~}{})`$ | $`𝔰𝔬(3,2)`$ | $`𝔰𝔬(4,2)`$ | $`𝔰𝔬(6,2)`$ | $`𝔰𝔬(10,2)`$ |
| $`L_2(𝕂,\stackrel{~}{𝕆})`$ | $`𝔰𝔬(5,4)`$ | $`𝔰𝔬(6,4)`$ | $`𝔰𝔬(8,4)`$ | $`𝔰𝔬(12,4)`$ |
.
As in the $`3\times 3`$ case, these Lie algebras can be identified with certain types of $`2\times 2`$ matrix algebras
| | $``$ | $``$ | $``$ | $`𝕆`$ |
| --- | --- | --- | --- | --- |
| $`\mathrm{Der}H_2(𝕂)L_2(𝕂,)`$ | $`𝔰𝔬(2)`$ | $`𝔰𝔲(2)`$ | $`𝔰𝔮(2)`$ | $`𝔰𝔬(9)`$ |
| $`\mathrm{Str}H_2(𝕂)L_2(𝕂,\stackrel{~}{})`$ | $`𝔰𝔩(2,)`$ | $`𝔰𝔩(2,)`$ | $`𝔰𝔩(2,)`$ | $`𝔰𝔩(2,𝕆)`$ |
| $`\mathrm{Con}H_2(𝕂)L_2(𝕂,\stackrel{~}{})`$ | $`𝔰𝔭(4,)`$ | $`𝔰𝔲(2,2)`$ | $`𝔰𝔭(4,)`$ | $`𝔰𝔭(4,𝕆)`$ |
| $`L_2(𝕂,\stackrel{~}{𝕆})`$ | $`𝔰𝔬(5,4)`$ | $`𝔰𝔬(6,4)`$ | $`𝔰𝔬(8,4)`$ | $`𝔰𝔬(12,4)`$ |
.
Again this extends the concepts of the Lie algebras $`𝔰𝔞(2,𝕂),𝔰𝔩(2,𝕂)`$ and $`𝔰𝔭(2,𝕂)`$ to $`𝕂=`$ and $`𝕆`$. Note that $`𝔰𝔲(2,2)𝔰𝔭(4,)`$.
## 4. Symmetry Property of the $`3\times 3`$ magic square.
In this section we will rearrange the definition
(18)
$$L_3(𝕂_1,𝕂_2)=\mathrm{Der}H_3(𝕂_1)\dot{+}H_3^{}(𝕂_1)𝕂_2^{}\dot{+}\mathrm{Der}𝕂_2$$
so as to make explicit the symmetry between $`𝕂_1`$ and $`𝕂_2`$. We need a new Lie algebra associated with any $`𝕂`$, defined as follows:
###### Definition 1.
Let $`𝕂`$ be a composition algebra over $``$. The *triality algebra* of $`𝕂`$ is
(19)
$$\mathrm{Tri}𝕂=\{(A,B,C)𝔰𝔬(𝕂)^3|A(xy)=x(By)+(Cx)y,x,y𝕂\}.$$
It is easy to verify that $`\mathrm{Tri}𝕂`$ is a Lie algebra with brackets defined componentwise, i.e. it is a Lie subalgebra of $`𝔰𝔬(𝕂)𝔰𝔬(𝕂)𝔰𝔬(𝕂)`$.
###### Lemma 1.
The triality algebras of the four positive-definite composition algebras can be identified as follows:
$`\mathrm{Tri}`$ $`=0`$
$`\mathrm{Tri}`$ $`^2`$
$`\mathrm{Tri}`$ $`𝔰𝔬(3)𝔰𝔬(3)𝔰𝔬(3)`$
$`\mathrm{Tri}𝕆`$ $`𝔰𝔬(8)`$
###### Proof.
$`\mathrm{Tri}=0`$ because $`𝔰𝔬()=0`$. For $``$ we can identify $`𝔰𝔬()`$ with the set of multiplication maps $`zhz`$ with $`h`$ pure imaginary, which is isomorphic to $``$ as a Lie algebra. Then $`\mathrm{Tri}`$ is the subspace of the abelian Lie algebra $`^3`$ given by
$$\mathrm{Tri}=\{(u,v,w):u=v+w\}$$
which is two dimensional.
Antisymmetric linear maps $`A:`$ are all of the form $`A=L_{a_1}+R_{a_2}`$ with $`a_1,a_2^{}`$. These are all independent (this is a reflection of the Lie algebra isomorphism $`𝔰𝔬(4)𝔰𝔬(3)𝔰𝔬(3)`$). Hence the condition for $`(A,B,C)\mathrm{Tri}`$ is of the form
$$a_1xy+xya_2=c_1xy+x(c_2+b_1)y+xyb_2,x,y$$
Taking $`y=1`$ and using the independence of the left and right multiplication maps gives
$$a_1=c_1\text{and}a_2=c_2+b_1+b_2.$$
Taking $`x=1`$ gives
$$a_1=c_1+c_2+b_1\text{and}a_2=b_2.$$
Hence $`c_2+b_1=0`$ and we have
$$A=L_{a_1}+R_{a_2},B=L_{b_1}+R_{a_2},C=L_{a_1}R_{b_1}.$$
Thus $`\mathrm{Tri}^{\mathrm{\hspace{0.17em}3}}𝔰𝔬(3)𝔰𝔬(3)𝔰𝔬(3)`$.
Finally, the infinitesimal version of the principle of triality asserts that for each $`A𝔰𝔬(8)`$ there are unique $`B,C𝔰𝔬(8)`$ such that
$$A(xy)=x(By)+(Cx)yx,y,𝕆.$$
This establishes an isomorphism between $`\mathrm{Tri}𝕆`$ and $`𝔰𝔬(8)`$. ∎
We will now describe the chain of inclusions
(20)
$$\mathrm{Der}𝕂\mathrm{Tri}𝕂\mathrm{Der}H_3(𝕂)$$
in a unified way, valid for any composition algebra $`𝕂`$. We will use a multiple notation to describe multiple direct sums, writing
$$nV=\underset{n}{\underset{}{V\dot{+}V\dot{+}\mathrm{}\dot{+}V}}$$
rather than $`V^n`$ (which might suggest $`V\mathrm{}V`$).
###### Lemma 2.
For any composition algebra $`𝕂`$,
$$\mathrm{Tri}𝕂=\mathrm{Der}𝕂\dot{+}2𝕂^{}$$
in which $`\mathrm{Der}𝕂`$ is a Lie subalgebra,
$`[D,(a,b)]`$ $`=(Da,Db)2𝕂^{}`$
$`[(a,0),(b,0)]`$ $`=\frac{2}{3}D_{a,b}+(\frac{1}{3}[a,b],\frac{2}{3}[a,b]),`$
$`[(a,0),(0,b)]`$ $`=\frac{1}{3}D_{a,b}(\frac{1}{3}[a,b],\frac{1}{3}[a,b]),`$
$`[(0,a),(0,b)]`$ $`=\frac{2}{3}D_{a,b}+(\frac{2}{3}[a,b],\frac{1}{3}[a,b]).`$
###### Proof.
Define $`T:\mathrm{Der}𝕂\dot{+}2𝕂^{}\mathrm{Tri}𝕂`$ by
(21)
$$\begin{array}{c}T(D,a,b)=\hfill \\ \hfill (D+L_aR_b,DL_aL_bR_b,D+L_a+R_a+R_b).\end{array}$$
This belongs to $`\mathrm{Tri}𝕂`$ as a consequence of the alternative law. The map $`T`$ is injective, for $`T(D,a,b)=0`$ implies
$$2L_a+L_b=0\text{and}R_a+2R_b=0$$
(subtracting the first component from the second and third in turn), so $`2a+b=a+2b=0`$ and hence $`a+b=0`$, which implies $`D=0`$. To show that $`T`$ is surjective, suppose $`(A,B,C)\mathrm{Tri}𝕂`$ and define $`a,b𝕂^{}`$ by
$`B(1)=a2b,`$
$`C(1)=2a+b.`$
Let
(22)
$$D=AL_a+R_b.$$
Since $`(A,B,C)\mathrm{Tri}𝕂`$ we have
$$B(x)=1.B(x)=A(1.x)C(1)x.$$
Thus
$`B`$ $`=A2L_aL_b`$
(23) $`=DL_aL_bR_b`$
and
(24)
$$C=D+L_a+R_a+R_b$$
Now
$`D(xy)`$ $`=A(xy)a(xy)+(xy)b`$
$`=x(By)+(Cx)ya(xy)+(xy)b`$
(25) $`=(Dx)y+x(Dy)`$
by equations (224) and the alternative law. Hence $`D`$ is a derivation and $`(A,B,C)=T(D,a,b)`$.
The first Lie bracket stated above, i.e.
$$[T(D,0,0),T(0,a,b)]=T(0,Da,Db),$$
follows from
(26)
$$[D,L_a]=L_{Da}\text{and}[D,R_a]=R_{Da}.$$
The other brackets follow from the commutators
$`[L_x,L_y]`$ $`=\frac{2}{3}D_{x,y}+\frac{1}{3}L_{[x,y]}+\frac{2}{3}R_{[x,y]}`$
$`[L_x,R_y]`$ $`=\frac{1}{3}D_{x,y}+\frac{1}{3}L_{[x,y]}\frac{1}{3}R_{[x,y]}`$
$`[R_x,R_y]`$ $`=\frac{2}{3}D_{x,y}\frac{2}{3}L_{[x,y]}\frac{1}{3}R_{[x,y]}`$
which can be calculated using equation (13). ∎
Any two elements $`x,y𝕂`$ are associated with a derivation $`D_{x,y}\mathrm{Der}𝕂`$ and also with an antisymmetric map $`S_{x,y}𝔰𝔬(𝕂)`$, the generator of rotations in the plane of $`x`$ and $`y`$ given by (17). There is also an element of $`\mathrm{Tri}𝕂`$ associated with $`x`$ and $`y`$:
###### Lemma 3.
For any $`x,y𝕂`$, let
$$T_{x,y}=(4S_{x,y},R_yR_{\overline{x}}R_xR_{\overline{y}},L_yL_{\overline{x}}L_xL_{\overline{y}}).$$
Then $`T_{x,y}\mathrm{Tri}𝕂`$.
###### Proof.
We can write the action of $`S_{x,y}`$ as
(27) $`2S_{x,y}z`$ $`=(x\overline{z}+z\overline{x})yx(\overline{z}y+\overline{y}z)`$
(28) $`=[x,y,z]+z(\overline{x}y)(x\overline{y})z`$
using the alternative law and the relation $`[x,y,\overline{z}]=[x,y,z]`$. Since $`\mathrm{Re}(\overline{x}y)=\mathrm{Re}(x\overline{y})`$, we can write the last two terms as
(29) $`z(\overline{x}y)(x\overline{y})z`$ $`=z\mathrm{Im}(\overline{x}y)\mathrm{Im}(x\overline{y})z`$
(30) $`=\frac{1}{2}z(\overline{x}y\overline{y}x)\frac{1}{2}(x\overline{y}y\overline{x})z.`$
Now, by equation (13), we have
(31)
$$S_{x,y}=\frac{1}{6}D_{x,y}+L_aR_b$$
with
$`a`$ $`=\frac{1}{6}[x,y]\frac{1}{4}(x\overline{y}y\overline{x})𝕂^{}`$
$`b`$ $`=\frac{1}{6}[x,y]\frac{1}{4}(\overline{x}y\overline{y}x)𝕂^{}.`$
Hence, by equation (21), there is an element $`(A,B,C)\mathrm{Tri}𝕂`$ with $`A=S_{x,y}`$ and
$`B`$ $`=\frac{1}{6}D_{x,y}L_aL_bR_b=S_{x,y}L_{2a+b},`$
$`C`$ $`=\frac{1}{6}D_{x,y}+L_a+R_a+R_b=S_{x,y}+R_{a+2b},`$
Writing $`[x,y]=\frac{1}{2}([\overline{x},y]+[x,\overline{y}])`$ gives
$`a+2b`$ $`=\frac{1}{4}(\overline{y}x\overline{x}y)`$
$`2a+b`$ $`=\frac{1}{4}(y\overline{x}x\overline{y})`$
so equations (27) and (29) give
(32)
$$S_{x,y}=\frac{1}{2}Q_{x,y}R_{a+2b}+L_{2a+b}$$
where $`Q_{x,y}z=[x,y,z]`$. Hence
$`Cz`$ $`=\frac{1}{2}[x,y,z]+\frac{1}{4}(y\overline{x}x\overline{y})z`$
$`=\frac{1}{4}y(\overline{x}z)\frac{1}{2}x(\overline{y}z)`$
i.e.
$$C=\frac{1}{4}(L_yL_{\overline{x}}L_xL_{\overline{y}})$$
and similarly
$$B=\frac{1}{4}(R_yR_{\overline{x}}R_xR_{\overline{y}}).$$
Thus $`T=(4S_{x,y},4C,4B)`$ is an element of $`\mathrm{Tri}𝕂`$. ∎
Note that if $`x,y𝕂^{}`$, so that $`\overline{x}=x`$ and $`\overline{y}=y`$, then $`a=b=\frac{1}{12}[x,y]`$ and so
(33)
$$\begin{array}{c}T_{x,y}=(\frac{2}{3}D_{x,y}+\frac{1}{3}L_{[x,y]}\frac{1}{3}R_{[x,y]},\frac{2}{3}D_{x,y}+\frac{1}{3}L_{[x,y]}+\frac{2}{3}R_{[x,y]},\hfill \\ \hfill \frac{2}{3}D_{x,y}\frac{2}{3}L_{[x,y]}\frac{1}{3}R_{[x,y]}).\end{array}$$
The element $`T_{x,y}`$ will be needed to describe $`\mathrm{Tri}𝕂`$ as a Lie subalgebra of $`\mathrm{Der}H_3(𝕂)`$. We will also need an automorphism of $`\mathrm{Tri}𝕂`$ defined as follows. For any linear map $`A:𝕂𝕂`$, let $`\overline{A}=KAK`$, where $`K:𝕂𝕂`$ is the conjugation $`x\overline{x}`$ in $`𝕂`$, i.e.
$$\overline{A}(x)=\overline{A(\overline{x})}.$$
Then $`\overline{\overline{A}}=A`$ and $`\overline{AB}=\overline{A}\overline{B}`$. Note also that
$$\overline{L}_x=R_{\overline{x}}$$
and $`D=\overline{D}`$ if $`D𝔰𝔬(𝕂^{})`$, in particular if D is a derivation of $`𝕂`$.
###### Lemma 4.
Given $`T=(A,B,C)\mathrm{Tri}𝕂`$, let
$$\theta (T)=(\overline{B},C,\overline{A}).$$
Then $`\theta (T)\mathrm{Tri}𝕂`$ and $`\theta `$ is a Lie algebra automorphism.
###### Proof.
By Lemma 2, $`T=T(D,a,b)`$ for some $`D\mathrm{Der}𝕂`$ and $`a,b𝕂^{}`$. Then
$`A`$ $`=D+L_aR_b`$
$`B`$ $`=DL_aL_bR_b`$
$`C`$ $`=D+L_a+R_a+R_b.`$
It follows that
$$\overline{B}=D+R_a+R_b+L_b=D+L_a^{}R_b^{}$$
which is the first component of $`T^{}=(A^{},B^{},C^{})\mathrm{Tri}𝕂`$, where
$`B^{}`$ $`=DL_a^{}L_b^{}R_b^{}`$
$`=DL_b+L_{a+b}+R_{a+b}=C`$
$`C^{}`$ $`=D+L_a^{}+R_a^{}+R_b^{}`$
$`=D+L_b+R_bR_{a+b}=\overline{A},`$
i.e. $`T^{}=(\overline{B},C,\overline{A})=\theta (T)`$. It is clear that $`\theta `$ is a Lie algebra automorphism. ∎
Given $`T=(A,B,C)\mathrm{Tri}𝕂`$, it is convenient to define $`(T_1,T_2,T_3)=(A,\overline{B},\overline{C})`$. Then $`\theta (T)_i=T_{\sigma (i)}`$ where $`\sigma 𝒮_3`$ is the cyclic permutation
$$\sigma (1)=2,\sigma (2)=3,\sigma (3)=1.$$
Write $`a_1=ab`$, $`a_2=a`$, $`a_3=b`$. Then the triality obtained from $`(a,b)`$ can be written in the symmetric form $`T(0,a,b)=(T_1,\overline{T}_2,\overline{T}_3)`$ where
$`T_1`$ $`=L_{a_2}R_{a_3}`$
$`T_2`$ $`=L_{a_3}R_{a_1}`$
$`T_3`$ $`=L_{a_1}R_{a_2}`$
i.e. $`T_i=L_{a_j}R_{a_k}`$ where $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$.
###### Theorem 1.
For any composition algebra $`𝕂`$,
$$\mathrm{Der}H_3(𝕂)=\mathrm{Tri}𝕂\dot{+}3𝕂$$
in which $`\mathrm{Tri}𝕂`$ is a Lie subalgebra, and the brackets in $`[\mathrm{Tri}𝕂,3𝕂]`$ are
(34)
$$[T,F_i(x)]=F_i(T_ix)3𝕂,$$
if $`T=(T_1,\overline{T}_2,\overline{T}_3)\mathrm{Tri}𝕂`$ and $`F_1(x)+F_2(y)+F_3(z)=(x,y,z)3𝕂`$; and the brackets in $`[\mathrm{Tri}𝕂,\mathrm{Tri}𝕂]`$ are given by
(35)
$$[F_i(x),F_j(y)]=F_k(\overline{y}\overline{x})3𝕂,$$
if $`x,y𝕂`$ and $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$; and
(36)
$$[F_i(x),F_i(y)]=\theta ^{1i}(T_{x,y})\mathrm{Tri}𝕂.$$
###### Proof.
Define elements $`e_i,P_i(x)`$ of $`H_3(𝕂)`$ (where $`i=1,2,3;x𝕂`$) by the equation
(37)
$$\left(\begin{array}{ccc}\alpha & z& \overline{y}\\ \overline{z}& \beta & x\\ y& \overline{x}& \gamma \end{array}\right)=\alpha e_1+\beta e_2+\gamma e_3+P_1(x)+P_2(y)+P_3(z)$$
for $`\alpha ,\beta ,\gamma ;x,y,z𝕂`$. Then the Jordan product in $`H_3(𝕂)`$ is given by
(38a) $`e_ie_j`$ $`=2\delta _{ij}e_i`$
(38b) $`e_iP_j(x)`$ $`=(1\delta _{ij})P_j(x)`$
(38c) $`P_i(x)P_i(y)`$ $`=2(x,y)(e_j+e_k)`$
(38d) $`P_i(x)P_j(y)`$ $`=P_k(\overline{y}\overline{x})`$
where in each of the last two equations $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$.
Now let $`D:H_3(𝕂)H_3(𝕂)`$ be a derivation of this algebra. First suppose that
$$De_i=0,i=1,2,3.$$
Then
$`e_iDP_i(x)`$ $`=0`$
$`e_iDP_j(x)`$ $`=DP_j(x)\text{if }ij`$
Thus $`DP_j(x)`$ is an eigenvector of each of the multiplication operators $`L_{e_i}`$, with eigenvalue $`0`$ if $`i=j`$ and $`1`$ if $`ij`$. It follows that
(39)
$$DP_j(x)=P_j(T_jx)$$
for some $`T_j:𝕂𝕂`$. Now
$$DP_j(x)P_j(y)+P_j(x)DP_j(y)=0$$
gives $`T_j𝔰𝔬(𝕂)`$; and the derivation property of $`D`$ applied to equation (38d) gives
$$T_k(\overline{y}\overline{x})=\overline{y}(\overline{T_ix})+(\overline{T_jy})\overline{x}$$
i.e. $`(T_k,\overline{T_i},\overline{T_j})\mathrm{Tri}𝕂`$ and therefore $`(T_1,\overline{T_2},\overline{T_3})\mathrm{Tri}𝕂`$.
If $`De_i0`$, then from equation (38a) with $`i=j`$,
$$2e_iDe_i=2De_i$$
so $`De_i`$ is an eigenvector of the multiplication $`L_{e_i}`$ with eigenvalue $`1`$, i.e. $`De_iP_j(𝕂)+P_k(𝕂)`$ where $`(i,j,k)`$ are distinct. Write
$$De_i=P_j(x_{ij})+P_k(x_{ik});$$
then equation (38a) with $`ij`$ gives
$$e_iP_k(x_{jk})+e_iP_i(x_{ji})+P_j(x_{ij})e_j+P_k(x_{ik})e_j=0.$$
Thus
$$P_k(x_{jk}+x_{ik})=0.$$
It follows that the action of any derivation on the $`e_i`$ must be of the form $`F_1(x)+F_2(y)+F_3(z)`$ where
$`F_i(x)e_i`$ $`=0`$
(40) $`F_i(x)e_j=`$ $`F_i(x)e_k=P_i(x),`$
$`(i,j,k)`$ being a cyclic permutation of $`(1,2,3)`$. Hence $`\mathrm{Der}H_3(𝕂)\mathrm{Tri}𝕂𝕂^3`$.
To show that such derivations $`F_i(x)`$ exist and therefore the inclusion just mentioned is an equality, consider the operation of commutation with the matrix
$`X`$ $`=\left(\begin{array}{ccc}0& z& \overline{y}\\ \overline{z}& 0& x\\ y& \overline{x}& 0\end{array}\right)`$
$`=X_1(x)+X_2(y)+X_3(z)`$
i.e. define $`F_i(x)=C_{X_i(x)}`$ where $`C_X:H_3(𝕂)H_3(𝕂)`$ is the commutator map
(41)
$$C_X(H)=XHHX.$$
This satisfies equation (1) and also
$`F_i(x)P_i(y)`$ $`=2(x,y)(e_je_k)`$
(42) $`F_i(x)P_j(y)`$ $`=P_k(\overline{y}\overline{x})`$
$`F_i(x)P_k(y)`$ $`=P_j(\overline{x}\overline{y}).`$
It is a derivation of $`H_3(𝕂)`$ by virtue of the matrix identity
(43)
$$[X,\{H,K\}]=\{[X,H],K\}+\{H,[X,K]\}$$
(in which square brackets denote commutators and round brackets denote anticommutators), which we will prove separately in lemma 7.
The Lie brackets of these derivations follow from another matrix identity which is also proved in lemma 7,
(44)
$$[X,[Y,H]][Y,[X,H]]=[[X,Y],H]E(X,Y)H$$
where $`E(X,Y)𝔰𝔬(𝕂^{})`$ is defined by
$$E(X,Y)z=\underset{ij}{}[x_{ij},y_{ji},z],$$
$`x_{ij},y_{ji}`$ being the matrix elements of $`X`$ and $`Y`$. If $`X=X_i(x)`$ and $`Y=X_j(y)`$ we have $`D(X,Y)=0`$ and
$$[X_i(x),X_j(y)]=X_k(\overline{y}\overline{x})$$
where $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$. This yields the Lie bracket (35). If $`X=X_i(x)`$ and $`Y=X_j(y)`$, the matrix commutator $`Z=[X,Y]`$ is diagonal with $`z_{ii}=0`$, $`z_{jj}=y\overline{x}x\overline{y}`$ and $`z_{kk}=\overline{y}x\overline{x}y`$ ($`i,j,k`$ cyclic). Hence the action of the commutator $`[F_i(x),F_i(y)]=C_z+E(X,Y)`$ on $`H_3(𝕂)`$ is
$`[F_i(x),F_i(y)]e_m`$ $`=0(m=i,j,k)`$
$`[F_i(x),F_i(y)]P_i(w)`$ $`=P_i(z_{jj}wwz_{kk}2[x,y,w])=P_i(T_1w)`$
$`=4P_i(S_{xy}w)\text{by equation (}\text{17}\text{).}`$
$`[F_i(x),F_i(y)]P_j(w)`$ $`=P_j(z_{kk}w2[x,y,w])`$
$`=P_j(\overline{y}(xw)\overline{x}(yw))`$
$`[F_i(x),F_i(y)]P_k(w)`$ $`=P_k(wz_{jj}2[x,y,w])`$
$`=P_k((wx)\overline{y}(wy)\overline{x}).`$
Comparing with lemma 3, we see that
$`[F_i(x),F_i(y)]P_i(w)`$ $`=P_i(T_1w)=P_i(T_i^{}w)`$
$`[F_i(x),F_i(y)]P_j(w)`$ $`=P_j(\overline{T_2}w)=P_j(\overline{T_j^{}}w)`$
$`[F_i(x),F_i(y)]P_k(w)`$ $`=P_k(\overline{T_3}w)=P_k(\overline{T_k^{}}w)`$
where $`(T_1,T_2,T_3)=T_{xy}`$, so that $`T^{}=\theta ^{1i}(T_{xy})`$. This establishes the Lie bracket (36). ∎
The matrix identities needed in the proof of Theorem 1 are contained in the following, in which we include a third identity for the sake of completeness:
###### Lemma 5.
Let $`𝕂`$ be a composition algebra, let $`H`$, $`K`$ and $`L`$ be hermitian $`3\times 3`$ matrices with entries from $`𝕂`$, and let $`X,Y`$ be traceless antihermitian matrices over $`𝕂`$. Then
(45a) $`[X,\{H,K\}]`$ $`=\{[X,H],K\}+\{H,[X,K]\}`$
(45b) $`[X,[Y,H]][Y,[X,H]]`$ $`=[[X,Y],H]+E(X,Y)H`$
(45c) $`\{H\{K,L\}\}\{K\{H,L\}\}`$ $`=[[H,K],L]+E(H,K)L`$
where $`E(X,Y)𝔰𝔬(𝕂)`$ is defined for any $`3\times 3`$ matrices $`X,Y`$ by
(46)
$$E(X,Y)z=\underset{ij}{}[x_{ij},y_{ji},z].$$
In these matrix identities the square brackets denote commutators and the chain brackets denote anticommutators.
###### Proof.
We consider first part (a). The difference between the two sides can be written in terms of matrix associators, where the $`(i,j)`$th element is
(47)
$$\begin{array}{c}\underset{mn}{}([x_{im},h_{mn},k_{nj}]+[x_{im},k_{mn},h_{nj}]\hfill \\ \hfill +[k_{im},h_{mn},x_{nj}][h_{im},x_{mn},k_{nj}][k_{im},x_{mn},h_{nj}]).\end{array}$$
Suppose $`ij`$ and let $`k`$ be the third index. Since the diagonal elements of $`H`$ and $`K`$ are real, any associator containing them vanishes. Hence the terms containing $`x_{ij}`$ or $`x_{ji}`$ are
$$\begin{array}{c}\underset{n}{}([x_{ij},h_{jn},k_{nj}]+[x_{ij},k_{jn},h_{nj}])+\underset{m}{}([h_{im},k_{mi},x_{ij}]+[k_{im},h_{mi},x_{ij}])\hfill \\ \hfill [h_{ij},x_{ji},k_{ij}]+[k_{ij},x_{ji},h_{ij}]=0\end{array}$$
by the alternative law, the hermiticity of $`H`$ and $`K`$, and the fact that an associator changes sign when one of its elements is conjugated. The terms containing $`x_{ik}`$ or $`x_{ki}`$ are
$$[x_{ik},h_{ki},k_{ij}]+[x_{ik},k_{ki},h_{ij}][h_{ik},x_{ki},k_{ij}][k_{ik},x_{ki},h_{ij}]=0$$
using also $`x_{ki}=\overline{x}_{ik}`$. Similarly, the terms containing $`x_{jk}`$ or $`x_{kj}`$ vanish. Finally, the terms containing $`x_{ii},x_{jj}`$ and $`x_{kk}`$ are
$$\begin{array}{c}[x_{ii},h_{ik},k_{kj}]+[x_{ii},k_{ik},h_{kj}]+[h_{ik},k_{kj},x_{jj}]+[k_{ik},h_{kj},x_{jj}]\hfill \\ \hfill [h_{ik},x_{kk},k_{kj}][k_{ik},x_{kk},h_{kj}]=0\end{array}$$
since $`x_{ii}+x_{jj}+x_{kk}=0`$.
Now consider the $`(i,i)`$th element. The last two terms of equation (47) become
$$\underset{mn}{}\left([h_{im},x_{mn},k_{ni}]+[k_{in},x_{nm},h_{mi}]\right)=0.$$
Let $`j`$ be one of the other two indices. The terms containing $`x_{ij}`$ or $`x_{ji}`$ are
$$[x_{ij},h_{jk},k_{ki}]+[x_{ij},k_{jk},h_{ki}]+[h_{ik},k_{kj},x_{ji}]+[k_{ik},h_{kj},x_{ji}]=0,$$
where $`k`$ is the third index. There are no terms containing $`x_{jk}`$ or $`x_{kj}`$. The terms containing $`x_{ii},x_{jj}`$ or $`x_{kk}`$ are
$$\begin{array}{c}\underset{n}{}\left([x_{ii},h_{in},k_{ni}]+[x_{ii},k_{in},h_{ni}]\right)\hfill \\ \hfill +\underset{m}{}\left([h_{im},k_{mi},x_{ii}]+[k_{im},h_{mi},x_{ii}]\right)=0.\end{array}$$
Thus in all cases the expression (47) vanishes, proving (a). Parts (b) and (c) are proved by similar arguments, which the reader will find more entertaining to write than to read. ∎
In any Jordan algebra, the commutator of two multiplication operators $`L_x`$ and $`L_y`$ is a derivation (this fact is used in the construction of the magic square Lie algebras $`L_3(𝕂_1,𝕂_2)`$; see ). In the case of the Jordan algebra $`H_3(𝕂)`$, we can identify these derivations as follows:
###### Lemma 6.
In $`H_3(𝕂)`$, where $`𝕂`$ is any composition algebra,
$`[L_{e_i},L_{e_j}]`$ $`=0`$
$`[L_{e_i},L_{P_j(x)}]`$ $`=ϵ_{ij}L_{P_j(x)}`$
where $`ϵ_{ij}=0`$ if $`i=j`$, otherwise $`ϵ_{ij}`$ is the sign of the permutation $`(i,j,k)`$ of $`(1,2,3)`$ where $`k`$ is the third index,
$`[L_{P_i(x)},L_{P_i(y)}]`$ $`=T_{x,y}`$
$`[L_{P_i(x)},L_{P_j(y)}]`$ $`=F_k(\overline{y}\overline{x})`$
where $`e_i,P_i(x)H_3(𝕂)`$ are defined by (37) and $`F_k(x)\mathrm{Der}H_3(𝕂)`$ is given by (1) and (1).
###### Proof.
Straightforward calculation from (38a38d). ∎
The proof of Theorem 1 suggests an alternative description of $`\mathrm{Der}H_3(𝕂)`$, which leads us to identify it as $`𝔰𝔞(3,𝕂)`$:
###### Theorem 2.
For any composition algebra $`𝕂`$,
(48)
$$\mathrm{Der}H_3(𝕂)=\mathrm{Der}𝕂\dot{+}A_3^{}(𝕂)$$
in which $`\mathrm{Der}𝕂`$ is a Lie subalgebra, the Lie brackets between $`\mathrm{Der}𝕂`$ and $`A_3^{}(𝕂)`$ are given by the elementwise action of $`\mathrm{Der}𝕂`$ on $`3\times 3`$ matrices, and
$$[X,Y]=(XYYX)^{}+\frac{1}{3}D(X,Y)$$
where $`X,YA_3^{}(𝕂)`$,
$$(XYYX)^{}=XYYX\frac{1}{3}\mathrm{tr}(XYYX)\text{1}\text{1}A_3^{}(𝕂)$$
and
$$D(X,Y)=\underset{ij}{}D(x_{ij},y_{ji})\mathrm{Der}𝕂$$
$`x_{ij},y_{ji}`$ being the matrix elements of $`X`$ and $`Y`$ and $`D(x,y)`$ being the derivation $`D_{x,y}`$ defined in equation (12).
###### Proof.
By Lemma 2 and Theorem 1
(49)
$$\mathrm{Der}H_3(𝕂)=\mathrm{Der}𝕂\dot{+}2𝕂^{}\dot{+}3𝕂.$$
Identify $`(a,b)+(x,y,z)2𝕂^{}\dot{+}3𝕂`$ with the traceless antihermitian matrix
$$X=\left(\begin{array}{ccc}ab& z& \overline{y}\\ \overline{z}& a& x\\ y& \overline{x}& b\end{array}\right)A_3^{}(𝕂);$$
then the actions of $`2𝕂^{}`$ and $`3𝕂`$ on $`H_3(𝕂)`$ defined in Theorem 1 are together equivalent to the commutator action $`C_x`$ defined by equation (41). By Lemma 5(b),
$$[C_X,C_Y]=C_{(XYYX)^{}}+C_{t\text{1}\text{1}}+E(X,Y)$$
where
$`t`$ $`=\frac{1}{3}\mathrm{tr}(XYYX)`$
$`=\frac{1}{3}{\displaystyle \underset{ij}{}}(x_{ij}y_{ji}y_{ji}x_{ij}).`$
Now $`C_{t\text{1}\text{1}}+E(X,Y)`$ acts elementwise on matrices in $`H_3(𝕂)`$ according to the map $`D:𝕂𝕂`$ given by
$`Dz`$ $`=[t,z]+E(X,Y)z`$
$`={\displaystyle \underset{ij}{}}\left(\frac{1}{3}[[x_{ij},y_{ji}],z][x_{ij},y_{ji},z]\right)`$
$`=\frac{1}{3}D(X,Y)Z.`$
Hence the bracket $`[X,Y]`$ is as stated. ∎
Finally we use Theorem 1 to give a description of $`L_3(𝕂_1,𝕂_2)`$ which makes manifest the symmetry between $`𝕂_1`$ and $`𝕂_2`$.
###### Theorem 3.
For any two composition algebras $`𝕂_1,𝕂_2`$,
(50)
$$L_3(𝕂_1,𝕂_2)=\mathrm{Tri}𝕂_1\mathrm{Tri}𝕂_2\dot{+}3𝕂_1𝕂_2$$
in which $`\mathrm{Tri}𝕂_1\mathrm{Tri}𝕂_2`$ is a Lie subalgebra;
(51) $`[T_1,F_i(xy)]`$ $`=F_i(T_{1i}x_1x_2)3𝕂_1𝕂_2`$
(52) $`[T_2,F_i(xy)]`$ $`=F_i(x_1T_{2i}x_2)3𝕂_1𝕂_2`$
if $`T_\alpha =(T_{\alpha 1},\overline{T}_{\alpha 2},\overline{T}_{\alpha 3})\mathrm{Tri}𝕂(\alpha =1,2)`$, and
$`F_1(x_1x_2)+F_2(y_1y_2)+F_3(z_1z_2)`$ $`=(x_1x_2,y_1y_2,z_1z_2)`$
$`3𝕂_1𝕂_2;`$
(53) $`[F_i(x_1x_2),F_j(y_1y_2)]`$ $`=F_k(\overline{y}_1\overline{x}_1\overline{y}_2\overline{x}_2)`$
$`3𝕂_1𝕂_2`$
if $`x_\alpha ,y_\alpha 𝕂_2`$ and $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$; and
(54)
$$\begin{array}{c}[F_i(x_1x_2),F_i(y_1y_2)]=x_2,y_2\theta ^{1i}T_{x_1y_1}+x_1,y_1\theta ^{1i}T_{x_2y_2}\hfill \\ \hfill \mathrm{Tri}𝕂_1\mathrm{Tri}𝕂_2\end{array}$$
###### Proof.
We can write
$$H_3^{}(𝕂_1)=23𝕂$$
by identifying $`(\alpha ,\beta )+(x,y,z)23𝕂`$ with the matrix
$`\left(\begin{array}{ccc}\alpha \beta & z& \overline{y}\\ \overline{z}& \alpha & x\\ y& \overline{x}& \beta \end{array}\right)H_3^{}(𝕂),`$
$`=\alpha (e_2e_1)+\beta (e_3e_1)+P_1(x)+P_2(y)+P_3(z)`$
in the notation of theorem 1. Then the vector space structure (10) of $`L_3(𝕂_1,𝕂_2)`$ can be written using Theorem 1, as
$`L_3(𝕂_1,𝕂_2)`$ $`=\mathrm{Der}H_3(𝕂_1)\dot{+}H_3^{}(𝕂_1)𝕂_2^{}\dot{+}\mathrm{Der}𝕂_2`$
$`=(\mathrm{Tri}𝕂\dot{+}3𝕂_1)\dot{+}(2𝕂_2^{}\dot{+}3𝕂_12𝕂_2^{})\dot{+}\mathrm{Der}𝕂_2`$
$`=\mathrm{Tri}𝕂_1\dot{+}(\mathrm{Der}𝕂_2\dot{+}2𝕂_2^{})\dot{+}(3𝕂_1𝕂_2^{}\dot{+}3𝕂_1)`$
$`\mathrm{Tri}𝕂_1\dot{+}\mathrm{Tri}𝕂_2\dot{+}3𝕂_1𝕂_2.`$
We use the following notation for the elements of the five subspaces of $`L_3(𝕂_1,𝕂_2)`$:
1. $`\mathrm{Tri}𝕂\mathrm{Der}H_3(𝕂_1)`$ contains elements $`T=(T_1,\overline{T}_2,\overline{T}_3)`$ acting on $`H_3^{}(𝕂_1)`$ as in Theorem 1:
$$Te_i=0,TP_i(x)=P_i(T_ix)(x𝕂;i=1,2,3)$$
2. $`3𝕂_1`$ is the subspace of $`\mathrm{Der}H_3(𝕂_1)`$ containing the elements $`F_i(x)`$ defined in Theorem 1; these will be identified with the elements $`F_i(x1)3𝕂_1𝕂_2^{}`$.
3. $`2𝕂_2^{}`$ is the subspace $`\mathrm{\Delta }𝕂_2^{}`$ of $`H_3(𝕂_1)𝕂_2^{}`$, where $`\mathrm{\Delta }H_3^{}(𝕂_1)`$ is the subspace of real, diagonal, traceless matrices and is identified with the subspace of $`\mathrm{Tri}𝕂`$ as described in Lemma 2. We will regard $`2𝕂_2^{}`$ as a subspace of $`3𝕂_2^{}`$, namely
$$2𝕂_2^{}=\{(a_1,a_2,a_3)3𝕂_2^{}:a_1+a_2+a_3=0\}$$
and identify $`𝐚=(a_1,a_2,a_3)`$ with the $`3\times 3`$ matrix
$$\mathrm{\Delta }(𝐚)=\left(\begin{array}{ccc}a_1& 0& 0\\ 0& a_2& 0\\ 0& 0& a_3\end{array}\right)H_3^{}(𝕂_1)𝕂_2^{}$$
and with the triality $`T(𝐚)=(T_1,\overline{T}_2,\overline{T}_3)`$ where $`T_i=L_{a_j}R_{a_k}`$ (see the remark after the proof of Lemma 4).
4. $`3𝕂_1𝕂_2^{}`$ is the subspaces of $`H_3(𝕂_1𝕂_2^{}`$ spanned by elements $`P_i(x)a`$ $`(i=1,2,3:x𝕂_1,a𝕂_2^{})`$; it is also a subspace of $`3𝕂_1𝕂_2`$ in the obvious way.
5. $`\mathrm{Der}𝕂_2`$ is a subspace of $`\mathrm{Tri}𝕂_2`$, a derivation $`D`$ being identified with $`(D,D,D)\mathrm{Tri}𝕂_2`$.
To complete the proof we must verify that the Lie brackets defined by Tits (see Section 3.1) coincide with those in the statement of the theorem. The above decomposition of $`L_3(𝕂_1,𝕂_2)`$ into five parts gives us fifteen types of bracket to examine. We will write $`[,]_{\text{Tits}}`$ for the bracket defined in section 3.1 and $`[,]_{\text{here}}`$ for that defined above.
1. $`[\mathrm{Tri}𝕂_1,\mathrm{Tri}𝕂_2]`$: For $`T_1,T_2\mathrm{Tri}𝕂_1`$, $`[T_1,T_2]_{\text{Tits}}`$ is the bracket in $`\mathrm{Der}H_3(𝕂_1)`$, which by theorem 1 is the same as $`[T_1,T_2]_{\text{here}}`$.
2. $`[\mathrm{Tri}𝕂_1,3𝕂_1]`$: For $`T\mathrm{Tri}𝕂_1`$, $`F_i(x1)3𝕂_1`$,
$`[T,F_i(x)]_{\text{Tits}}`$ $`=F_1(T_ix)\text{see Theorem }\text{1}`$
$`=F_i(T_ix1)=[T,F_i(x1)]_{\text{here}}.`$
3. $`[\mathrm{Tri}𝕂_1,2𝕂_2]`$: For $`T_1\mathrm{Tri}𝕂_1`$, $`(a,b,c)2𝕂_2^{}`$,
$$[T_1,(a,b,c)]_{\text{Tits}}=[T,e_1a+e_2b+e_3c]=0$$
since in Theorem 1 $`\mathrm{Tri}𝕂_1`$ was obtained as the subspace of derivations which annihilate the diagonal matrices $`e_i`$. On the other hand,
$$[T_1,(a,b,c)]_{\text{here}}=[T_1,T_2(a,b,c)]=0.$$
4. $`[\mathrm{Tri}𝕂_1,3𝕂_1𝕂_2^{}]`$: For $`T_1\mathrm{Tri}𝕂_1`$, $`P_i(xa)3𝕂_1𝕂_2^{}`$,
(55)
$$[T_1,P_i(xa)=P_i(T_{1i}xa)=[T_1,P_i(xa)]_{\text{here}}.$$
5. $`[\mathrm{Tri}𝕂_1,\mathrm{Der}𝕂_2]_{\text{Tits}}[\mathrm{Der}H_3(𝕂_1),\mathrm{Der}𝕂_2]=0`$, while
$`[\mathrm{Tri}𝕂_1,\mathrm{Der}𝕂_2]_{\text{here}}[\mathrm{Tri}𝕂_1,\mathrm{Tri}𝕂_2]=0`$.
6. $`[3𝕂_1,3𝕂_1]`$: $`3𝕂_1=3𝕂_1`$ is spanned by $`F_i(x)=F_i(x1)`$
$`(i=1,2,3;x𝕂_1)`$, and $`[F_i(x),F_j(y)]_{\text{Tits}}`$ is given by Theorem 1, while $`[F_i(x1),F_j(y1)]_{\text{here}}`$ is the same since $`T_{x_2,y_2}=0`$ if $`x_2,y_2`$.
7. $`[3𝕂_1,2𝕂_2^{}]`$: For $`F_i(x)3𝕂_1`$, $`𝐚=(a_1,a_2,a_3)2𝕂_2^{}`$ with $`(a_1+a_2+a_3=0)`$,
$`[F_i(x),𝐚]_{\text{Tits}}`$ $`=[F_i(x),{\displaystyle e_ia_i}][\mathrm{Der}H_3(𝕂_1),H_3^{}(𝕂_1)𝕂_2^{}]`$
$`=P_i(x)(a_ja_k)\text{by (}\text{1}\text{)}`$
while
$`[F_i(x),𝐚]_{\text{here}}`$ $`=[F_i(x1),T(𝐚)][3𝕂_1𝕂_2,\mathrm{Tri}𝕂_2]`$
$`=F_i(x(a_ja_k)).`$
8. $`[3𝕂_1,3𝕂_1𝕂_2^{}]`$: For $`F_i(x)3𝕂_1,F_j(ya)3𝕂_1𝕂_2^{}`$,
$`[F_i(x),F_i(ya)]_{\text{Tits}}`$ $`=[F_i(x),P_i(y)a][\mathrm{Der}H_3(𝕂_1),H_3^{}(𝕂_1)𝕂_2^{}]`$
$`=2x,y(e_je_k)aH_3^{}(𝕂_1𝕂_2^{}`$
$`=2x,y(a_1,a_2,a_3)2𝕂_2^{}`$
where $`a_i=0,a_j=a,a_k=a`$ ($`i,j,k`$ cyclic). On the other hand
$`[F_i(x),F_i(ya)]_{\text{here}}`$ $`=[F_i(x1),F_i(ya)]`$
$`=x,y\theta ^{1i}T_{1,a}`$
$`=[F_i(x),F_i(ya)]_{\text{Tits}}`$
since $`T_{1,a}=(2L_a+R_a,2R_a,2L_a)`$ which is identified with $`(0,2a,2a)`$
$`𝕂_2^{}`$ in paragraph 3 above. If $`ij`$ and $`(i,j,k)`$ is a cyclic permutation of $`1,2,3)`$,
$`[F_i(x),F_j(ya)]_{\text{Tits}}`$ $`=[F_i(x),P_j(ya)][\mathrm{Der}H_3(𝕂_1),H_3^{}(𝕂_1)𝕂_2]`$
$`=P_k(\overline{y}\overline{x})aH_3^{}(𝕂_1)𝕂_2^{}`$
$`=F_k(\overline{y}\overline{x})a3𝕂_1𝕂_2^{}`$
while $`[F_i(x),F_j(ya)]_{\text{here}}=F_k(\overline{y}\overline{x})a`$ since $`\overline{a}=a`$. Similarly,
$$[F_i(x),F_k(ya)]_{\text{Tits}}=F_k(\overline{x}\overline{y})a=[F_i(x),F_k(ya)]_{\text{here}}.$$
9. $`[3𝕂_1,\mathrm{Der}𝕂_2]_{\text{Tits}}[\mathrm{Der}H_3(𝕂_1),\mathrm{Der}𝕂_2]=0`$
and
$$[3𝕂_1,\mathrm{Der}𝕂_2]_{\text{here}}[3𝕂_1,\mathrm{Der}𝕂_2]=0.$$
10. $`[2𝕂_2^{},2𝕂_2^{}]`$: For $`𝐚,𝐛2𝕂_2^{}`$, with $`𝐚=(a_1,a_2,a_3)`$ and $`𝐛=(b_1,b_2,b_3)`$ where $`a_1+a_2+a_3=b_1+b_2+b_3=0`$,
$`[𝐚,𝐛]_{\text{Tits}}`$ $`=[{\displaystyle e_ia_i},{\displaystyle e_jb_j}]`$
$`={\displaystyle \underset{i,j}{}}\left(e_i,e_jD_{a_i,b_j}+(e_ie_j)\mathrm{Im}(a_ib_j)+a_i,b_j[L_{e_i},L_{e_j}]\right)`$
$`={\displaystyle \underset{ij}{}}\left(2\delta _{ij}D_{a_i,b_j}+\delta _{ij}(2e_i\frac{2}{3}\text{1}\text{1})\frac{1}{2}[a_i,b_j]\right)`$
$`={\displaystyle \underset{i}{}}\left(2D_{a_i,b_i}+\frac{1}{3}e_i(2[a_i,b_i][a_j,b_j][a_k,b_k])\right)`$
$`=[𝐚,𝐛]_{\text{here}}`$
11. $`[2𝕂_2^{},3𝕂_1𝕂_2^{}]`$: For $`𝐚=(a_1,a_2,a_3)2𝕂_2^{}`$ and $`F_i(xb)3𝕂_1𝕂_2^{}`$,
$`[𝐚,F_i(xb)]_{\text{Tits}}`$ $`=[e_ia_i+e_ja_j+e_ka_k,P_i(x)b]`$
$`=P_i(x)\frac{1}{2}[a_j,b]+P_i(x)\frac{1}{2}[a_k,b]a_ja_k,bF_i(x)`$
using Lemma 6. The first two terms belong to the subspace $`3𝕂_1𝕂_2^{}`$ of $`H_3^{}𝕂_2^{}`$ and the third to the subspace $`3𝕂_1`$ of $`\mathrm{Der}H_3(𝕂_1)`$, so together they constitute an element of $`3𝕂_1𝕂_2`$:
$`[𝐚,F_i(xb)]_{\text{Tits}}`$ $`=F_i(x(\frac{1}{2}[a_j,b]=\frac{1}{2}[a_k,b]a_ja_k,b))`$
$`=F_i(x(a_jbba_k))`$
$`=F_i(xT(𝐚)_ib)`$
$`=[𝐚,F_i(xb)]_{\text{here}}`$
12. $`[2𝕂_2^{},\mathrm{Der}𝕂_2]`$: Tits’s bracket (3.1) coincides with the bracket in $`\mathrm{Tri}𝕂_2`$ as given by Lemma 2.
13. $`[3𝕂_1𝕂_2^{},3𝕂_1𝕂_2^{}]`$: For $`P_i(x),P_i(y)3𝕂_1`$ and $`a,b𝕂_2^{}`$, if $`ij`$ and $`(i,j,k)`$ is a cyclic permutation of $`(1,2,3)`$ then
$$[P_i(x)a,P_j(y)b]_{\text{Tits}}=P_k(\overline{y}\overline{x})\frac{1}{2}[a,b]a,bF_k(\overline{y}\overline{x})$$
by equation (38d) and Lemma 6. This is an element of $`3𝕂_1𝕂_2^{}\dot{+}3𝕂_1`$ which is identified with the following element of $`3𝕂_1𝕂_2`$:
$`F_k(\overline{y}\overline{x}\frac{1}{2}([a,b](a\overline{b}+b\overline{a}))`$ $`=F_k(\overline{y}\overline{x}\overline{b}\overline{a})`$
$`=[F_i(xa),F_j(yb)]_{\text{here}}.`$
If $`i=j`$, then
$$\begin{array}{c}[P_i(x)a,P_i(y)b]_{\text{Tits}}=4x,yD_{a,b}+\hfill \\ \hfill \frac{1}{2}x,y(2e_i+e_j+e_k)[a,b]a,b\theta ^{1i}T_{x,y}\end{array}$$
by Lemma 6. The second term belongs to the subspace $`2𝕂_2^{}`$ and it is to be identified with the triality $`\frac{1}{3}x,y\theta ^{1i}T`$ where $`T_1=\frac{1}{3}(L_{[a,b]}R_{[a,b]})`$, $`\overline{T}_2=\frac{1}{3}(R_{[a,b]}+2L_{[a,b]})`$ and $`\overline{T}_3=\frac{1}{3}(2R_{[a,b]}+L_{[a,b]})`$. By (33), $`T=T_{a,b}`$. Hence
$$[P_i(x)a,P_i(y)b]_{\text{Tits}}=x,y\theta ^{1i}T_{a,b}a,b\theta ^{1i}T_{x,y}.$$
14. $`[3𝕂_1𝕂_2^{},\mathrm{Der}𝕂_2]`$ is given by the action of $`\mathrm{Der}𝕂_2`$ on the second factor of the tensor product in both cases.
15. $`[\mathrm{Der}𝕂_2,\mathrm{Der}𝕂_2]`$ is given by the Lie bracket of $`\mathrm{Der}𝕂_2`$ in both cases.
## 5. Magic Squares of $`2\times 2`$ Matrix Algebras: Proofs
In this section we prove the following theorems.
###### Theorem 4.
For $`𝕂=,,`$ and $`𝕆`$,
$`L_2(𝕂_1,𝕂_2)`$ $`𝔰𝔬(𝕂_1\dot{+}𝕂_2)`$
$`L_2(𝕂_1,\stackrel{~}{𝕂}_2)`$ $`𝔰𝔬(\frac{1}{2}(\nu _1+\nu _2),\frac{1}{2}\nu _2).`$
###### Theorem 5.
The following isomorphisms are true for $`𝕂=,,`$ and $`𝕆`$.
(56a) $`L_2(𝕂,)`$ $`\mathrm{Der}H_2(𝕂)`$
(56b) $`L_2(𝕂,\stackrel{~}{})`$ $`\mathrm{Str}H_2(𝕂)`$
(56c) $`L_2(𝕂,\stackrel{~}{})`$ $`\mathrm{Con}H_2(𝕂)`$
We prove Theorem 4 by first showing that $`L_2(𝕂_1,𝕂_2)`$ is isomorphic to the Lie algebra of the pseudo-orthogonal group $`\mathrm{O}(𝕂_1\dot{+}𝕂_2)`$ of linear transformations of $`𝕂_1\dot{+}𝕂_2`$ preserving the quadratic form
$$|x_1+x_2|^2=|x_1|^2+|x_2|^2.(x_1𝕂_1,x_2𝕂_2)$$
We then prove equation (56c) and notice that the proof of this contains the isomorphisms for equations (56a) and (56b). The proofs of these equations will require the use of the following Theorem, and associated Lemmas.
###### Theorem 6.
The derivation algebra of $`H_2(𝕂)`$ can be expressed in the form
(57)
$$\mathrm{Der}H_2(𝕂)=A_2^{}(𝕂)\dot{+}𝔰𝔬(𝕂^{}).$$
Our proof requires the use of the lemma
###### Lemma 7.
Let $`AA_n(𝕂)`$ and $`X,YH_n(𝕂)`$ where $`𝕂`$ is any alternative algebra. The identity
(58)
$$[A,\{X,Y\}]=\{[A,X],Y\}+\{X,[A,Y]\},$$
(where the brackets denote commutators of matrices) holds if $`n=2`$ or if $`n=3`$ and $`\mathrm{tr}A=0`$.
###### Proof.
The $`3\times 3`$ case was proved in Lemma 5. The $`2\times 2`$ case can be deduced from it by considering the $`3\times 3`$ matrices
$$\stackrel{~}{A}=\left(\begin{array}{cc}A& 0\\ 0& \mathrm{tr}A\end{array}\right),\stackrel{~}{X}=\left(\begin{array}{cc}X& 0\\ 0& 1\end{array}\right),\stackrel{~}{Y}=\left(\begin{array}{cc}Y& 0\\ 0& 1\end{array}\right).$$
###### Proof of theorem 6.
From lemma 7 we see that for each $`AA_2(𝕂)`$ there is a derivation $`D(A)`$ of $`H_2(𝕂)`$ given by
$$D(A)(X)=AXXA.$$
We consider $`H_2(𝕂)`$ as a Jordan algebra with product
$$\left(\begin{array}{cc}\alpha & x\\ \overline{x}& \beta \end{array}\right)\left(\begin{array}{cc}\gamma & y\\ \overline{y}& \delta \end{array}\right)=\left(\begin{array}{cc}2\alpha \gamma +2\mathrm{Re}(x\overline{y})& (\gamma +\delta )x+(\alpha +\beta )y\\ (\gamma +\delta )\overline{x}+(\alpha +\beta )\overline{y}& 2\beta \delta +2\mathrm{Re}(\overline{x}y)\end{array}\right).$$
We can write a matrix $`AH_2(𝕂)`$ as follows
$$\left(\begin{array}{cc}\alpha & x\\ \overline{x}& \beta \end{array}\right)=\lambda I+\mu E+P(x)$$
where $`\lambda =\frac{1}{2}(\alpha +\beta )`$, $`\mu =\frac{1}{2}(\alpha \beta )`$, $`E=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)`$ and $`P(x)=\left(\begin{array}{cc}0& x\\ \overline{x}& 0\end{array}\right)`$. Then the Jordan multiplication can be rewritten as
$`EE`$ $`=I`$
$`P(x)P(y)`$ $`=2x,yI`$
$`EP(x)`$ $`=0.`$
Thus $`H_2(𝕂)`$ can be identified with $`𝕁(V)`$, the Jordan algebra associated with the inner product space $`V=𝕂`$. $`𝕁(V)`$ is a subalgebra of the anticommutator algebra of $`\mathrm{Cl}(V)`$, where $`𝐯𝐰=𝐯,𝐰1`$. Derivations of this algebra must satisfy
$`D(1)`$ $`=0`$
$`1,D(𝐯)`$ $`=0.`$
Thus
$$D(𝐯),𝐰+D(𝐰),𝐯=0$$
i.e. $`D`$ is an antisymmetric map of $`𝐯`$. Hence $`\mathrm{Der}H_2(𝕂)=𝔬(𝕂\dot{+})`$.
Considering the matrix structure of $`𝔬(𝕂\dot{+})`$ we can write this as $`𝔬(𝕂)\dot{+}𝕂`$. Consider the action of $`𝕂`$ on the $`(\nu +1)\times 1`$ column vectors $`\left(\begin{array}{c}0\\ 1\end{array}\right)`$ and $`\left(\begin{array}{c}x\\ 0\end{array}\right)`$. We express $`k𝕂`$ as the final row and column in a $`(\nu +1)\times (\nu +1)`$ block matrix. Then
$`\left(\begin{array}{cc}0& k\\ k^t& 0\end{array}\right)\left(\begin{array}{c}0\\ 1\end{array}\right)`$ $`=\left(\begin{array}{c}k\\ 0\end{array}\right)`$
$`\left(\begin{array}{cc}0& k\\ k^t& 0\end{array}\right)\left(\begin{array}{c}x\\ 0\end{array}\right)`$ $`=\left(\begin{array}{c}0\\ k^tx\end{array}\right)`$
Thus $`k`$ maps $`E`$ to $`P(k)`$ and $`P(x)`$ to $`k,xE`$. Now
$$[\left(\begin{array}{cc}0& k\\ \overline{k}& 0\end{array}\right),\left(\begin{array}{cc}0& x\\ \overline{x}& 0\end{array}\right)]=2\left(\begin{array}{cc}k,x& k\\ \overline{k}& k,x\end{array}\right)$$
i.e. multiplication by $`\left(\begin{array}{cc}0& k\\ k^t& 0\end{array}\right)`$ in $`𝕁(V)`$ is equivalent to commutation with $`\left(\begin{array}{cc}0& \frac{k}{2}\\ \frac{\overline{k}}{2}& 0\end{array}\right)`$ in $`H_2(𝕂)`$.
We can split $`𝔬(𝕂)=𝔬(𝕂^{})`$ into $`𝔬(𝕂^{})\dot{+}𝕂`$. Consider the action of the $`\nu \times \nu `$ matrix $`\left(\begin{array}{cc}0& l\\ l^t& 0\end{array}\right)`$ with $`l𝕂^{}`$ on the vectors $`\left(\begin{array}{c}y\\ 0\end{array}\right)`$ and $`\left(\begin{array}{c}0\\ 1\end{array}\right)`$ with $`y𝕂^{}`$:
$`\left(\begin{array}{cc}0& l\\ l^t& 0\end{array}\right)\left(\begin{array}{c}y\\ 0\end{array}\right)`$ $`=\left(\begin{array}{c}l^tx\\ 0\end{array}\right)`$
$`\left(\begin{array}{cc}0& l\\ l^t& 0\end{array}\right)\left(\begin{array}{c}0\\ 1\end{array}\right)`$ $`=\left(\begin{array}{c}0\\ l\end{array}\right)`$
and we obtain (by a similar method) that multiplication by $`\left(\begin{array}{cc}0& l\\ l^t& 0\end{array}\right)`$ in $`𝕁(V)`$ is equivalent to commutation with $`\left(\begin{array}{cc}\frac{l}{2}& 0\\ 0& \frac{l}{2}\end{array}\right)`$ in $`H_2(𝕂)`$. Further $`𝔰𝔬(𝕂^{})`$ acts in $`𝕁(V)`$ precisely as it does in $`H_2(𝕂)`$. Thus we have
$$\mathrm{Der}H_2(𝕂)=A_2^{}(𝕂)\dot{+}𝔰𝔬(𝕂^{})$$
as required. ∎
The brackets in $`A_2^{}(𝕂_1)\dot{+}𝔰𝔬(𝕂_1)`$ are given by
$`[A,A^{}]`$ $`=AA^{}A^{}A`$
$`[S,A]`$ $`=S(A)`$
$`[S,S^{}]`$ $`=SS^{}S^{}S`$
with $`A,A^{}A_2^{}(𝕂_1)`$ and $`S,S^{}𝔰𝔬(𝕂_1)`$ and $`S(A)`$ describes $`S`$ acting elementwise on $`A`$. When they arise in calculations we consider multiples of the $`2\times 2`$ identity matrix $`I_2`$ to be elements of $`𝔰𝔬(𝕂_1)`$. Finally, the following lemma holds.
###### Lemma 8.
The Jacobi identity
$$[A,[B,H]]+[B,[H,A]]+[H,[A,B]]=0$$
holds for $`A,BA_2^{}(𝕂)`$ and $`XH_2(𝕂)`$.
###### Proof of Theorem 4.
Using Theorem 6 we can write $`L_2(𝕂_1,𝕂_2)`$ as
$$L_2(𝕂_1,𝕂_2)=A_2^{}(𝕂_1)\dot{+}𝔰𝔬(𝕂_1)\dot{+}H_2^{}(𝕂_1)𝕂_2^{}\dot{+}𝔰𝔬(𝕂_2^{}).$$
This can be considered to contain the following elements
$`J`$ $`=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)A_2^{}()`$
$`A_1`$ $`=\left(\begin{array}{cc}a_1& 0\\ 0& a_1\end{array}\right)A_2^{}(𝕂_1^{})`$
$`A_2`$ $`=\left(\begin{array}{cc}0& a_2\\ a_2& 0\end{array}\right)A_2^{}(𝕂_1^{})`$
(59) $`F`$ $`𝔰𝔬(𝕂_1^{})`$
$`S`$ $`𝔰𝔬(𝕂_2^{})`$
$`B_1`$ $`=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)b_1H_2^{}()𝕂_2^{}`$
$`B_2`$ $`=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)b_1H_2^{}()𝕂_2^{}`$
$`C`$ $`=\left(\begin{array}{cc}0& c_1\\ c_1& 0\end{array}\right)c_2H_2^{}(𝕂_1^{})𝕂_2^{}`$
with $`a_1,b_1,c_1𝕂_1^{}`$ and $`a_2,b_2,c_2𝕂_2^{}`$.
We define $`\phi :L_2(𝕂_1,𝕂_2)𝔰𝔬(\nu _1+\nu _2)`$ by
(60)
$$\begin{array}{c}\phi (J+A_1+A_2+F+S+B_1+B_2+C)=\hfill \\ \hfill \left(\begin{array}{cccc}S& 2Gb_1& 2Gb_2& 2Gc_2c_1^t\\ 2b_1^tG& 0& 2& 2a_2^t\\ 2b_2^tG& 2& 0& 2a_1^t\\ 2c_1c_2^tG& 2a_2& 2a_1& F\end{array}\right)\end{array}$$
where now elements of $`𝕂_i^{}`$ are identified with column vectors in $`^{\nu _i1}`$ and $`G`$ is the metric matrix for $`𝕂_2^{}`$. In the Euclidean case it is merely the identity matrix whereas in the non-Euclidean case it is the diagonal matrix consisting of $`(\frac{\nu _2}{2})`$ positive $`1`$’s and $`(\frac{\nu _2}{2}1)`$ negative $`1`$’s. The order of the positive and negative elements is determined by the choice of $`\pm 1`$ in the Cayley-Dickson calculation for $`𝕂_2`$. We show that $`\psi `$ is a Lie algebra isomorphism by calculating the multiplication tables for the Lie brackets between the elements listed in $`L_2(𝕂_1,𝕂_2)`$. We then calculate the equivalent brackets in $`𝔰𝔬(𝕂_1𝕂_2)`$ and show that they are equivalent. The relevant tables are found on the following two pages and are obtained simply by applying the stated Lie brackets for each algebra to these basis elements.
| | $`J`$ | $`A_1`$ | $`A_2`$ | $`B_1`$ | $`B_2`$ | $`C`$ | $`F`$ | $`S`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`J`$ | $`0`$ | $`2\left(\begin{array}{cc}0& a_1\\ a_1& 0\end{array}\right)`$ | $`2\left(\begin{array}{cc}a_2& 0\\ 0& a_2\end{array}\right)`$ | $`2\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)b_1`$ | $`2\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)b_2`$ | $`0`$ | $`0`$ | $`0`$ |
| $`A_1^{}`$ | $`2\left(\begin{array}{cc}0& a_1^{}\\ a_1^{}& 0\end{array}\right)`$ | $`2\mathrm{Im}\left(a_1a_1^{}\right)I`$ | $`2\mathrm{Re}\left(a_1^{}a_1\right)J`$ | $`0`$ | $`2\left(\begin{array}{cc}0& a_1^{}\\ a_1^{}& 0\end{array}\right)b_2`$ | $`2\mathrm{Re}\left(c_1a_1^{}\right)\times \left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)c_2`$ | $`F\left(A_1^{}\right)`$ | $`0`$ |
| $`A_2^{}`$ | $`2\left(\begin{array}{cc}a_2^{}& 0\\ 0& a_2^{}\end{array}\right)`$ | $`2\mathrm{Re}\left(a_1a_2^{}\right)J`$ | $`2\mathrm{Im}\left(a_2a_2^{}\right)I`$ | $`2\left(\begin{array}{cc}0& a_2^{}\\ a_2^{}& 0\end{array}\right)b_1`$ | $`0`$ | $`2\mathrm{Re}\left(a_2^{}c_1\right)\times \left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)c_2`$ | $`F(A_2^{}`$ | $`0`$ |
| $`B_1^{}`$ | $`2\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)b_1^{}`$ | $`0`$ | $`2\left(\begin{array}{cc}0& a_2\\ a_2& 0\end{array}\right)b_1^{}`$ | $`D_{b_1^{},b_1}`$ | $`2(b_1^{},b_2)J`$ | $`2(b_1^{},c_2)\times \left(\begin{array}{cc}0& c_1\\ c_1& 0\end{array}\right)`$ | $`0`$ | $`\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)S\left(b_1^{}\right)`$ |
| $`B_2^{}`$ | $`2\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)b_2^{}`$ | $`2\left(\begin{array}{c}0a_1\\ a_1& 0\end{array}\right)b_2^{}`$ | $`0`$ | $`2(b_1,b_2^{})J`$ | $`D_{b_2^{},b_2}`$ | $`2(b_2^{},c_2)\times \left(\begin{array}{cc}c_1& 0\\ 0& c_1\end{array}\right)`$ | $`0`$ | $`\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)S\left(b_2^{}\right)`$ |
| $`C^{}`$ | $`0`$ | $`2\mathrm{Re}\left(c_1^{}a_1\right)\times \left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)c_2^{}`$ | $`2\mathrm{Re}\left(a_2c_1^{}\right)\times \left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)c_2^{}`$ | $`2(b_1,c_2^{})\times \left(\begin{array}{cc}0& c_1^{}\\ c_1^{}& 0\end{array}\right)`$ | $`2(b_2,c_2^{})\times \left(\begin{array}{cc}c_1^{}& 0\\ 0& c_1^{}\end{array}\right)`$ | $`(c_1,c_1^{})D_{c_2^{},c_2}2(c_2,c_2^{})\times \mathrm{Im}\left(c_1^{}c_1\right)I`$ | $`F\left(\begin{array}{cc}0& c_1^{}\\ c_1^{}& 0\end{array}\right)c_2^{}`$ | $`\left(\begin{array}{cc}0& c_1^{}\\ c_1^{}& 0\end{array}\right)S\left(c_2^{}\right)`$ |
| $`F^{}`$ | $`0`$ | $`F^{}\left(A_1\right)`$ | $`F^{}\left(A_2\right)`$ | $`0`$ | $`0`$ | $`F^{}\left(\begin{array}{cc}0& c_1\\ c_1& 0\end{array}\right)c_2`$ | $`F^{}FFF^{}`$ | $`0`$ |
| $`S^{}`$ | $`0`$ | $`0`$ | $`0`$ | $`\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)S^{}\left(b_1\right)`$ | $`\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)S^{}\left(b_2\right)`$ | $`\left(\begin{array}{cc}0& c_1\\ c_1& 0\end{array}\right)S^{}\left(c_2\right)`$ | $`0`$ | $`S^{}SSS^{}`$ |
| | $`\psi \left(J\right)`$ | $`\psi \left(A_1\right)`$ | $`\psi \left(A_2\right)`$ | $`\psi \left(B_1\right)`$ | $`\psi \left(B_2\right)`$ | $`\psi \left(C\right)`$ | $`\psi \left(F\right)`$ | $`\psi \left(S\right)`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`\psi \left(J^{}\right)`$ | $`0`$ | $`4a_1^tG_{24}+4a_1G_{42}`$ | $`4a_2^tG_{34}+4a_2G_{43}`$ | $`4b_1GG_{13}+4Gb_1^tG_{31}`$ | $`4b_2GG_{12}4Gb_2^tG_{21}`$ | $`0`$ | $`0`$ | $`0`$ |
| $`\psi \left(A_1^{}\right)`$ | $`4a_1^tG_{24}4a_1^{}G_{42}`$ | $`\left(4a_1^{}a_1^t+4a_1a_1^t\right)G_{44}`$ | $`4a_2^ta_1^{}G_{23}+4a_1^ta_2G_{32}`$ | $`0`$ | $`4Gb_2a_1^tG_{14}4a_1^{}b_2^tGG_{41}`$ | $`4Gc_2c_1^ta_1^{}G_{13}+4a_1^tc_1c_2^tGG_{31}`$ | $`2a_1^tFG_{34}+2F^ta_1^{}G_{43}`$ | $`0`$ |
| $`\psi \left(A_2^{}\right)`$ | $`4a_2^tG_{34}4a_2^{}G_{43}`$ | $`4a_2^ta_1G_{23}4a_1^ta_2^{}G_{32}`$ | $`\left(4a_2^{}a_2^t+4a_2a_2^t\right)G_{44}`$ | $`4a_2^{}b_1^tGG_{41}4Gb_1a_2^tG_{14}`$ | $`0`$ | $`4Gc_2c_1^ta_2^{}G_{12}4a_2^tc_1c_2^tGG_{21}`$ | $`2a_2^tFG_{24}2Fa_2G_{42}`$ | $`0`$ |
| $`\psi \left(B_1^{}\right)`$ | $`4b_1^{}GG_{13}4b_1^tGG_{31}`$ | $`0`$ | $`4Gb_1^{}a_2^tG_{14}4a_2b_1^tGG_{41}`$ | $`\left(4b_1^{}b_1^t+4b_1b_1^t\right)G_{11}`$ | $`4b_2^tGGb_1^{}G_{32}4b_1^tGGb_2G_{23}`$ | $`4Gc_2c_1^tGb_1^{}G_{42}4b_1^tGGc_2c_1^tG_{24}`$ | $`0`$ | $`2b_1^tGS^tG_{21}2SGb_1^{}G_{12}`$ |
| $`\psi \left(B_2^{}\right)`$ | $`4b_2^tGG_{21}4Gb_2^{}G_{12}`$ | $`4a_1b_2^tGG_{41}4b_2^{}Ga_1^tG_{14}`$ | $`0`$ | $`4b_1^tGGb_2^{}G_{23}4b_2^tGGb_1G_{32}`$ | $`4\left(Gb_2b_2^tGGb_2^{}b_2^tG\right)G_{11}`$ | $`4c_1c_2^tGGb_2^{}G_{43}4b_2^tGGc_2c_1^tG_{34}`$ | $`0`$ | $`2b_2^tGS^tG_{31}2Sb_2^{}GG_{13}`$ |
| $`\psi \left(C^{}\right)`$ | $`0`$ | $`4Gc_2^{}c_1^ta_1G_{13}4a_1^tc_1^{}c_2^tGG_{31}`$ | $`4a_2^tc_1^{}c_2^tGG_{21}4Gc_2^{}c_1^ta_2G_{12}`$ | $`4b_1^tGGc_2^{}c_1^tG_{24}4Gc_2^{}c_1^tGb_1G_{42}`$ | $`4b_2^tGGc_2^{}c_1^tG_{34}4c_1^{}c_2^tGGb_2G_{43}`$ | $`4\left(Gc_2c_1^tc_1^{}c_2^tGGc_2^{}c_1^tc_1c_2^tG\right)G_{11}+4\left(c_1c_2^tGGc_2^{}c_1^tc_1^{}c_2^tGGc_2c_1^t\right)G_{44}`$ | $`2Gc_2^{}c_1^tFG_{14}2Fc_1^{}c_2^tGG_{41}`$ | $`2c_1^{}c_2^tGS^tG_{41}2SGc_2^{}c_1^tG_{14}`$ |
| $`\psi \left(F^{}\right)`$ | $`0`$ | $`2a_1^tF^{}G_{34}2F^ta_1G_{43}`$ | $`2F^ta_2G_{42}2a_2^tF^{}G_{24}`$ | $`0`$ | $`0`$ | $`2F^{}c_1c_2^tGG_{41}2Gc_2c_1^tF^{}G_{14}`$ | $`\left(F^{}FFF^{}\right)G_{44}`$ | $`0`$ |
| $`\psi \left(S^{}\right)`$ | $`0`$ | $`0`$ | $`0`$ | $`2S^{}Gb_1G_{12}2b_1^tGS^tG_{21}`$ | $`2S^{}Gb_2G_{13}2b_2^tGS^tG_{31}`$ | $`2S^{}Gc_2c_1^tG_{14}2c_1c_2^tGS^tG_{41}`$ | $`0`$ | $`\left(S^{}SSS^{}\right)G_{11}`$ |
Since Lie brackets are antisymmetric the tables show that there are 25 non-zero brackets to compare. We will use the following lemma.
###### Lemma 9.
If $`𝐚_1`$, $`𝐚_2^{\nu 1}`$ are the vector representations of the hypercomplex numbers $`a_1,a_2𝕂^{}`$ then
1. $`4(𝐚_1𝐚_2^t+𝐚_2𝐚_1^t)𝔰𝔬(𝕂_1\dot{+}𝕂_2)`$ is equivalent to $`2\mathrm{Im}(a_2a_1)IL_2(𝕂_1,𝕂_2)`$,
2. $`𝐚_2^t𝐚_1𝔰𝔬(𝕂_1\dot{+}𝕂_2)`$ is equivalent to $`(a_1,a_2)=\mathrm{Re}(a_1a_2)L_2(𝕂_1,𝕂_2)`$.
###### Proof.
We shall use the notation that if $`a=a_ie_i`$ is a hypercomplex number in $`𝕂^{}`$ then $`𝐚=(a_i)`$ is its vector representation as a column vector in $`^{\nu 1}`$ where $`\nu `$ is the dimension of $`𝕂`$.
1. The $`(i,j)`$th component of the matrix $`(\mathrm{𝐚𝐛}^t+\mathrm{𝐛𝐚}^t)`$ is the element $`a_ib_j+b_ia_j`$. If this multiplies a third hypercomplex number, say $`𝐜=(c_k)`$ then the $`n`$th element of the resulting vector will be
$$\underset{k=1}{\overset{\nu 1}{}}(a_nb_k+b_na_k)c_k.$$
Now we can write $`2\mathrm{Im}(ba)=2_ie_i(b_ja_kb_ka_j)`$ where $`(i,j,k)`$ is a cyclic permutation of a quaternionic triple. Commuting this with $`c`$ (which is equivalent to the action of $`2\mathrm{Im}(ba)I`$ on an element in a matrix $`H_2(𝕂_1)`$) gives
$$2\mathrm{Im}(ba)=2(b_ja_ka_jb_k)c_m(e_ie_me_me_i)$$
Now $`e_ie_m=e_me_i`$, thus
$$[2\mathrm{Im}(ba),c]=\underset{j}{}4((b_ja_kb_ka_j)c_me_j$$
where $`(e_i,e_m,e_j)`$ form a quaternionic triple. Therefore $`4(𝐚_1𝐚_2^t+𝐚_2𝐚_1^t)`$ is equivalent to $`2\mathrm{Im}(a_2a_1)I`$.
2. Now $`𝐛^t𝐚=_{i=1}^{\nu 1}b_ia_i`$. But
$`(a,b)`$ $`=\mathrm{Re}(a\overline{b})`$
$`={\displaystyle \underset{i=1}{\overset{\nu 1}{}}}a_ib_i(e_i)^2`$
$`={\displaystyle \underset{i=1}{\overset{\nu 1}{}}}a_ib_i`$
as required. Clearly $`\mathrm{Re}(ab)=(a,b)`$.
Using these results and the fact that $`G^2=I_{\nu 1}`$ the two tables are clearly equivalent and $`\psi `$ is a Lie algebra isomorphism. ∎
### 5.1. Proof of equation (56c)
We start by defining the structure and conformal algebras as they apply to the $`2\times 2`$ hermitian matrix case. We will see that the structure algebra is a subalgebra of the conformal algebra and the derivation algebra a subalgebra of the structure algebra. This means that our isomorphism for the conformal algebra includes the isomorphisms for the structure algebra, and trivially for the derivation algebra. From equations (7) and (57) we can deduce that
(61)
$$\mathrm{Str}^{}H_2(𝕂)=H_2^{}(𝕂)\dot{+}A_2^{}(𝕂)\dot{+}𝔰𝔬(𝕂^{})$$
which is a Lie algebra with brackets defined by the statement that $`\mathrm{Der}(H_2(𝕂))=A_2^{}(𝕂)\dot{+}𝔰𝔬(𝕂^{})`$ is a subalgebra. Denoting the elements of the algebra by
$`D`$ $`\mathrm{Der}(H_2(𝕂))`$
$`H`$ $`H_2^{}(𝕂)`$
we then have the brackets
$`[D,H]`$ $`=D(H)`$
$`[H,H^{}]`$ $`=[L_H,L_H^{}]`$
where we consider the derivation $`[L_H,L_H^{}]`$ to be an element of $`\mathrm{Der}𝕂`$. For the conformal algebra given by (8) we define the Lie brackets by taking $`\mathrm{Str}𝕁`$ as a subalgebra and the brackets as below. We take the elements of $`\mathrm{Con}H_2(𝕂)`$ to be
$`T`$ $`\mathrm{Str}𝕁`$
$`(X,Y)`$ $`[H_2(𝕂)]^2.`$
Then if $`RR^{}`$ is an involutive automorphism which is the identity on $`\mathrm{Der}𝕁`$ and multiplies elements of $`𝕁^2`$ by $`1`$ then the Lie brackets are
$`[T,(X,Y)]`$ $`=(TX,T^{}Y)`$
$`[(X,0),(Y,0)]=[(0,X),(0,Y)]`$ $`=0`$
$`[(X,0),(0,Y)]`$ $`=2L_{XY}+2[L_X,L_Y].`$
In the case of the conformal algebra for $`2\times 2`$ hermitian matrices we substitute $`H_2(𝕂)`$ for $`𝕁`$ in equation (8) to obtain
$$\mathrm{Con}H_2(𝕂)=\mathrm{Str}H_2(𝕂)\dot{+}[H_2(𝕂)]^2$$
which we can expand to
(62)
$$\mathrm{Con}H_2(𝕂)=H_2^{}(𝕂)\dot{+}A_2^{}(𝕂)\dot{+}𝔰𝔬(𝕂^{})\dot{+}\dot{+}[H_2(𝕂)]^2.$$
We can now define the Lie brackets explicitly for $`\mathrm{Con}H_2(𝕂)`$ by considering the following elements
$`D`$ $`\mathrm{Der}H_2(𝕂)`$
$`H`$ $`H_2^{}(𝕂)`$
$`r,r^{}`$ $``$
$`(X,Y)`$ $`[H_2(𝕂)]^2.`$
Then the brackets are
$`[D,r]=[r,H]`$ $`=[r,r^{}]=0`$
$`[D,H]`$ $`=D(H)`$
$`[D,(X,Y)]`$ $`=(D(X),D(Y))`$
$`[r,(X,Y)]`$ $`=(rX,rY)`$
$`[H,(X,Y)]`$ $`=(HX,HY)`$
with the brackets for $`(X,Y)`$ defined as above. We can also think of $`\mathrm{Str}H_2(𝕂)`$ and $`\mathrm{Con}H_2(𝕂)`$ in terms of $`2\times 2`$ matrices over $`H_2(𝕂)`$. Writing $`\mathrm{Str}H_2(𝕂)`$ and $`\mathrm{Con}H_2(𝕂)`$ in this way gives
$`\mathrm{Str}H_2(𝕂)`$ $`=\mathrm{Der}H_2(𝕂)\dot{+}\left\{\left(\begin{array}{cc}A& 0\\ 0& A\end{array}\right)AH_2(𝕂)\right\}`$
$`\mathrm{Con}H_2(𝕂)`$ $`=\mathrm{Der}H_2(𝕂)\dot{+}\left\{\left(\begin{array}{cc}A& B\\ C& A\end{array}\right)A,B,CH_2(𝕂)\right\}.`$
In this section we denote the elements of $`L_2(𝕂,\stackrel{~}{})`$ by
$`A`$ $`A_2^{}(𝕂)`$
$`S`$ $`𝔰𝔬(𝕂^{})`$
$`B\stackrel{~}{i},C\stackrel{~}{j},D\stackrel{~}{k}H_2^{}(𝕂)\stackrel{~}{}^{}`$
along with the basis elements of $`𝔰𝔬(\stackrel{~}{}^{})𝔰𝔬(2,1)`$ which we will call $`s_{12},s_{13}`$ and $`s_{23}`$, where
$`s_{12}`$ $`=\left(\begin{array}{ccc}0& 1& 0\\ 1& 0& 0\\ 0& 0& 0\end{array}\right)`$
$`s_{13}`$ $`=\left(\begin{array}{ccc}0& 0& 1\\ 0& 0& 0\\ 1& 0& 0\end{array}\right)`$
$`s_{23}`$ $`=\left(\begin{array}{ccc}0& 0& 0\\ 0& 0& 1\\ 0& 1& 0\end{array}\right).`$
Now we define an isomorphism $`\psi :L_2(𝕂,\stackrel{~}{})\mathrm{Con}H_2(𝕂)`$ by
$`\psi (A)`$ $`=A`$ $`\psi (S)`$ $`=S`$
$`\psi (B\stackrel{~}{i})`$ $`=B`$
$`\psi (C\stackrel{~}{j})`$ $`=\frac{1}{2}(C,C),`$ $`\psi (D\stackrel{~}{k})`$ $`=\frac{1}{2}(D,D)`$
$`\psi (s_{12})`$ $`=\frac{1}{2}(I,I),`$ $`\psi (s_{13})`$ $`=\frac{1}{2}(I,I)`$
$`\psi (s_{23})`$ $`=1.`$
We can also define this in terms of our $`4\times 4`$ matrices by
$`\psi (A)`$ $`=A`$ $`\psi (S)`$ $`=S`$
$`\psi (B\stackrel{~}{i})`$ $`=\left(\begin{array}{cc}B& 0\\ 0& B\end{array}\right)`$
$`\psi (C\stackrel{~}{j})`$ $`=\left(\begin{array}{cc}0& \frac{1}{2}C\\ 0& 0\end{array}\right),`$ $`\psi (D\stackrel{~}{k})`$ $`=\left(\begin{array}{cc}0& 0\\ \frac{1}{2}D& 0\end{array}\right)`$
$`\psi (s_{12})`$ $`=\left(\begin{array}{cc}0& 0\\ \frac{1}{2}I& 0\end{array}\right),`$ $`\psi (s_{13})`$ $`=\left(\begin{array}{cc}0& \frac{1}{2}I\\ 0& 0\end{array}\right)`$
$`\psi (s_{23})`$ $`=\left(\begin{array}{cc}I& 0\\ 0& I\end{array}\right).`$
The proof is a series of routine calculations of each product in both algebras showing that the isomorphism holds in all cases,which can be found on the next page.
Now $`L_2(𝕂,\stackrel{~}{})`$ is embedded in $`L_2(𝕂,\stackrel{~}{})`$ because $`\stackrel{~}{}`$ is embedded in $`\stackrel{~}{}`$ and $`\psi `$ maps $`L_2(𝕂,\stackrel{~}{})`$ to $`\mathrm{Str}^{}H_2(𝕂)`$. Thus if we define $`\psi _1=\psi L_2(𝕂,\stackrel{~}{})`$ then $`\psi _1`$ is an isomorphism between $`L_2(𝕂,\stackrel{~}{})`$ and $`\mathrm{Str}^{}H_2(𝕂)`$. Similarly we can define $`\psi _2=\psi L_2(𝕂,)`$ as an isomorphism between $`L_2(𝕂,)`$ and $`\mathrm{Der}H_2(𝕂)`$ by restricting $`\psi `$ to $`L_2(𝕂,)`$. Thus we have obtained proofs for all the $`2\times 2`$ isomorphisms.
| | $`A^{}`$ | $`D^{}`$ | $`B^{}\stackrel{~}{i}`$ | $`C^{}\stackrel{~}{j}`$ | $`E^{}\stackrel{~}{k}`$ | $`s_{12}`$ | $`s_{13}`$ | $`s_{23}`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`A`$ | $`AA^{}A^{}A`$ | $`D^{}\left(A\right)`$ | $`\left(AB^{}B^{}A\right)`$ | $`\left(AC^{}C^{}A\right)`$ | $`\left(AE^{}E^{}A\right)`$ | $`0`$ | $`0`$ | $`0`$ |
| | | | $`\stackrel{~}{i}`$ | $`\stackrel{~}{j}`$ | $`\stackrel{~}{k}`$ | | | |
| $`D`$ | $`D\left(A^{}\right)`$ | $`DD^{}D^{}D`$ | $`D\left(B^{}\right)\stackrel{~}{i}`$ | $`D\left(C^{}\right)\stackrel{~}{j}`$ | $`D\left(E^{}\right)\stackrel{~}{k}`$ | $`0`$ | $`0`$ | $`0`$ |
| $`B\stackrel{~}{i}`$ | $`\left(BA^{}A^{}B\right)`$ | $`D^{}\left(B\right)\stackrel{~}{i}`$ | $`[L_B,L_B^{}]`$ | $`\frac{1}{2}B,C^{}s_{12}+`$ | $`\frac{1}{2}B,E^{}s_{13}+`$ | $`B\stackrel{~}{j}`$ | $`B\stackrel{~}{k}`$ | $`0`$ |
| | $`\stackrel{~}{i}`$ | | | $`\left(BC^{}\right)\stackrel{~}{k}`$ | $`\left(BE^{}\right)\stackrel{~}{j}`$ | | | |
| $`C\stackrel{~}{j}`$ | $`\left(CA^{}A^{}C\right)`$ | $`D^{}\left(C\right)\stackrel{~}{j}`$ | $`\frac{1}{2}C,B^{}s_{12}`$ | $`[L_C,L_C^{}]`$ | $`\frac{1}{2}C,E^{}s_{23}`$ | $`C\stackrel{~}{i}`$ | $`0`$ | $`C\stackrel{~}{k}`$ |
| | $`\stackrel{~}{j}`$ | | $`\left(CB^{}\right)\stackrel{~}{k}`$ | | $`\left(CE^{}\right)\stackrel{~}{i}`$ | | | |
| $`E\stackrel{~}{k}`$ | $`\left(EA^{}A^{}E\right)`$ | $`D^{}\left(E\right)\stackrel{~}{k}`$ | $`\frac{1}{2}E,B^{}s_{13}`$ | $`\frac{1}{2}E,C^{}s_{23}+`$ | $`[L_E,L_E^{}]`$ | $`0`$ | $`E\stackrel{~}{i}`$ | $`E\stackrel{~}{j}`$ |
| | $`\stackrel{~}{k}`$ | | $`\left(EB^{}\right)\stackrel{~}{j}`$ | $`\left(EC^{}\right)\stackrel{~}{i}`$ | | | | |
| $`s_{12}`$ | $`0`$ | $`0`$ | $`B^{}\stackrel{~}{j}`$ | $`C^{}\stackrel{~}{i}`$ | $`0`$ | $`0`$ | $`s_{23}`$ | $`s_{13}`$ |
| $`s_{13}`$ | $`0`$ | $`0`$ | $`B^{}\stackrel{~}{k}`$ | $`0`$ | $`E^{}\stackrel{~}{i}`$ | $`s_{23}`$ | $`0`$ | $`s_{12}`$ |
| $`s_{23}`$ | $`0`$ | $`0`$ | $`0`$ | $`C^{}\stackrel{~}{k}`$ | $`E^{}\stackrel{~}{j}`$ | $`s_{13}`$ | $`s_{12}`$ | $`0`$ |
The multiplication table for $`L_2(𝕂,\stackrel{~}{})`$.
| | $`A^{}`$ | $`D^{}`$ | $`B^{}`$ | $`\frac{1}{2}(C^{},C^{})`$ | $`\frac{1}{2}(E^{},E^{})`$ | $`\frac{1}{2}(I,I)`$ | $`\frac{1}{2}(I,I)`$ | $`1`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`A`$ | $`AA^{}A^{}A`$ | $`D^{}\left(A\right)`$ | $`AB^{}B^{}A`$ | $`\frac{1}{2}(AC^{}C^{}A,`$ | $`\frac{1}{2}(AE^{}E^{}A,`$ | $`0`$ | $`0`$ | $`0`$ |
| | | | | $`AC^{}C^{}A)`$ | $`E^{}AAE^{})`$ | | | |
| $`D`$ | $`D\left(A^{}\right)`$ | $`DD^{}D^{}D`$ | $`D\left(B^{}\right)`$ | $`\frac{1}{2}(D\left(C^{}\right),D\left(C^{}\right))`$ | $`\frac{1}{2}(D\left(E^{}\right),D\left(E^{}\right))`$ | $`0`$ | $`0`$ | $`0`$ |
| $`B`$ | $`BA^{}A^{}B`$ | $`D^{}\left(B\right)`$ | $`[L_B,L_B^{}]`$ | $`\frac{1}{4}B,C^{}(I,I)+`$ | $`\frac{1}{4}B,E^{}(I,I)+`$ | $`\frac{1}{2}(B,B)`$ | $`\frac{1}{2}(B,B)`$ | $`0`$ |
| | | | | $`(BC^{},BC^{})`$ | $`(BE^{},BE^{})`$ | | | |
| $`\frac{1}{2}(C,C)`$ | $`\frac{1}{2}(CA^{}A^{}C,`$ | $`\frac{1}{2}(D^{}\left(C\right),D^{}\left(C\right))`$ | $`\frac{1}{4}C,B^{}(I,I)`$ | $`[L_C,L_C^{}]`$ | $`\frac{1}{2}C,E^{}1CE^{}`$ | $`C`$ | $`0`$ | $`\frac{1}{2}(C,C)`$ |
| | $`CA^{}A^{}C)`$ | | $`(CB^{},CB^{})`$ | | | | | |
| $`\frac{1}{2}(E,E)`$ | $`\frac{1}{2}(EA^{}A^{}E,`$ | $`\frac{1}{2}(D^{}\left(E\right),D^{}\left(E\right))`$ | $`\frac{1}{4}E,B^{}(I,I)`$ | $`\frac{1}{2}E,C^{}1+EC^{}`$ | $`[L_E,L_E^{}]`$ | $`0`$ | $`E`$ | $`\frac{1}{2}(E,E)`$ |
| | $`A^{}EEA^{})`$ | | $`(EB^{},EB^{})`$ | | | | | |
| $`\frac{1}{2}(I,I)`$ | $`0`$ | $`0`$ | $`\frac{1}{2}(B^{},B^{})`$ | $`C^{}`$ | $`0`$ | $`0`$ | $`1`$ | $`\frac{1}{2}(I,I)`$ |
| $`\frac{1}{2}(I,I)`$ | $`0`$ | $`0`$ | $`\frac{1}{2}(B^{},B^{})`$ | $`0`$ | $`E^{}`$ | $`1`$ | $`0`$ | $`\frac{1}{2}(I,I)`$ |
| $`1`$ | $`0`$ | $`0`$ | $`0`$ | $`\frac{1}{2}(C^{},C^{})`$ | $`\frac{1}{2}(E^{},E^{})`$ | $`\frac{1}{2}(I,I)`$ | $`\frac{1}{2}(I,I)`$ | $`0`$ |
The multiplication table for $`\mathrm{Con}H_2\left(𝕂\right)`$.
## 6. Magic squares of $`3\times 3`$ Matrix Algebras: Proofs
In this section we extend the results of the last section to the $`3\times 3`$ matrix case. We then develop these ideas by showing the maximal compact subalgebras for each of the exceptional Lie algebras that appear in the magic square.
We begin by showing that
(63)
$$L_3(𝕂,\stackrel{~}{})\mathrm{Con}H_3(𝕂)$$
and then show the maximal compact subalgebras are as stated previously.
### 6.1. Proof of equation (63)
In the case of $`3\times 3`$ matrices we know from (7) and Theorem 2 that
$$\mathrm{Str}^{}H_3(𝕂)=L_3^{}(𝕂)\dot{+}\mathrm{Der}𝕂$$
and also from (8),
$$\mathrm{Con}H_3(𝕂)=\mathrm{Str}H_3(𝕂)\dot{+}[H_3(𝕂)]^2.$$
Thus
$$\mathrm{Con}H_3(𝕂)=\mathrm{Der}𝕂\dot{+}A_3^{}(𝕂)\dot{+}H_3^{}(𝕂)\dot{+}\dot{+}[H_3(𝕂)]^2.$$
We have
$$L_3(𝕂,\stackrel{~}{})=\mathrm{Der}H_3(𝕂)\dot{+}H_3^{}(𝕂)\stackrel{~}{}^{}\dot{+}\mathrm{Der}\stackrel{~}{}.$$
It is known from that $`\mathrm{Der}=C(^{})`$ and can be shown similarly that $`\mathrm{Der}\stackrel{~}{}=C(\stackrel{~}{}^{})`$. Thus, using Theorem 2
$$L_3(𝕂,\stackrel{~}{})=\mathrm{Der}𝕂\dot{+}A_3^{}(𝕂)\dot{+}H_3^{}(𝕂)\stackrel{~}{}^{}\dot{+}C(\stackrel{~}{}^{}).$$
If we consider the elements of $`L_3(𝕂,\stackrel{~}{})`$
$`A`$ $`A_3^{}(𝕂)`$
$`D`$ $`\mathrm{Der}(𝕂)`$
$`B\stackrel{~}{i},C\stackrel{~}{j},E\stackrel{~}{k}`$ $`H_3^{}(𝕂)\stackrel{~}{}^{}`$
$`C_{\stackrel{~}{i}},C_{\stackrel{~}{j}},C_{\stackrel{~}{k}}`$ $`C(\stackrel{~}{}^{})`$
and the elements of $`\mathrm{Con}H_3(𝕂)`$
$`A`$ $`A_3^{}(𝕂)`$
$`B`$ $`H_3^{}(𝕂)`$
$`D`$ $`\mathrm{Der}(𝕂)`$
$`r`$ $``$
$`(X,Y)`$ $`[H_3(𝕂)]^2`$
then we can define an isomorphism $`\varphi :L_3(𝕂,\stackrel{~}{})\mathrm{Con}H_3(𝕂)`$ by :
$`\varphi (A)`$ $`=A`$ $`\varphi (D)`$ $`=D`$
$`\varphi (B\stackrel{~}{i})`$ $`=B`$
$`\varphi (C\stackrel{~}{j})`$ $`=\frac{1}{2}(C,C)`$ $`\varphi (E\stackrel{~}{k})`$ $`=\frac{1}{2}(E,E)`$
$`\varphi (C_{\stackrel{~}{k}})`$ $`=(I,I)`$ $`\varphi (C_{\stackrel{~}{j}})`$ $`=(I,I)`$
$`\varphi (C_{\stackrel{~}{i}})`$ $`=2.`$
Again we can express $`\mathrm{Con}H_3(𝕂)`$ and $`\mathrm{Str}^{}H_3(𝕂)`$ in terms of $`2\times 2`$ matrices over $`H_3(𝕂)`$ given by
$`\mathrm{Str}^{}H_3(𝕂)`$ $`=\mathrm{Der}H_3(𝕂)\dot{+}\left\{\left(\begin{array}{cc}A& 0\\ 0& A\end{array}\right)AH_3^{}(𝕂)\right\}`$
$`\mathrm{Con}H_3(𝕂)`$ $`=\mathrm{Der}H_3(𝕂)\dot{+}\left\{\left(\begin{array}{cc}A& B\\ C& A\end{array}\right)A,B,CH_3(𝕂)\right\}.`$
Further we can also define our isomorphism in terms of these matrices
$`\varphi (A)`$ $`=A`$ $`\varphi (D)`$ $`=D`$
$`\varphi (B\stackrel{~}{i})`$ $`=\left(\begin{array}{cc}B& 0\\ 0& B\end{array}\right)`$
$`\varphi (C\stackrel{~}{j})`$ $`=\left(\begin{array}{cc}0& \frac{1}{2}C\\ 0& 0\end{array}\right)`$ $`\varphi (E\stackrel{~}{k})`$ $`=\left(\begin{array}{cc}0& 0\\ \frac{1}{2}E& 0\end{array}\right)`$
$`\varphi (C_{\stackrel{~}{k}})`$ $`=\left(\begin{array}{cc}0& 0\\ I& 0\end{array}\right)`$ $`\varphi (C_{\stackrel{~}{j}})`$ $`=\left(\begin{array}{cc}0& I\\ 0& 0\end{array}\right)`$
$`\varphi (C_{\stackrel{~}{i}})`$ $`=\left(\begin{array}{cc}2I& 0\\ 0& 2I\end{array}\right).`$
We can again show this is a Lie algebra isomorphism by comparisons of the multiplication tables on the following page for $`L_3(𝕂,\stackrel{~}{})`$ and $`\mathrm{Con}H_3(𝕂)`$.
| | $`A^{}`$ | $`D^{}`$ | $`B^{}\stackrel{~}{i}`$ | $`C^{}\stackrel{~}{j}`$ | $`E^{}\stackrel{~}{k}`$ | $`C_{\stackrel{~}{k}}`$ | $`C_{\stackrel{~}{j}}`$ | $`C_{\stackrel{~}{i}}`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`A`$ | $`AA^{}A^{}A`$ | $`D^{}\left(A\right)`$ | $`\left(AB^{}B^{}A\right)\stackrel{~}{i}`$ | $`\left(AC^{}C^{}A\right)\stackrel{~}{j}`$ | $`\left(AE^{}E^{}A\right)\stackrel{~}{k}`$ | $`0`$ | $`0`$ | $`0`$ |
| $`D`$ | $`D\left(A^{}\right)`$ | $`DD^{}D^{}D`$ | $`D\left(B^{}\right)\stackrel{~}{i}`$ | $`D\left(C^{}\right)\stackrel{~}{j}`$ | $`D\left(E^{}\right)\stackrel{~}{k}`$ | $`0`$ | $`0`$ | $`0`$ |
| $`B\stackrel{~}{i}`$ | $`\left(BA^{}A^{}B\right)\stackrel{~}{i}`$ | $`D^{}\left(B\right)\stackrel{~}{i}`$ | $`[L_B,L_B^{}]`$ | $`\frac{1}{6}B,C^{}C_{\stackrel{~}{k}}+\left(BC^{}\right)\stackrel{~}{k}`$ | $`\frac{1}{6}B,E^{}C_{\stackrel{~}{j}}+\left(BE^{}\right)\stackrel{~}{j}`$ | $`2B\stackrel{~}{j}`$ | $`2B\stackrel{~}{k}`$ | $`0`$ |
| $`C\stackrel{~}{j}`$ | $`\left(CA^{}A^{}C\right)\stackrel{~}{j}`$ | $`D^{}\left(C\right)\stackrel{~}{j}`$ | $`\frac{1}{6}C,B^{}C_{\stackrel{~}{k}}\left(CB^{}\right)\stackrel{~}{k}`$ | $`[L_C,L_C^{}]`$ | $`\frac{1}{6}C,E^{}C_{\stackrel{~}{i}}\left(CE^{}\right)\stackrel{~}{i}`$ | $`2C\stackrel{~}{i}`$ | $`0`$ | $`2C\stackrel{~}{k}`$ |
| $`E\stackrel{~}{k}`$ | $`\left(EA^{}A^{}E\right)\stackrel{~}{k}`$ | $`D^{}\left(E\right)\stackrel{~}{k}`$ | $`\frac{1}{6}E,B^{}C_{\stackrel{~}{j}}\left(EB^{}\right)\stackrel{~}{j}`$ | $`\frac{1}{6}E,C^{}C_{\stackrel{~}{i}}+\left(EC^{}\right)\stackrel{~}{i}`$ | $`[L_E,L_E^{}]`$ | $`0`$ | $`2E\stackrel{~}{i}`$ | $`2E\stackrel{~}{j}`$ |
| $`C_{\stackrel{~}{k}}`$ | $`0`$ | $`0`$ | $`2B^{}\stackrel{~}{j}`$ | $`2C^{}\stackrel{~}{i}`$ | $`0`$ | $`0`$ | $`2C_{\stackrel{~}{i}}`$ | $`2C_{\stackrel{~}{j}}`$ |
| $`C_{\stackrel{~}{j}}`$ | $`0`$ | $`0`$ | $`2B^{}\stackrel{~}{k}`$ | $`0`$ | $`2E^{}\stackrel{~}{i}`$ | $`2C_{\stackrel{~}{i}}`$ | $`0`$ | $`2C_{\stackrel{~}{k}}`$ |
| $`C_{\stackrel{~}{i}}`$ | $`0`$ | $`0`$ | $`0`$ | $`2C^{}\stackrel{~}{k}`$ | $`2E^{}\stackrel{~}{j}`$ | $`2C_{\stackrel{~}{j}}`$ | $`2C_{\stackrel{~}{k}}`$ | $`0`$ |
The multiplication table for $`L_3(𝕂,\stackrel{~}{})`$.
| | $`A^{}`$ | $`D^{}`$ | $`B^{}`$ | $`\frac{1}{2}(C^{},C^{})`$ | $`\frac{1}{2}(E^{},E^{})`$ | $`(I,I)`$ | $`(I,I)`$ | $`2`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`A`$ | $`AA^{}A^{}A`$ | $`D^{}\left(A\right)`$ | $`AB^{}B^{}A`$ | $`\frac{1}{2}(AC^{}C^{}A,`$ | $`(AE^{}E^{}A,`$ | $`0`$ | $`0`$ | $`0`$ |
| | | | | $`AC^{}C^{}A)`$ | $`AE^{}+E^{}A)`$ | | | |
| $`D`$ | $`D\left(A^{}\right)`$ | $`DD^{}D^{}D`$ | $`D\left(B^{}\right)`$ | $`\frac{1}{2}(D\left(C^{}\right),D\left(C^{}\right))`$ | $`\frac{1}{2}(D\left(E^{}\right),D\left(E^{}\right))`$ | $`0`$ | $`0`$ | $`0`$ |
| $`B`$ | $`BA^{}A^{}B`$ | $`D^{}\left(B\right)`$ | $`[L_B,L_B^{}]B`$ | $`\frac{1}{6}B,C^{}(I,I)+`$ | $`\frac{1}{6}B,E^{}(I,I)+`$ | $`(B,B)`$ | $`(B,B)`$ | $`0`$ |
| | | | | $`(BC^{},BC^{})`$ | $`(BE^{},BE^{})`$ | | | |
| $`\frac{1}{2}(C,C)`$ | $`\frac{1}{2}(CA^{}A^{}C,`$ | $`\frac{1}{2}(D^{}\left(C\right),D^{}\left(C\right))`$ | $`\frac{1}{6}C,B^{}(I,I)`$ | $`[L_C,L_C^{}]`$ | $`\frac{1}{3}C,E^{}CE^{}`$ | $`2C`$ | $`0`$ | $`(C,C)`$ |
| | $`CA^{}A^{}C)`$ | | $`(CB^{},CB^{})`$ | | | | | |
| $`\frac{1}{2}(E,E)`$ | $`\frac{1}{2}(EA^{}A^{}E,`$ | $`\frac{1}{2}(D^{}\left(E\right),D^{}\left(E\right))`$ | $`\frac{1}{6}E,B^{}(I,I)`$ | $`\frac{1}{3}E,C^{}+EC^{}`$ | $`[L_E,L_E^{}]`$ | $`0`$ | $`2E`$ | $`(E,E)`$ |
| | $`EA^{}+A^{}E)`$ | | $`(EB^{},EB^{})`$ | | | | | |
| $`(I,I)`$ | $`0`$ | $`0`$ | $`(B^{},B^{})`$ | $`2C^{}`$ | $`0`$ | $`0`$ | $`4`$ | $`2(I,I)`$ |
| $`(I,I)`$ | $`0`$ | $`0`$ | $`(B^{},B^{})`$ | $`0`$ | $`2E^{}`$ | $`4`$ | $`0`$ | $`2(I,I)`$ |
| $`2`$ | $`0`$ | $`0`$ | $`0`$ | $`(C^{},C^{})`$ | $`(E^{},E^{})`$ | $`2(I,I)`$ | $`2(I,I)`$ | $`0`$ |
The multiplication table for $`\mathrm{Con}H_3\left(𝕂\right)`$.
Invoking the same method as used in the similar proof for $`2\times 2`$ matrices we note that we can define $`\varphi _1:L_3(𝕂,\stackrel{~}{})\mathrm{Str}^{}H_3(𝕂)`$ by:
$`\varphi _1(A)`$ $`=A,`$ $`\varphi _1(D)`$ $`=D`$
$`\varphi _1(B\stackrel{~}{i})`$ $`=B`$
and also, trivially, $`\varphi _2:L_3(𝕂,)\mathrm{Der}H_3(𝕂)`$ by :
$$\varphi _2(A)=A,\varphi _2(D)=D$$
Thus we have proved the following.
###### Theorem 7.
$`L_3(𝕂,)`$ $`\mathrm{Der}H_3(𝕂)`$
$`L_3(𝕂,\stackrel{~}{})`$ $`\mathrm{Str}^{}H_3(𝕂)`$
$`L_3(𝕂,\stackrel{~}{})`$ $`\mathrm{Con}H_3(𝕂)`$
hold for $`𝕂=,,`$ and $`𝕆`$.
The relation to the matrix algebras described in the Introduction in equation (1) is:
###### Theorem 8.
$`L_3(𝕂,)`$ $`𝔰𝔲(3,𝕂)`$
$`L_3(𝕂,\stackrel{~}{})`$ $`𝔰𝔩(3,𝕂)`$
$`L_3(𝕂,\stackrel{~}{})`$ $`𝔰𝔭(6,𝕂).`$
Thus we have the table
| | $``$ | $``$ | $``$ | $`𝕆`$ |
| --- | --- | --- | --- | --- |
| $`\mathrm{Der}H_3(𝕂)L_3(𝕂,)`$ | $`𝔰𝔲(3,)`$ | $`𝔰𝔲(3,)`$ | $`𝔰𝔲(3,)`$ | $`𝔰𝔲(3,𝕆)`$ |
| $`\mathrm{Str}^{}H_3(𝕂)L_3(𝕂,\stackrel{~}{})`$ | $`𝔰𝔩(3,)`$ | $`𝔰𝔩(3,)`$ | $`𝔰𝔩(3,)`$ | $`𝔰𝔩(3,𝕆)`$ |
| $`\mathrm{Con}H_3(𝕂)L_3(𝕂,\stackrel{~}{})`$ | $`𝔰𝔭(6,)`$ | $`𝔰𝔭(6,)`$ | $`𝔰𝔭(6,)`$ | $`𝔰𝔭(6,𝕆)`$ |
###### Proof.
1. We recall that $`𝔰𝔲(3,𝕂)`$ is the set of matrices with entries in $`𝕂`$ satisfying $`A^{}G+GA=0`$, i.e. the set of anti-hermitian matrices. In this case we have $`L_3(𝕂,)=A_3^{}(𝕂)\dot{+}\mathrm{Der}𝕂`$. Take $`AA_3^{}(𝕂)`$ and $`D\mathrm{Der}𝕂`$. Then $`\psi :L_3(𝕂,)𝔰𝔲(3,𝕂)`$ by
$$\psi (A+D)=A+DI$$
where $`I`$ is the $`3\times 3`$ matrix identity.
2. We define $`𝔰𝔩(3,𝕂)`$ to be the family of $`3\times 3`$ matrices with entries in $`𝕂`$ with the real part of the trace equal to zero. Also
$$L_3(𝕂,\stackrel{~}{})=A_3^{}(𝕂)\dot{+}\mathrm{Der}𝕂\dot{+}H_3^{}(𝕂)\stackrel{~}{}.$$
If we take $`AA_3^{}(𝕂)`$, $`D\mathrm{Der}𝕂`$ and $`HH_3^{}(𝕂)`$ (since we can regard the tensor product $`H_3^{}(𝕂)\stackrel{~}{}`$ as being one copy of $`H_3^{}(𝕂)`$) then the isomorphism $`\varphi :L_3(𝕂,\stackrel{~}{})𝔰𝔩(3,𝕂)`$ can be written
$$\varphi (A+D+H)=A+DI+H.$$
3. The Lie algebra $`𝔰𝔭(6,𝕂)`$ is defined to be the Lie algebra of $`6\times 6`$ matrices satisfying the equation $`A^{}J+JA=0`$ with the added condition that the trace is also zero. This can be written as the matrix $`\left(\begin{array}{cc}A& B\\ C& A^{}\end{array}\right)`$ where $`A`$, $`B`$ and $`C`$ are $`3\times 3`$ block matrices and $`B`$ and $`C`$ are hermitian. We have
$$L_3(𝕂,\stackrel{~}{})=A_3^{}()\dot{+}\mathrm{Der}\dot{+}H_3^{}()\stackrel{~}{}^{}\dot{+}\mathrm{Der}\stackrel{~}{}.$$
Taking $`AA_3^{}(𝕂)`$, $`D\mathrm{Der}𝕂`$, $`H_1\stackrel{~}{i}`$, $`H_2\stackrel{~}{j}`$ and $`H_3\stackrel{~}{k}H_3^{}(𝕂)\stackrel{~}{}`$ and $`r_1C_{\stackrel{~}{i}}+r_2C_{\stackrel{~}{j}}+r_3C_{\stackrel{~}{k}}\mathrm{Der}\stackrel{~}{}`$ then the isomorphism $`\chi :L_3(𝕂,\stackrel{~}{})𝔰𝔭(6,𝕂)`$ can be written explicitly as
$$\begin{array}{c}\chi (A+DI+H_1\stackrel{~}{i}+H_2\stackrel{~}{j}+H_3\stackrel{~}{k}+r_1C_{\stackrel{~}{i}}+r_2C_{\stackrel{~}{j}}+r_3C_{\stackrel{~}{k}})=\hfill \\ \hfill \left(\begin{array}{cc}A+DI+H_1+\frac{1}{3}r_1I& (H_2+r_2IH_3\frac{1}{3}r_3I)\\ (H_2+r_2I+H_3+\frac{1}{3}r_3I)& A+DIH_1\frac{1}{3}r_1I\end{array}\right).\end{array}$$
### 6.2. Maximal Compact Subalgebras
A semi-simple Lie algebra is called *compact* if it has a negative-definite killing form. It is called *non-compact* if its killing form is not negative-definite.
A non-compact real form, $`𝔤`$, of a semi-simple complex Lie algebra, $`L`$, has a *maximal compact subalgebra* $`𝔣`$ with an *orthogonal complementary subspace* $`𝔭`$ such that $`𝔤=𝔣\dot{+}𝔭`$ and the brackets
$`[𝔣,𝔣]`$ $`𝔣`$
$`[𝔣,𝔭]`$ $`𝔭`$
$`[𝔭,𝔭]`$ $`𝔣`$
$`(𝔣,𝔭)`$ $`=0`$
are satisfied. We denote by $`(,)`$ the killing form of $`L`$. There exists an involutive automorphism $`\sigma :𝔤𝔤`$ such that $`𝔣`$ and $`𝔭`$ are eigenspaces of $`\sigma `$ with eigenvalues $`+1`$ and $`1`$ respectively. A compact real form, $`𝔤^{}`$, of $`L`$ will also contain $`𝔣`$ as a compact subalgebra of $`𝔤^{}`$ but clearly in this case the maximal compact subalgebra will be $`𝔤^{}`$ itself. We can obtain $`𝔤^{}`$ from $`𝔤`$ by keeping the same brackets in $`[𝔣,𝔣]`$ and $`[𝔣,𝔭]`$ but multiplying the brackets in $`[𝔭,𝔭]`$ by $`1`$, i.e. by performing the *Weyl unitary trick* (putting $`𝔤^{}=𝔣\dot{+}i𝔭`$).
We will now give an overview of the method used to show that the algebras given in the table on page 3.1 are maximal compact, which is essentially the same in each case. We know that $`L_3(𝕂_1,𝕂_2)`$ gives a compact real form of each Lie algebra (from ). Thus if $`L_3(𝕂_1,\stackrel{~}{𝕂}_2)`$ shares a common subalgebra with $`L_3(𝕂_1,𝕂_2)`$, say $`𝔣`$, where
$`L_3(𝕂_1,𝕂_2)`$ $`=𝔣\dot{+}𝔭_1`$
$`L_3(𝕂_1,\stackrel{~}{𝕂}_2)`$ $`=𝔣\dot{+}𝔭_2,`$
and the brackets in $`[𝔣,𝔭_1]`$ are the same as those in $`[𝔣,𝔭_2]`$ but the brackets in $`[𝔭_1,𝔭_1]`$ are $`1`$ times the equivalent brackets in $`[𝔭_2,𝔭_2]`$, then $`𝔣`$ will be the maximal compact subalgebra of $`L_3(𝕂_1,\stackrel{~}{𝕂}_2)`$ and $`𝔭_2`$ will be its orthogonal complementary subspace. Moreover, because of the nature of the split composition algebras, this sign change in the brackets will reflect precisely the change in sign in the Cayley-Dickson process when moving from the non-split to the split form of the composition algebra.
We will briefly consider the nature of $`\mathrm{Der}𝕆`$ and $`\mathrm{Der}\stackrel{~}{𝕆}`$ since the structure of these algebras form a fundamental part of this proof. It is well known that $`\mathrm{Der}𝕆G_2`$ (see ). The derivation algebra of $`𝕆`$ is the Lie algebra of the automorphism group of $`𝕆`$. The derivations can be split into two types, those that are the infinitesimal versions of the automorphisms of $`𝕆`$ fixing the complex subspace $`=\dot{+}i`$ and those which are the infinitesimal versions of automorphisms fixing $``$. The derivations of the first type (leaving $`i`$ invariant) form a subalgebra isomorphic to $`𝔰𝔲(3)`$.
Further we can express the elements of $`\mathrm{Der}𝕆`$ in terms of pairs of basis elements $`s_{ij}`$ of $`𝔰𝔬(7)`$, as defined previously, since $`\mathrm{Der}𝕆𝔰𝔬(7)`$. Similarly $`\mathrm{Der}\stackrel{~}{𝕆}𝔰𝔬(4,3)`$, giving a representation of $`\mathrm{Der}\stackrel{~}{𝕆}`$ in terms of pairs of basis elements of $`𝔰𝔬(4,3)`$. It can be shown that the 14 elements of $`\mathrm{Der}𝕆`$ (by a method similar to that found in ) are
$`g_1`$ $`=s_{23}s_{45}`$ $`g_2`$ $`=s_{45}+s_{67}`$
$`g_3`$ $`=s_{25}s_{34}`$ $`g_4`$ $`=s_{27}s_{36}`$
$`g_5`$ $`=s_{47}s_{56}`$ $`g_6`$ $`=s_{24}s_{35}`$
$`g_7`$ $`=s_{26}+s_{37}`$ $`g_8`$ $`=s_{46}+s_{57}`$
$`g_9`$ $`=s_{12}+s_{47}`$ $`g_{10}`$ $`=s_{13}+s_{57}`$
$`g_{11}`$ $`=s_{14}+s_{27}`$ $`g_{12}`$ $`=s_{15}+s_{37}`$
$`g_{13}`$ $`=s_{16}+s_{25}`$ $`g_{14}`$ $`=s_{17}+s_{24}.`$
The derivations in $`\mathrm{Der}\stackrel{~}{𝕆}`$ are obtained directly from these by writing, for example
$$\stackrel{~}{g_1}=\stackrel{~}{s}_{23}\stackrel{~}{s}_{45},$$
where, if $`S_{ij}`$ is the matrix representation of $`s_{ij}`$ then $`G_2S_{ij}`$ is the matrix representation of $`\stackrel{~}{s_{ij}}`$ (recall that $`G_2`$ is the metric for $`𝕂_2`$). The other elements of $`\mathrm{Der}\stackrel{~}{𝕆}`$ are obtained by a similar method. Then $`\mathrm{Der}𝕆`$ and $`\mathrm{Der}\stackrel{~}{𝕆}`$ have a common subalgebra $`𝔰𝔬(3)\dot{+}𝔰𝔬(3)`$ (see, for example, ) which is the subalgebra with basis elements $`\{g_1,g_9,g_{10},g_2,g_5,g_8\}`$, these being invariant under multiplication by the metric $`G_2`$. We will denote by $`\stackrel{~}{G}_2`$ the form of the exceptional Lie algebra $`G_2`$ isomorphic to $`\mathrm{Der}\stackrel{~}{𝕆}`$.
We now state explicity the result we are about to prove.
###### Theorem 9.
The maximal compact subalgebras, stated explicity, are of the forms given in the table below.
| $`𝔤`$ | $`𝔣`$ | |
| --- | --- | --- |
| $`E_{6,1}`$ | $`F_4`$ | $`\mathrm{Der}H_3(𝕆)`$ |
| $`E_{7,1}`$ | $`E_6𝔰𝔬(2)`$ | $`\mathrm{Der}H_3(𝕆)\dot{+}H_3^{}(𝕆)\{i\}\dot{+}\{C_i\}`$ |
| $`E_{8,1}`$ | $`E_7𝔰𝔬(3)`$ | $`\mathrm{Der}H_3(𝕆)\dot{+}H_3^{}(𝕆)^{}\dot{+}M_{G_2}`$ |
| $`E_{6,2}`$ | $`𝔰𝔮(3)𝔰𝔬(3)`$ | $`\mathrm{Der}H_3()\dot{+}H_3^{}()^{}\dot{+}M_{G_2}`$ |
| $`E_{7,2}`$ | $`𝔰𝔲(6)𝔰𝔬(3)`$ | $`\mathrm{Der}H_3()\dot{+}H_3^{}()^{}\dot{+}M_{G_2}`$ |
.
where $`M_{G_2}(=\{g_1,g_9,g_{10},g_2,g_5,g_8\})`$, is the maximal compact subalgebra of $`\stackrel{~}{G_2}`$.
###### Proof.
We consider first each of the <sub>,1</sub> type algebras and then move on to the <sub>,2</sub> type. Denote by $`A^N`$ the non- compact form of the algebra $`A`$ and by $`A^C`$ the compact form of $`A`$.
1. For $`E_{6,1}^N`$
$`𝔣`$ $`=\mathrm{Der}H_3(𝕆)`$
$`𝔭`$ $`=H_3^{}(𝕆)\stackrel{~}{}^{}.`$
$`E_6^C`$ also has $`𝔣`$ as a subalgebra but in this case the remaining subspace is $`H_3^{}(𝕆)\stackrel{~}{}^{}`$. Thus there is only one set of brackets to check. If we consider $`H_1\stackrel{~}{i},H_2\stackrel{~}{i}H_3(𝕆)\stackrel{~}{}`$ and $`H_1i,H_2iH_3(𝕆)`$ then clearly, using the definitions for the brackets found on page 10
$`[H_1\stackrel{~}{i},H_2\stackrel{~}{i}]`$ $`=[L_{H_1},L_{H_2}]`$
$`[H_1i,H_2i]`$ $`=[L_{H_1},L_{H_2}],`$
as required.
2. In the case of $`E_{7,1}^N`$ the orthogonal complementary subspace is
$$\stackrel{~}{𝔭}=H_3^{}(𝕆)\{\stackrel{~}{j},\stackrel{~}{k}\}\dot{+}\{C_{\stackrel{~}{j}},C_{\stackrel{~}{k}}\},$$
where $`𝔣`$ is the maximal compact subalgebra given in the table above. Then $`𝔣`$ is also a subalgebra in $`E_7^C`$ and the remaining subspace is $`𝔭=H_3^{}(𝕆)\{j,k\}\dot{+}\{C_j,C_k\}`$. Now
$`[C_{\stackrel{~}{j}},C_{\stackrel{~}{k}}]`$ $`=2C_i`$ $`[C_j,C_k]`$ $`=2C_i`$
$`[C_{\stackrel{~}{j}},H_1\stackrel{~}{k}]`$ $`=H_12i`$ $`[C_j,H_1k]`$ $`=H_12i`$
$`[C_{\stackrel{~}{k}},H_1\stackrel{~}{j}]`$ $`=H_12i`$ $`[C_k,H_1j]`$ $`=H_12i.`$
Recall that
$$\begin{array}{c}[H_1x,H_2y]=2\mathrm{tr}(H_1H_2)D_{x,y}+\hfill \\ \hfill (HG)\mathrm{Im}(xy)Re(x\overline{y})[L_{H_1},L_{H_2}].\end{array}$$
We have to consider the two cases (1) when $`x=y`$ and (2) when $`xy`$. Case (1) is considered in $`E_6`$. Case (2) gives
$`[H_1\stackrel{~}{j},H_2\stackrel{~}{k}]`$ $`=2\mathrm{tr}(H_1H_2)C_i(H_1H_2)i`$
$`[H_1j,H_2k]`$ $`=2\mathrm{tr}(H_1H_2)C_i+(H_1H_2)i.`$
Clearly, $`𝔭`$ and $`\stackrel{~}{𝔭}`$ are orthogonal complementary subspaces with their brackets with themselves giving opposite signs and thus we can deduce that the choice of maximal compact subalgebra is correct.
3. In $`E_8^N`$ we have the orthogonal complementary subspace
$$\stackrel{~}{𝔭}=H_3^{}(𝕆)\{\stackrel{~}{l},\stackrel{~}{il},\stackrel{~}{jl},\stackrel{~}{kl}\}\dot{+}\{\stackrel{~}{g_a}a=3,4,6,7,11,12,13,14\}$$
to the maximal compact subalgebra $`𝔣`$ as shown in the previous table. Then $`𝔣`$ is also a subalgebra of $`E_8^C`$ and the remaining subspace in $`E_8^C`$ will be $`𝔭=H_3^{}(𝕆)\{l,il,jl,kl\}\dot{+}\{g_aa=3,4,6,7,11,12,13,14\}`$. For convenience we will label the orthogonal complementary subspaces of $`G_2`$ $`𝔭_{G_2}`$ and $`\stackrel{~}{𝔭}_{G_2}`$ for the compact and non-compact cases respectively. The calculations for $`E_8`$ are much the same as those for $`E_7`$. For brackets between $`H_3^{}(𝕆)\{l,il,jl,kl\}`$ and itself we again have two cases with $`x=y`$ and $`xy`$. These are resolved in the same way as before. To calculate the brackets between $`H_3^{}(𝕆)\{l,il,jl,kl\}`$ and $`𝔭_{G_2}`$ and between $`𝔭_{G_2}`$ and itself involves a set of long but relatively simple calculations involving $`s_{ij}`$ and $`\stackrel{~}{s_{ij}}`$. These produce the signs as expected, however, in the interest of the rainforests, we have not reproduced them here.
Now notice that these proofs do not in fact involve the matrices in $`H_3(𝕂)`$ since the change between compactness and non-compactness does not involve $`𝕂_1`$ but only $`𝕂_2`$. Thus since the orthogonal compact subspaces of $`E_{6,2}`$ and $`E_{7,2}`$ are
$`\stackrel{~}{𝔭_1}`$ $`=H_3^{}()\{\stackrel{~}{l},\stackrel{~}{il},\stackrel{~}{jl},\stackrel{~}{kl}\}\dot{+}\stackrel{~}{𝔭}_{G_2}`$
$`\stackrel{~}{𝔭_1}`$ $`=H_3^{}()\{\stackrel{~}{l},\stackrel{~}{il},\stackrel{~}{jl},\stackrel{~}{kl}\}\dot{+}\stackrel{~}{𝔭}_{G_2},`$
the proofs for these maximal compact subalgebras are contained within the that of $`E_8`$
Thus we have covered all of the subalgebras and our proof is complete. ∎ |
warning/0001/hep-ph0001091.html | ar5iv | text | # Breakdown of the KLN Theorem for Charged Particles in Condensed Matter
## 1 Introduction
The KLN theorem (Kinoshita 1962, Lee and Nauenberg 1964) refers to the removable nature of singularities in the production probabilities of particles in the limit of zero mass. For example, that the photon has zero mass in the vacuum led to the quantum electrodynamic “infrared catastrophe” wherein more and more photons were created at lower and lower frequencies. The catastrophe was cured with the realization that photon detectors have finite energy resolution. If the resolution was kept finite as a formal photon mass went to zero, no infinities appeared in the final results.
A generalization may be made for ultra-relativistic charged particles wherein the mass appears as a small parameter. For example, consider the weak decay
$$\mu ^{}e^{}+\overline{\nu }_e+\nu _\mu .$$
(1)
The vacuum radiatively corrected width (to lowest orders in $`G_F`$ and $`\alpha `$) is given by (Kinoshita T and Sirlin A 1959, Berman S 1958, Lee T D 1988)
$$\mathrm{\Gamma }_\mu =\mathrm{\Gamma }_\mu ^{(0)}\left\{1\frac{\alpha }{2\pi }\left(\pi ^2\frac{25}{4}\right)\left(\frac{8m^2}{M^2}\right)+\mathrm{}\right\},$$
(2)
where $`\mathrm{\Gamma }_\mu ^{(0)}=(Mc^2/192\pi ^3\mathrm{})(G_FM^2/\mathrm{}c)^2`$, $`M`$ is the muon mass, $`m`$ is electron mass, and $`\alpha =(e^2/\mathrm{}c)`$. Note, in the fictitious zero electron mass limit $`m0`$, there is no singularity in $`\mathrm{\Gamma }_\mu `$. In its general form, the KLN theorem protects the production probability (in the vacuum) from zero mass limit singularities.
However, when a muon decays from rest in condensed matter (as in laboratory measurements of $`\mathrm{\Gamma }_\mu `$) the vacuum KLN protection no longer holds true. This becomes evident from the retardation force $`F`$ in the condensed matter felt by an ultra-relativistic electron. It is well known theoretically (Berestetskii V B et. al. 1994) and experimentally (Caso C et. al. 1998) that
$$F\left(\frac{e^2\omega _p^2}{c^2}\right)\left\{ln\left(\frac{2E^2}{Imc^2}\right)1+\mathrm{}\right\},E>>mc^2,$$
(3)
where $`I`$ is the log mean ionization potential of the atoms, and the plasma frequency
$$\omega _p^2=\left(\frac{4\pi ne^2}{m}\right).$$
(4)
The finite number density $`n`$ of electrons in the material is responsible for decelerating the fast electron in accordance with Eq.(3), and by a similar process is responsible for stopping the muon in the first place. Clearly, the occurrence of the electron mass in the logarithm of Eq.(3) precludes a smooth limit $`m0`$ in the retardation force $`F`$ for ultra-relativistic particles. The KLN violation is present as long as the density of electrons in the condensed matter is finite. This is the first instance for which the KLN theorem for decays in condensed matter fails.
In more detail, the wave function $`\psi (x)`$ for a fast lepton of mass $`(\mathrm{}\kappa /c)`$ traveling through a condensed matter medium obeys a non-local Dirac equation of the form (Dyson F J 1949, Schwinger J 1951)
$$\left(i\gamma ^\mu _\mu +\kappa \right)\psi (x)+\mathrm{\Sigma }(xy)\psi (y)d^4y=0,$$
(5)
where the “self energy” $`\mathrm{\Sigma }(xy)`$ depends on the continuous medium and is certainly not well approximated by a local potential. The fact that condensed matter induces a non-locality in the self energy part is merely another description of the ability of condensed matter to stop a high energy particle. To lowest order in $`\alpha `$, the ultra-relativistic lepton has a self energy
$$\mathrm{\Sigma }(xy)=i\alpha \gamma ^\mu S(xy)\gamma ^\nu 𝒟_{\mu \nu }(xy)+\mathrm{},$$
(6)
where $`S(xy)`$ obeys the free lepton Dirac equation
$$\left(i\gamma ^\mu _\mu +\kappa \right)S(xy)=\delta (xy),$$
(7)
and the photon propagator in the condensed matter medium has the form
$$𝒟_{\mu \nu }(x)=D_{\mu \nu }(Q)e^{iQx}\frac{d^4Q}{(2\pi )^4}.$$
(8)
In Sec.2, we shall discuss the form of the photon propagator $`D_{\mu \nu }(Q)`$ in continuous media described by a dielectric response function $`\epsilon (\zeta )`$ for complex frequency $`\zeta `$. In Sec.3, it will be shown how the retardation force on an ultra-relativistic lepton can be related to the condensed matter induced non-local self energy part $`\mathrm{\Sigma }(xy)`$. This induced self energy renders the KLN protection in the limit $`\kappa 0`$ invalid. In the concluding Sec.4, we explore a notion of how KLN violating condensed matter effects can be made manifest in transition rates, e.g. $`\mathrm{\Gamma }_\mu `$.
## 2 Electromagnetic Fields in Continuous Media
The photon propagator in a material medium will be reviewed. All discussions of electromagnetic fields in materials should start with the Maxwell’s equations
$$_\mu ^{}F^{\mu \nu }=0,$$
(9)
$$_\mu F^{\mu \nu }=\left(\frac{4\pi }{c}\right)J^\nu .$$
(10)
The current may be broken up into a polarization current plus an external current
$$J^\nu =c_\mu P^{\mu \nu }+J_{ext}^\nu .$$
(11)
Eq.(11) reads as $`\rho =div𝐏+\rho _{ext}`$ and $`𝐉=c\mathrm{𝐜𝐮𝐫𝐥}𝐌+(𝐏/t)+𝐉_{ext}`$, in a specific Lorentz frame. If we now define
$$H^{\mu \nu }=F^{\mu \nu }+4\pi P^{\mu \nu },$$
(12)
then Eqs.(10)-(12) imply
$$_\mu H^{\mu \nu }=\left(\frac{4\pi }{c}\right)J_{ext}^\nu .$$
(13)
Just as $`F^{\mu \nu }`$ yields $`𝐄`$ and $`𝐁`$, we have $`H^{\mu \nu }`$ yielding $`𝐃=𝐄+4\pi 𝐏`$ and $`𝐇=𝐁4\pi 𝐌`$ via Eq.(12). Maxwell’s Eqs.(9) and (13) determine the electromagnetic fields in material media. However the electromagnetic properties of the medium must be included as constitutive equations (in linear order) to compute the photon propagator.
The constitutive equations are most easily expressed in the rest frame of the condensed matter. However, we here take the option of employing a strictly covariant notation. This may be done by considering the stress tensor $`T_{\mu \nu }`$ for the material. The stress tensor has four eigenvalues found from $`detT_\mu ^\nu P\delta _\mu ^\nu =0`$. Three of the eigenvalues $`(P_I,P_{II},P_{III})`$ correspond to possible values of the material pressure in three spatial eigenvector directions. The fourth eigenvalue is determined by the material mass density $`(P_{IV}/c^2)=\rho _{mass}`$ and is described by a time-like eigenvector
$$N_\mu N^\mu =1,$$
(14)
$$T_\mu ^\nu N_\nu =(\rho _{mass}c^2)N_\mu .$$
(15)
The dielectric response is a non-local (in space-time) linear operator $`\widehat{\epsilon }`$ describing
$$(N_\mu H^{\mu \nu })=\widehat{\epsilon }(N_\mu F^{\mu \nu }),$$
(16)
i.e. $`𝐃=\widehat{\epsilon }𝐄`$ in the rest frame of the material. We shall further assume that
$$N_\mu ^{}H^{\mu \nu }=N_\mu ^{}F^{\mu \nu }$$
(17)
i.e. $`𝐇=𝐁`$ in the rest frame of the material.
Thus far our description is strictly gauge invariant since we have not yet introduced vector potentials, and manifestly Lorentz covariant. However, in order to discuss the photon propagator one must choose a vector potential in a particular gauge. One satisfies Eq.(9) by introducing a vector potential $`A^\mu `$ via
$$F_{\mu \nu }=_\mu A_\nu _\nu A_\mu .$$
(18)
In the vacuum, it is often convenient to employ the covariant (Lorentz) gauge $`_\mu A_L^\mu =0`$. The close analogue to the covariant gauge is the material Lorentz gauge. In the condensed matter rest frame, with $`N^\mu =(\mathrm{𝟎},1)`$ and $`A^\mu =(𝐀,\varphi )`$, we fix the gauge according to
$$\frac{1}{c}\left(\frac{(\widehat{\epsilon }\varphi )}{t}\right)+div𝐀=0.$$
(19)
In this gauge Eqs.(13) and (16)-(18) read
$$\left\{\frac{\widehat{\epsilon }}{c^2}\left(\frac{}{t}\right)^2^2\right\}\varphi =\left(\frac{4\pi }{\widehat{\epsilon }}\right)\rho _{ext},$$
(20)
and
$$\left\{\frac{\widehat{\epsilon }}{c^2}\left(\frac{}{t}\right)^2^2\right\}𝐀=\left(\frac{4\pi }{c}\right)𝐉_{ext}.$$
(21)
From Eqs.(20) and (21), we find the material Lorentz gauge photon propagator in the space like directions
$$D_{ij}(𝐐,\omega )=\left(\frac{4\pi \delta _{ij}}{\left(|𝐐|^2\epsilon (|\omega |+i0^+)(\omega /c)^2i0^+\right)}\right),$$
(22)
and in the time-like direction
$$D_{00}(𝐐,\omega )=\left(\frac{4\pi }{\epsilon (|\omega |+i0^+)\left(|𝐐|^2\epsilon (|\omega |+i0^+)(\omega /c)^2i0^+\right)}\right).$$
(23)
The expressions in Eqs.(22) and (23) for the photon propagator in matter in the gauge of Eq.(19) are well known (Abrikosov A A et. al. 1963). These have been fully discussed for materials, along with the other standard gauges (e.g. radiation or temporal) with regard to the statistical physics of radiation in matter (Lifshitz E M and Pitaevskii L P 1984).
## 3 The Lepton Self Energy
In $`k`$-space, the self energy of Eqs.(6) and (7) has the form shown in Fig.1; It is
$$\mathrm{\Sigma }(k)=i\alpha \gamma ^\mu \left(\frac{1}{\kappa +\gamma (kQ)i0^+}\right)\gamma ^\nu D_{\mu \nu }(Q)\frac{d^4Q}{(2\pi )^4}.$$
(24)
If an ultra-relativistic charged lepton having four velocity $`v`$ is moving through the condensed matter and is described by a Dirac spinor $`u`$, then the transition rate per unit proper time $`\mathrm{\Gamma }_\gamma `$ to emit a photon in the medium is determined by the imaginary part of the self energy; i.e. $`\mathrm{\Gamma }_\gamma =2c\mathrm{}m\left(\overline{u}\mathrm{\Sigma }(k=\kappa v/c)u\right)`$. This amounts to the quasi-classical Dirac matrix replacement $`\gamma ^\mu (v^\mu /c)`$. Thus Eq.(24) yields a photon production rate
$$\mathrm{\Gamma }_\gamma =2\pi \alpha v^\mu \delta (vQ)v^\nu \mathrm{}m\left(D_{\mu \nu }(Q)\right)\frac{d^4Q}{(2\pi )^4}$$
(25)
per unit proper time of the ultra-relativistic particle.
It is of interest to compare our derivation of the photon production rate in Eq.(25), via $`\mathrm{\Sigma }(k)`$, to an earlier derivation (Schwinger J 1976) which starts from the non-local action for the current,
$$S=\left(\frac{1}{2c^3}\right)J^\mu (x)J^\nu (y)𝒟_{\mu \nu }(xy)d^4xd^4y.$$
(26)
If one starts from a “classical current” representing the ultra-relativistic lepton
$$J^\mu (x)=ec\delta (xv\tau )v^\mu 𝑑\tau ,$$
(27)
then radiated photons are created with a Poisson distribution $`p_n=(n^{N_\gamma }/n!)e^{N\gamma }`$. The mean number of radiated photons is $`N_\gamma `$. The probability of radiating zero photons $`p_0=|e^{iS/\mathrm{}}|^2=e^{N_\gamma }`$ yields $`N_\gamma =\mathrm{}m(2S/\mathrm{})`$; i.e.
$$N_\gamma =\mathrm{}m\left(\frac{1}{\mathrm{}c^3}J^\mu (x)J^\nu (y)𝒟_{\mu \nu }(xy)d^4xd^4y\right).$$
(28)
With $`L^\mu (Q,\tau )=_{(\tau /2)}^{(\tau /2)}\left(v^\mu e^{iQv\tau ^{}}\right)𝑑\tau ^{},`$ Eqs.(8), (27) and (28) yield
$$N_\gamma =\alpha \mathrm{}m\left(L^\mu (Q,\tau )L^\nu (Q,\tau )^{}D_{\mu \nu }(Q)\frac{d^4Q}{(2\pi )^4}\right).$$
(29)
Eq.(29) yields the same number of radiated photons per unit proper time
$$\mathrm{\Gamma }_\mu =\underset{\tau \mathrm{}}{lim}\left(\frac{N_\gamma }{\tau }\right)$$
(30)
as does Eq.(25), which was based on the self energy part $`\mathrm{\Sigma }(k)`$ in Eq.(24).
In the rest frame of the material, employing a laboratory time $`T`$,
$$\mathrm{\Gamma }_\mu ^{lab}=\underset{T\mathrm{}}{lim}\left(\frac{N_\gamma }{T}\right)=\mathrm{\Gamma }_\mu \sqrt{1\left(\frac{|𝐯|}{c}\right)^2},$$
(31)
it follows from Eqs.(22), (23), (25) and (30) that
$$\left(\frac{d\mathrm{\Gamma }_\mu ^{lab}}{d\omega }\right)=4\pi \alpha 𝒦(\omega ),\mathrm{restricted}\mathrm{to}(\omega >0),$$
(32)
$$𝒦(\omega )=2\mathrm{}m\left\{\frac{\delta (\omega 𝐐𝐯)\left((𝐯/c)^2\epsilon ^1(\omega +i0^+)\right)}{\left(|𝐐|^2\epsilon (\omega +i0^+)(\omega /c)^2i0^+\right)}\right\}\frac{d^3𝐐}{(2\pi )^3}.$$
(33)
To calculate the retardation force $`𝐅`$ one may compute the power loss according to
$$𝐅𝐯=_0^{\mathrm{}}\mathrm{}\omega \left(\frac{d\mathrm{\Gamma }_\mu ^{lab}}{d\omega }\right)𝑑\omega =$$
(34)
$$\left(\frac{2e^2}{\pi vc^2}\right)\mathrm{}m_0^{\mathrm{}}_0^{q_{max}}\left\{\frac{\left((v/c)^2\epsilon ^1\right)\omega qdqd\omega }{\left(q^2(\epsilon (c/v)^2)(\omega /c)^2i0^+\right)}\right\}.$$
(35)
In Eqs.(34) and (35) we have decomposed $`d^3𝐐=d^2𝐪dQ_{||}`$, where $`Q_{||}=(𝐐𝐯/|𝐯|)`$, and introduced a high wave number cutoff $`q_{max}`$.
After playing many ingenious games with logarithms, Eq.(35) in combination with the plasma frequency sum rule
$$\left(\frac{2}{\pi }\right)_0^{\mathrm{}}\omega \mathrm{}m\epsilon (\omega +i0^+)𝑑\omega =\omega _p^2,$$
(36)
yields a high energy retardation force formula in agreement with the standard result of Eq.(3),
$$F\left(\frac{e^2\omega _p^2}{c^2}\right)\left\{ln\left(\frac{2E^2}{Imc^2}\right)1+\mathrm{}\right\}\mathrm{as}(E/mc^2)\mathrm{}.$$
(37)
The fact that the logarithm in unavoidable in retardation force calculations in both theory and laboratory experiments clearly shows the failure of the KLN theorem in condensed matter. The theorem is (of course) valid and useful for particles moving through the vacuum (Sterman G and Weinberg S 1977). But the theorem cannot properly be applied to experiments wherein particles move through materials.
## 4 Conclusions
We have shown that the standard condensed matter retardation force formulas follow from the self energy part $`\mathrm{\Sigma }(k)`$ of a charged lepton. The lowest order in $`\alpha `$ correction is shown in Fig.1. Furthermore, the structure of the inclusive transition rate to excite the condensed matter, i.e. $`\left(2c\mathrm{}m\mathrm{\Sigma }(k)\right)`$, violates the required hypotheses of the KLN theorem. This violation has profound implications for ultra-high precision measurements of lepton lifetimes.
For example, consider the decays both with and without real photon radiation. If the photons are merely virtual, as in Figs.3, 4 and 5, then the reaction is in reality
$$\mu ^{}e^{}+\nu _\mu +\overline{\nu }_e.$$
(38)
From the interference of the superposition of amplitudes shown in Figs.3, 4 and 5, with the amplitude shown in Fig.2, one finds a decrease in the total muon decay rate. On the other hand there is a somewhat smaller increase in the total muon decay rate due the emission of real photons as shown in Figs.6 and 7. The real photon emission reaction is
$$\mu ^{}e^{}+\nu _\mu +\overline{\nu }_e+\gamma .$$
(39)
Both Eqs.(38) and (39) should include radiative corrections of order $`\alpha `$. The total effect to order $`\alpha `$ is to decrease the total decay rate of the muon.
In order to properly include initial and final state interactions, the self energy parts $`\mathrm{\Sigma }(k)`$ of both the muon and the electron, i.e. the diagrams both in Figs.3 and 4 must be calculated including condensed matter effects as fully discussed in this work. It makes no sense to ignore the simple experimental fact that charged particles may be stopped in condensed matter via the induced lepton self energy parts $`\mathrm{\Sigma }(k)`$.
Once this simple experimental fact is realized, it becomes evident (for reasons of gauge invariance) that all photons, virtual or real, must be renormalized by the condensed matter. Since all photons propagate as Eqs.(20) and (21), condensed matter effects must be included in the final state interactions of the photon. Thus, the condensed matter corrections up to order $`\alpha `$ shown become somewhat larger than the vacuum terms of order $`\alpha ^2`$.
The breakdown of the KLN theorem in condensed matter dictates that internal electromagnetic fields must be seriously considered in precision computations of experimental muon lifetimes. The condensed matter electric fields are sufficiently large to actually stop an energetic lepton. This would be most difficult to do in a vacuum. The requirement of considering condensed matter effects in precision determinations of the muon lifetime is most easily proved (or disproved) by experimental studies in which the muons are stopped in different sorts of materials.
## References |
warning/0001/hep-ex0001056.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In spite of its remarkable success in describing all electroweak data available today, the Standard Model (SM) leaves many questions unanswered. In particular, it explains neither the origin of the number of fermion generations nor the fermion mass spectrum. The precise measurements of the electroweak parameters at the Z pole have shown that the number of species of light neutrinos is three ; however, this does not exclude a fourth generation, or other massive fermions, if these particles have masses greater than half the mass of the Z-boson ($`M_\mathrm{Z}/2`$).
New fermions could be of the following types (for reviews see References ): sequential fermions, mirror fermions (with chirality opposite to that in the SM), vector fermions (with left- and right-handed doublets), and singlet fermions. These could be produced at high-energy $`\mathrm{e}^+\mathrm{e}^{}`$ colliders such as LEP, where two production mechanisms are possible: pair-production and single-production in association with a light standard fermion.
Lower limits on the masses of heavy leptons were obtained in $`\mathrm{e}^+\mathrm{e}^{}`$ collisions at centre-of-mass energies, $`\sqrt{s}`$, around $`M_\mathrm{Z}`$ , and recent searches at $`\sqrt{s}=`$ 130-140 GeV , $`\sqrt{s}=`$ 161 GeV , $`\sqrt{s}=`$ 172 GeV and $`\sqrt{s}=`$ 130-183 GeV have improved these limits. Excited leptons have been sought at $`\sqrt{s}M_\mathrm{Z}`$ , $`\sqrt{s}=`$ 130-140 GeV , $`\sqrt{s}=`$ 161 GeV , $`\sqrt{s}=`$ 172 GeV , $`\sqrt{s}=`$ 183 GeV , $`\sqrt{s}=`$ 189 GeV , and at the HERA ep collider . If direct production is kinematically forbidden, the cross-sections of processes such as $`\mathrm{e}^+\mathrm{e}^{}\gamma \gamma `$ and $`\mathrm{e}^+\mathrm{e}^{}\mathrm{f}\overline{\mathrm{f}}`$ are sensitive to new particles at higher masses.
This paper concentrates on the search conducted by OPAL in a wide range of topologies for the pair-production of new unstable heavy leptons and for both pair- and single-production of excited leptons of the known generations, using data collected in 1997 at $`\sqrt{s}=181184`$ GeV, with an average energy of 182.7 GeV. The integrated luminosity used depends on the final-state topologies, and is between 52 and 58 pb<sup>-1</sup>. The results are combined with those obtained earlier from 10 pb<sup>-1</sup> of data at $`\sqrt{s}=`$ 161.3 GeV and 10 pb<sup>-1</sup> at $`\sqrt{s}=`$ 172.1 GeV.
### 1.1 Heavy Leptons
Heavy neutral leptons are particularly interesting in the light of recent evidence for massive SM neutrinos . One method of generating neutrino mass is the see-saw mechanism , which predicts additional heavy neutral leptons. In this mechanism, if the mass of a heavy neutral lepton satisfies the relation $`m_\mathrm{N}=m_\mathrm{e}^2/m_{\nu _\mathrm{e}}`$, and if $`m_{\nu _\mathrm{e}}`$ is as massive as $`2.5`$ eV, then $`m_\mathrm{N}100`$ GeV, which is within the reach of LEP2.
In general, new heavy leptons N and $`\mathrm{L}^\pm `$ could in principle decay through the charged (CC) or neutral current (NC) channels:
$$\mathrm{N}\mathrm{}^\pm \mathrm{W}^{},\mathrm{N}\mathrm{L}^\pm \mathrm{W}^{},\mathrm{N}\nu _{\mathrm{}}\mathrm{Z},$$
$$\mathrm{L}^\pm \nu _{\mathrm{}}\mathrm{W}^\pm ,\mathrm{L}^\pm \mathrm{N}_\mathrm{L}\mathrm{W}^\pm ,\mathrm{L}^\pm \mathrm{}^\pm \mathrm{Z},$$
where $`\mathrm{N}_\mathrm{L}`$ is a stable or long-lived neutral heavy lepton, and $`\mathrm{}=\mathrm{e},\mu ,\mathrm{or}\tau `$. For heavy lepton masses less than the gauge boson masses, $`M_\mathrm{W}`$ or $`M_\mathrm{Z}`$, the vector bosons are virtual, leading to 3-body decay topologies. For masses greater than $`M_\mathrm{W}`$ or $`M_\mathrm{Z}`$, the decays are 2-body decays, and the CC and NC branching ratios can be comparable. For masses close to $`M_\mathrm{W}`$ or $`M_\mathrm{Z}`$, it is important to treat the transition from the 3-body to the 2-body decay properly, including effects from the vector boson widths. Expressions for the computation of partial decay widths with an off-shell W or Z boson can be found in .
The mixing of a heavy lepton with the standard lepton flavour is governed by a mixing angle $`\zeta `$. A mixing of 0.01 radians yields a decay length $`c\tau `$ of $`𝒪`$(1 nm). Since the decay length is proportional to $`1/\zeta ^2`$, looking for unstable heavy leptons which decay within the first cm, the analyses described in this paper are sensitive to $`\zeta ^2>𝒪`$(10<sup>-12</sup>). The presently existing upper limit on $`\zeta ^2`$ is approximately 0.005 radians<sup>2</sup> .
The searches for heavy leptons presented in this paper utilise only the case where N and $`\mathrm{L}^\pm `$ decay via the CC channel, as would be expected in a naive fourth generation extension to the SM. The NC channel does not contribute significantly in the heavy lepton searches due to kinematics. Searches for stable or long-lived charged heavy leptons, $`\mathrm{L}^\pm `$, are described in a separate paper .
### 1.2 Excited Leptons
Compositeness models attempt to explain the hierarchy of masses in the SM by the existence of a substructure within the fermions. Several of these models predict excited states of the known leptons. Excited leptons are assumed to have the same electroweak SU(2) and U(1) gauge couplings, $`g`$ and $`g^{}`$, to the vector bosons, but are expected to be grouped into both left- and right-handed weak isodoublets with vector couplings. The existence of the right-handed doublets is required to protect the ordinary light leptons from radiatively acquiring a large anomalous magnetic moment via the $`\mathrm{}^{}\mathrm{}V`$ interaction (where $`V`$ is a $`\gamma `$, Z, or W<sup>±</sup> and $`\mathrm{}^{}`$ refers in this case to both charged and neutral excited leptons).
In $`\mathrm{e}^+\mathrm{e}^{}`$ collisions, excited leptons could be produced in pairs via the process $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^{}\overline{\mathrm{}}^{}`$, or singly via the process $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^{}\overline{\mathrm{}}`$, as a result of the $`\mathrm{}^{}\mathrm{}V`$ couplings. Depending on the details of these couplings, excited leptons could be detected in the photonic, CC, or NC channels:
$$\nu _{\mathrm{}}^{}\nu _{\mathrm{}}\gamma ,\nu _{\mathrm{}}^{}\mathrm{}^\pm \mathrm{W}^{},\nu _{\mathrm{}}^{}\nu _{\mathrm{}}\mathrm{Z},$$
$$\mathrm{}^\pm \mathrm{}^\pm \gamma ,\mathrm{}^\pm \nu _{\mathrm{}}\mathrm{W}^\pm ,\mathrm{}^\pm \mathrm{}^\pm \mathrm{Z},$$
where $`\nu _{\mathrm{}}^{}`$ and $`\mathrm{}^\pm `$ are neutral and charged excited leptons, respectively.
The branching fractions of the excited leptons into the different vector bosons are determined by the strength of the three $`\mathrm{}^{}\mathrm{}V`$ couplings. We use the effective Lagrangian :
$$_{\mathrm{}\mathrm{}^{}}=\frac{1}{2\mathrm{\Lambda }}\overline{\mathrm{}}^{}\sigma ^{\mu \nu }\left[gf\frac{𝝉}{2}\text{W}_{\mu \nu }+g^{}f^{}\frac{Y}{2}B_{\mu \nu }\right]\mathrm{}_\mathrm{L}+\mathrm{hermitian}\mathrm{conjugate},$$
(1)
which describes the generalized magnetic de-excitation of the excited states. The matrix $`\sigma ^{\mu \nu }`$ is the covariant bilinear tensor, $`𝝉`$ are the Pauli matrices, $`𝐖_{\mu \nu }`$ and $`B_{\mu \nu }`$ represent the fully gauge-invariant field tensors, and $`Y`$ is the weak hypercharge. The parameter $`\mathrm{\Lambda }`$ has units of energy and can be regarded as the compositeness scale, while $`f`$ and $`f^{}`$ are the weights associated with the different gauge groups.
The relative values of $`f`$ and $`f^{}`$ also affect the size of the single-production cross-sections and their detection efficiencies. Depending on their relative values, either the photonic decay, the CC decay, or the NC decay will have the largest branching fraction, depending on the respective couplings :
$$f_\gamma =e_ff^{}+I_{3L}(ff^{}),f_W=\frac{f}{\sqrt{2}s_w},f_Z=\frac{4I_{3L}(c_w^2f+s_w^2f^{})4e_fs_w^2f^{}}{4s_wc_w}$$
where $`e_f`$ is the excited fermion charge, $`I_{3L}`$ is the weak isospin, and $`s_w(c_w)`$ are the sine (cosine) of the Weinberg angle $`\theta _w`$.
Our results will be interpreted using the two complementary coupling assignments, $`f=f^{}`$ and $`f=f^{}`$. For example, for the case $`f=f^{}`$, the photonic coupling to excited electrons is suppressed and the dominant production of excited electrons is via the $`s`$-channel. For $`ff^{}`$, the $`t`$-channel production of excited electrons dominates. In the case of excited neutrinos, if $`ff^{}`$, then the photonic coupling is allowed. In addition to the results for the two assignments $`f=f^{}`$ and $`f=f^{}`$, a new method is introduced which gives limits on excited leptons independent of the relative values of $`f`$ and $`f^{}`$.
## 2 Monte Carlo Simulation
The Monte Carlo (MC) generator EXOTIC has been used for the simulation of heavy lepton pair-production, $`\mathrm{e}^+\mathrm{e}^{}\mathrm{N}\overline{\mathrm{N}}`$ and $`\mathrm{e}^+\mathrm{e}^{}\mathrm{L}^+\mathrm{L}^{}`$, of excited lepton pair-production, $`\mathrm{e}^+\mathrm{e}^{}\nu _{\mathrm{}}^{}\overline{\nu }_{\mathrm{}}^{}`$ and $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^+\mathrm{}^{}`$, and of single excited lepton production, $`\mathrm{e}^+\mathrm{e}^{}\nu _{\mathrm{}}^{}\overline{\nu }_{\mathrm{}}`$ and $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^{}\mathrm{}`$. The code is based on formulae given in . The matrix elements include all spin correlations in the production and decay processes, and describe the transition from 3-body to 2-body decays of heavy fermions, involving virtual or real vector bosons, including the effects from vector boson widths. The JETSET package is used for the fragmentation and hadronization of quarks.
For $`\mathrm{N}\overline{\mathrm{N}}`$ and $`\mathrm{L}^+\mathrm{L}^{}`$ production, MC samples were generated for a set of masses from 40 to 90 GeV. Separate samples were generated for Dirac and Majorana heavy neutral leptons, taking into account the different angular distributions. For the case where $`\mathrm{L}^{}\mathrm{N}_\mathrm{L}\mathrm{W}^{}`$, samples were simulated at 25 points in the ($`M_\mathrm{L}`$,$`M_{\mathrm{N}_\mathrm{L}}`$) plane with $`M_\mathrm{L}`$ ranging from 50 to 90 GeV and $`M_{\mathrm{N}_\mathrm{L}}`$ from 40 to 87 GeV, and with a mass difference $`M_\mathrm{L}M_{\mathrm{N}_\mathrm{L}}`$ larger than 3 GeV. Excited lepton MC samples were generated for the pair-production channels with masses in the range from 40 to 90 GeV and for the single-production channels with masses in the range from 90 to 180 GeV.
A variety of MC generators was used to study the multihadronic background from SM processes: 4-fermion background processes were simulated using the generator grc4f , multihadronic background from 2-fermion final states was modelled using PYTHIA , while 2-photon processes were generated with PHOJET and HERWIG . To study the background from low-multiplicity events, the generators BHWIDE (large-angle Bhabha scattering) and TEEGG ($`t`$-channel Bhabha scattering) were used for the $`\mathrm{e}^+\mathrm{e}^{}\gamma (\gamma )`$ topology. The KORALZ generator was used for the $`\mu ^+\mu ^{}\gamma (\gamma )`$ and $`\tau ^+\tau ^{}\gamma (\gamma )`$ topologies. These generators include initial and final-state radiation, which is particularly important for the analyses with photons in the final states. Low-multiplicity 4-fermion final states produced in 2-photon interactions were modelled by using VERMASEREN .
Finally, SM processes with only photons in the final state are an important background to the analysis of excited neutral leptons with photonic decays. The RADCOR program was used to simulate the process $`\mathrm{e}^+\mathrm{e}^{}\gamma \gamma (\gamma )`$ and KORALZ was used to simulate the process $`\mathrm{e}^+\mathrm{e}^{}\nu \overline{\nu }\gamma (\gamma )`$.
All signal and background MC samples were processed through the full OPAL detector simulation and passed through the same analysis chain as the data.
## 3 Selection
The searches presented in this paper involve many different experimental topologies. Three classes of different analyses are used which rely on slightly different criteria for such details as track and cluster quality requirements and lepton identification methods. The first class of analyses includes the selections for high-multiplicity topologies, with hadronic jets in the final state from the hadronic CC decays of heavy leptons and the hadronic CC and NC decays of excited neutral and charged leptons. The second class includes selections for low-multiplicity topologies, and covers the photonic decays of excited charged leptons. The third class covers purely photonic event topologies arising from the photonic decays of excited neutral leptons.
### 3.1 High-Multiplicity Topologies
The searches for topologies with hadronic jets in the final state share a common high-multiplicity preselection. All charged tracks and calorimeter clusters are subjected to established quality criteria . Events are required to have at least 8 tracks and 15 clusters. The total visible energy measured in the detector, $`E_{\mathrm{vis}}`$, calculated from tracks and clusters , must be greater than 20 GeV.
Global event properties after this preselection are shown in Figure 1, which compares data and MC distributions. The visible energy is well described for $`E_{\mathrm{vis}}>75`$ GeV, where 2-fermion and 4-fermion processes dominate. For $`E_{\mathrm{vis}}<20`$ GeV, the data are not as well modelled. This region is dominated by events from 2-photon processes, which are not a significant background to the majority of analyses described in this paper. After a cut of $`E_{\mathrm{vis}}>75`$ GeV, the global event properties shown in Figure 1 (b-f) compare well between data and the SM expectation.
Figure 2 shows the energy distribution of identified electrons, muons, taus, and photons after the preselection. The lepton identification is similar to that described in , with modified isolation requirements for the high-multiplicity selections. Identified electrons and muons must have more than 1.5 GeV of visible energy, and taus more than 3 GeV. In addition, leptons have to be isolated within a cone of $`15^{}`$. The typical lepton identification efficiencies are 85-90% for electrons and muons and 70% for taus.
#### 3.1.1 Pair-Production of Heavy Leptons
##### $`𝐋^+𝐋^{}`$ Candidates
give rise to final states produced from flavour-mixing decays into light leptons via $`\mathrm{L}\nu _\mathrm{e}\mathrm{W}`$, $`\mathrm{L}\nu _\mu \mathrm{W}`$, and $`\mathrm{L}\nu _\tau \mathrm{W}`$. The resulting topologies, $`\nu \nu \mathrm{WW}`$, consist of the decay products of the two W-bosons along with missing transverse momentum. The main background to this selection is SM W-pair production. To optimize the sensitivity, selections for all high-multiplicity final states $`\nu \nu \mathrm{jjjj}`$, $`\nu \nu \mathrm{e}\nu \mathrm{jj}`$, and $`\nu \nu \mu \nu \mathrm{jj}`$ are performed, where “j” refers to a hadronic jet. The final state $`\nu \nu \tau \nu \mathrm{jj}`$ does not improve the sensitivity, and is not considered. In the selection for $`\nu \nu \mathrm{jjjj}`$ events, at least 12 GeV of missing momentum and at least 8 GeV of missing transverse momentum are required, and the missing momentum vector must not point along the beam direction ($`|\mathrm{cos}(\theta _{\mathrm{miss}})|<0.9`$). In addition, the events must satisfy higher track (10) and cluster multiplicity (35) requirements, and are vetoed if a charged lepton is identified. The number of events after applying these selections is shown in Table 1 for data and for the SM expectation. The selection efficiency, including the W branching ratio, is about 17-25%, depending on the heavy lepton mass. In the selection of $`\nu \nu \mathrm{e}\nu \mathrm{jj}`$ and $`\nu \nu \mu \nu \mathrm{jj}`$ events, an isolated lepton is required together with significant visible energy and missing transverse momentum. In this case, the selection efficiency, including the W branching ratio, is about 9-11%.
##### $`𝐍\overline{𝐍}`$ Candidates
give rise to final states produced through the flavour-mixing decay into a light charged lepton, via $`\mathrm{N}\mathrm{eW}`$, $`\mathrm{N}\mu \mathrm{W}`$, or $`\mathrm{N}\tau \mathrm{W}`$. The topologies are defined as $`\mathrm{}\mathrm{}\mathrm{WW}`$, where at least one W-boson decays hadronically and produces jets in the final state. At least two charged leptons of the same flavour are required and the jet resolution parameters have to be consistent with at least a 5-jet topology (isolated leptons are treated as “jets”). In order to optimize the sensitivity for the case $`\mathrm{N}\tau \mathrm{W}`$, the selection has been divided into $`\tau \tau \mathrm{jjjj}`$ topologies with fully-hadronic W-decays (2 charged leptons) and $`\tau \tau \mathrm{}\nu \mathrm{jj}`$ topologies with semileptonic W-decays, (3 charged leptons). The number of events after applying these selections is shown in Table 1. The signal efficiencies for a Dirac or Majorana lepton are about 50% for $`\mathrm{N}\mathrm{eW}`$, 57% for $`\mathrm{N}\mu \mathrm{W}`$, and 30-42% for $`\mathrm{N}\tau \mathrm{W}`$.
#### 3.1.2 Pair-Production of Long-Lived Heavy Neutral Leptons
##### $`𝐍_\mathrm{L}\overline{𝐍}_\mathrm{L}\mathrm{𝐖𝐖}`$ Candidates
originate from the process $`\mathrm{e}^+\mathrm{e}^{}\mathrm{L}^+\mathrm{L}^{}`$ with $`\mathrm{L}^{}\mathrm{N}_\mathrm{L}\mathrm{W}^{}`$, where $`\mathrm{N}_\mathrm{L}`$ is a stable or long-lived neutral heavy lepton which decays outside the detector. This production is possible if $`\mathrm{N}_\mathrm{L}`$ is a member of a fourth-generation SU(2) doublet which does not mix with the three known lepton generations and satisfies $`M_{\mathrm{L}^\pm }>M_{\mathrm{N}_\mathrm{L}}`$. This signal leads to very low visible energy if the mass difference, $`\mathrm{\Delta }MM_{\mathrm{L}^\pm }M_{\mathrm{N}_\mathrm{L}}`$, is small. Events are required to have a visible energy between 8 and 90 GeV, the missing momentum vector must be at least 20% of the visible energy and lie within the barrel region, a minimum transverse energy of 12 GeV is required, and the topology should correspond to a pair of acoplanar jets ($`>14^{}`$). The total number of candidates and expected background are shown in Table 1. The typical signal efficiencies are about 30-40% for $`\mathrm{\Delta }M>10`$ GeV, dropping to a few per-cent for $`\mathrm{\Delta }M=5`$ GeV. A total of 78 events is observed, compared to an expected background of 52.7 events. The discrepancy may be due to mismodelling of the background, which is dominated by 2-photon processes with an estimated systematic error on the background of 20%. These events typically have energy deposited in the forward region ($`|\mathrm{cos}(\theta )|>0.9`$), while from the signal MC one does not expect significant forward energy.
#### 3.1.3 Pair-Production of Excited Leptons with hadronic decays
Excited leptons, which could be pair-produced at LEP2, are expected to decay dominantly via CC or photonic interactions, depending on their coupling assignments. The final states with both de-excitations via CC decays, $`\mathrm{}^{}\nu \mathrm{W}`$ and $`\nu ^{}\mathrm{}\mathrm{W}`$, are similar to the decay topologies of heavy leptons, $`\mathrm{L}\nu \mathrm{W}`$ and $`\mathrm{N}\mathrm{}\mathrm{W}`$, so the same selections are applied, with the results shown in Table 1. The doubly-photonic decays give low-multiplicity topologies, and are discussed in Section 3.2.
#### 3.1.4 Single-Production of Excited Leptons with hadronic decays
##### Candidates for the processes $`\mathbf{}^\pm \mathbf{}^{}\mathbf{}\mathbf{}^\pm \mathbf{}^{}𝐙`$ and $`𝝂_{\mathbf{}}^{}{}_{}{}^{}𝝂_{\mathbf{}}\mathbf{}𝝂_{\mathbf{}}𝝂_{\mathbf{}}𝐙`$
followed by the hadronic decay of the Z-boson are selected by requiring two identified leptons of the same flavour and significant visible energy or significant missing transverse momentum ($`>`$25 GeV), respectively. In the latter case, events containing charged leptons are vetoed. Accepted events have to be consistent with an acoplanar 2-jet topology. For the charged lepton final states, a kinematic fit is performed requiring energy and momentum conservation, and the fit probability has to be consistent with a $`\mathrm{}^+\mathrm{}^{}\mathrm{jj}`$ final state. For the $`\tau \tau \mathrm{jj}`$ final state, to reduce the background further, additional cuts on the jet resolution parameter $`y_{34}`$, on the missing momentum vector, and on the ratio of fitted tau energy to visible tau energy ($`>1.15`$) are applied. The results for excited leptons are shown in Table 2 for data and for the SM expectation.
The selection efficiencies for excited leptons depend on the mass, and in the case of excited electrons also depend strongly on the coupling assignments for $`f`$ and $`f^{}`$ (see Section 1.2). For the case $`f=f^{}`$, the $`s`$-channel production of excited electrons is dominant, and the selection efficiency is typically 30-50%. For the case $`ff^{}`$, the $`t`$-channel production of excited electrons dominates, in which the scattered electron is preferentially scattered at low angles and is not detected. The selection efficiency for this case is typically 7-10%. For excited muons and taus only $`s`$-channel production is allowed and the selection efficiencies range from 40-50% and 3-25%, respectively. The selection efficiencies for excited neutrinos are typically 30-40%.
To improve the sensitivity in the excited electron search for the case $`ff^{}`$ a dedicated selection for the $`\mathrm{e}(\mathrm{e})\mathrm{jj}`$ channel has been designed, where the scattered electron is not observed in the detector. Events of this topology are selected by requiring exactly one electron to be identified. Tighter isolation cuts on the electron are applied and the event must have small missing transverse momentum ($`<25`$ GeV), in order to reject $`\mathrm{e}\nu \mathrm{jj}`$ final states from W-pair production. The event has to be consistent with a 3-jet topology, and events with high energy photons ($`>50`$ GeV) in the final state are rejected to reduce background from radiative returns to the Z. The fitted kinematics must be consistent with having an undetected electron in the beam direction opposite to the detected electron. The selection efficiency for the case $`ff^{}`$ is typically 20-40% and the numbers of observed and expected events are shown in Table 2.
##### Candidates for the processes $`\mathbf{}^\pm \mathbf{}^{}\mathbf{}\mathbf{}^\pm 𝝂_{\mathbf{}}`$W and $`𝝂_{\mathbf{}}𝝂_{\mathbf{}}^{}{}_{}{}^{}\mathbf{}\mathbf{}^\pm 𝝂_{\mathbf{}}`$W
followed by a hadronic decay of the W boson are selected by requiring an isolated lepton to be identified. The total energy of the event must be at least 30% of the centre-of-mass energy. For the case $`\mathrm{}=e`$ and $`\mu `$, the sum of the lepton energy and the missing transverse momentum must be at least 40% of the beam energy. Background from W-pair production is reduced by applying a kinematic fit to the $`\mathrm{jj}\mathrm{}\nu `$ system, requiring energy and momentum-conservation. If the resulting masses of the jet-jet and $`\mathrm{}\nu `$ systems are consistent with the W mass, the event is rejected. For the case $`\mathrm{}=\tau `$, it is assumed that the direction of the $`\tau `$ is given by the direction of the leading particle of the $`\tau `$ candidate. Further background suppression is obtained by requiring that the ratio of energy to mass of the dijet system be greater than 1.1. The number of observed events and the SM expectation are shown in Table 2. The selection efficiency for these channels, for $`f=f^{}`$, is typically from 20–30%.
For excited electron production in the case $`ff^{}`$, in which the scattered electron is not observed, a dedicated selection is applied. In this case, the total energy of the event must be at least 40% of the centre-of-mass energy, the missing transverse momentum must be at least 7.5% of the visible energy, and there must be no electron identified in the event. The event is forced into two jets, which are required to be acoplanar ($`>50^{}`$). The dijet mass $`M_{\mathrm{jj}}`$ must be consistent with the W mass ($`<100`$ GeV) and the ratio of energy to mass of the dijet system must exceed 1.1. The number of observed events and the SM expectation are shown in Table 2. The selection efficiency for this channel, for $`ff^{}`$ , ranges from 10–30%.
### 3.2 Low-Multiplicity Topologies
In this section the selection of photonic decays of singly-produced or pair-produced excited charged leptons is discussed. The final states consist of 2 like-flavour leptons and 1 or 2 photons. The lepton and photon identification and analysis techniques are described in Reference .
#### 3.2.1 Pair-Production of Excited Leptons with photonic decays
##### Candidates for the process $`\mathbf{}^{}\mathbf{}^{}\mathbf{}\mathbf{}^+\mathbf{}^{}𝜸𝜸`$
are selected by requiring the identification of two leptons with the same flavour and of two photons. These particles must carry at least 80% of the centre-of-mass energy in the case of $`\mathrm{e}^+\mathrm{e}^{}\gamma \gamma `$ and $`\mu ^+\mu ^{}\gamma \gamma `$, and between 40% and 95% of the centre-of-mass energy in the case of $`\tau ^+\tau ^{}\gamma \gamma `$. The background in the $`\mathrm{}^+\mathrm{}^{}\gamma \gamma `$ topology from Bhabha scattering and di-lepton production is reduced by requiring the leptons and photons to be isolated, and the background from the doubly-radiative return process $`\mathrm{e}^+\mathrm{e}^{}\mathrm{Z}\gamma \gamma \mathrm{}^+\mathrm{}^{}\gamma \gamma `$ is reduced by vetoing events with di-lepton masses close to the Z mass. The selection efficiencies are insensitive to the excited lepton mass and are about 54% for $`\mathrm{e}^+\mathrm{e}^{}`$, 61% for $`\mu ^+\mu ^{}`$, and 40% for $`\tau ^+\tau ^{}`$. The number of observed events and the SM expectation is shown in Table 1.
#### 3.2.2 Single-Production of Excited Leptons with photonic decays
##### Candidates for the process $`\mathbf{}^{}\mathbf{}\mathbf{}\mathbf{}^+\mathbf{}^{}𝜸`$
are selected with a technique identical to the $`\mathrm{}^+\mathrm{}^{}\gamma \gamma `$ selection, except that only one photon is required. Figure 3 shows the resulting $`\mathrm{}^\pm \gamma `$ invariant mass distributions.
To improve the efficiency for the excited electron search with $`ff^{}`$, giving rise to t-channel production, a dedicated search is performed for final states with a single electron and single photon visible in the detector, assuming the other electron is missing along the beam axis. Events with one electron and one photon carrying together at least 40% of the total centre-of-mass energy are selected. The photon is additionally required to have $`|\mathrm{cos}(\theta _{\mathrm{miss}})|<0.7`$, greatly suppressing the Bhabha scattering background.
The number of observed events and the SM expectation are shown in Table 2, and the lepton-photon invariant masses for the selected events are shown in Figure 3. No peak is observed. The selection efficiencies are typically about 70% for $`\mathrm{e}^{}\mathrm{e}`$ and $`\mu ^{}\mu `$, and about 40% for $`\tau ^{}\tau `$. The excess in the $`\tau ^{}\tau `$ search is consistent with a statistical fluctuation in the SM $`\tau ^+\tau ^{}\gamma `$ background.
### 3.3 Photonic final states
The search for singly- and pair-produced excited neutral leptons, $`\nu ^{}\nu \nu \nu \gamma `$ and $`\nu ^{}\nu ^{}\nu \nu \gamma \gamma `$, uses the OPAL search for photonic events with missing energy, described in . The numbers of selected events and expected backgrounds are listed in Tables 1 and 2 for pair- and single-production, respectively. For excited neutrinos in the mass range 70–180 GeV, the selection efficiencies are 70% and 10–70% for pair- and single-production, respectively.
## 4 Results
The numbers of expected signal events are evaluated from the production cross-sections, the integrated luminosity, and the estimated detection efficiencies of the various analyses.
The systematic errors on the number of expected signal and background events are estimated from: the statistical error of the MC estimates (1-10%), the error due to the interpolation used to infer the efficiency at arbitrary masses from a limited number of MC samples (2-15%), the error on the integrated luminosity (0.6%), the uncertainties in modelling the lepton identification cuts and in the photon conversion finder efficiency (2-8%), and the error due to uncertainties of the energy scale, energy resolution, and error parameterisations (1-4%). The errors are considered to be independent and are added in quadrature to give the total systematic error which is taken into account for the limit calculations.
In the case of pair-production searches, the production cross-section is relatively model-independent, and limits on the masses of the heavy or excited leptons can be obtained directly. For the flavour-mixing heavy lepton decays in which the pair-produced heavy leptons undergo CC decays into light leptons, 95% confidence level (CL) lower limits on the mass of the heavy lepton are shown in Table 3. The results are given for both Dirac and Majorana heavy neutral leptons. These results are valid for a mixing angle squared, $`\zeta ^2`$, greater than about 10<sup>-12</sup> radians<sup>2</sup>. For the decays of charged heavy leptons into a massless neutral lepton, $`\mathrm{L}\nu _{\mathrm{}}\mathrm{W}`$, the searches for the $`\nu \nu \mathrm{jjjj}`$ decay and for the $`\nu \nu \mathrm{jj}\mathrm{}\nu _{\mathrm{}}`$ have been combined. Masses smaller than 84.1 GeV are excluded at the 95% CL.
In the case that a heavy charged lepton $`\mathrm{L}^\pm `$ decays into a stable heavy neutral lepton, $`\mathrm{N}_\mathrm{L}`$, the exclusion region depends on both $`M_\mathrm{L}`$ and $`M_{\mathrm{N}_\mathrm{L}}`$. In order to optimize the sensitivity of the analysis, a scan is performed in steps of $`\mathrm{\Delta }M`$. At each point, the expected minimum and maximum visible energy, ($`\mathrm{E}_{\mathrm{min}},\mathrm{E}_{\mathrm{max}}`$), are calculated analytically, and the corresponding number of observed and expected events is determined. The resulting region in ($`M_\mathrm{L},M_\mathrm{N}`$) excluded at the 95% CL is given in Figure 4, together with the mean expected limit.
The mass limits on excited leptons are somewhat better than for the heavy lepton case, primarily due to the nature of the vector couplings which lead to larger production cross-sections . The mass limits inferred from the pair-production searches are shown in Table 4 for charged excited leptons and in Table 5 for neutral excited leptons. The first two sections of each table give the mass limits in which the dominant decay mode is assumed to be either via photons or W bosons.
In the single-production searches, the production cross-section depends on parameters within the model, so limits on those parameters, as a function of the new particle masses, are inferred instead. From the single production searches, 95% CL limits have been calculated on $`\sigma \times \mathrm{BR}`$ for $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^{}\overline{\mathrm{}},\mathrm{}^{}\mathrm{}V`$ for the different particle types, including photonic decay results from $`\sqrt{s}=`$ 161 GeV and 172 GeV, where the branching ratio, “BR”, depends on the relative values of $`f`$ and $`f^{}`$.
In general, the mass limits have been evaluated using a sliding window technique , taking into account the expected mass resolution for the $`\mathrm{e}\nu \mathrm{jj}`$, $`\mu \nu \mathrm{jj}`$, $`\tau \nu \mathrm{jj}`$ topologies (CC decays) and the $`\mathrm{ee}\gamma `$, $`\mu \mu \gamma `$, $`\tau \tau \gamma `$ topologies (photonic decays), or by taking into account the kinematically allowed mass limits for the $`\nu \nu \gamma `$ topologies (photonic decays). For the $`\mathrm{eejj}`$, $`\mu \mu \mathrm{jj}`$, and $`\tau \tau \mathrm{jj}`$ topologies (NC decays), a likelihood fit method has been used for the limit calculation . The resulting limits on $`\sigma \times \mathrm{BR}`$ are shown for excited charged leptons in Figure 5 and for excited neutral leptons in Figure 6 for all generations decaying via photonic, neutral current, and charged current processes. The $`\sigma \times \mathrm{BR}`$ limits do not depend on the coupling assignments except for excited electrons, where the selection efficiencies depend on the ratio of $`t`$-channel and $`s`$-channel contributions, and the results are shown for the example assignments $`f=\pm f^{}`$. The limits for the photonic decays are valid only for one of the two coupling assignments, $`f=+f^{}`$ for excited charged leptons and $`f=f^{}`$ for excited neutral leptons.
From the single-production searches, limits on the ratio of the coupling to the compositeness scale, $`f/\mathrm{\Lambda }`$, can be inferred. The results are shown in Figure 7 for two coupling assumptions<sup>1</sup><sup>1</sup>1 For the figures of coupling versus compositeness scale given in , an error was discovered in the cross-section formula used; the resulting limits on $`f/\mathrm{\Lambda }`$ were over-conservative for most values of the excited lepton mass. This error has been corrected in the present paper.. Since the branching ratio of the excited lepton decays via the different vector bosons is not known, examples of coupling assignments, $`f=\pm f^{}`$, are used to calculate these branching ratios and then the photonic, NC and CC decay results are combined for the limits.
A new method is used to infer limits on the coupling strength $`f_0/\mathrm{\Lambda }`$, independently of the relative values of $`f`$ and $`f^{}`$. Here $`f_0`$ is a generalized coupling constant defined as $`f_0=\sqrt{\frac{1}{2}(f^2+f_{}^{}{}_{}{}^{2})}`$. It is also useful to define the parameter $`\mathrm{tan}\varphi _f=f/f^{}`$; the previous coupling assignments then correspond to $`\varphi _f=\pi /4`$ ($`f=f^{}`$) and $`\varphi _f=\pi /4`$ ($`f=f^{}`$).
From the Lagrangian in Equation 1, the cross-section depends on $`f`$ and $`f^{}`$ in the following way:
$`\sigma ={\displaystyle \frac{a_1f^2+a_2ff^{}+a_3f_{}^{}{}_{}{}^{2}}{\mathrm{\Lambda }^2}}=\left({\displaystyle \frac{f_0}{\mathrm{\Lambda }}}\right)^2A(\varphi _f),`$ (2)
where $`A(\varphi _f)=2(a_1\mathrm{sin}^2\varphi _f+a_2\mathrm{sin}\varphi _f\mathrm{cos}\varphi _f+a_3\mathrm{cos}^2\varphi _f)`$. The coefficients $`a_1`$, $`a_2`$, and $`a_3`$ can be calculated from the matrix elements. The production cross-section and also the decay of the excited leptons depend on the coupling assumption and on the angle $`\varphi _f`$. The branching ratio is calculated using the formulae given in . By combining both the production cross-section and the branching ratio, the likelihood function $`L(N_s)`$, which is a function of the number of observed signal events $`N_s`$, can be translated into a likelihood function depending on the coupling strength $`\frac{f_0}{\mathrm{\Lambda }}`$:
$$L\left(\frac{f_0^2}{\mathrm{\Lambda }^2}\right)=L\left(\frac{N_s}{A(\varphi _f)BR(\varphi _f)ϵ(\varphi _f)}\right),$$
(3)
where $`ϵ(\varphi _f)`$ is the total selection efficiency and $``$ the integrated luminosity.
In the case of $`\mu ^{}`$, $`\tau ^{}`$, $`\nu _\mu ^{}`$, and $`\nu _\tau ^{}`$ production the efficiency is constant, while the case of the single production of excited $`\mathrm{e}^{}`$ and $`\nu _\mathrm{e}^{}`$ is more complicated. Due to the different $`\varphi _f`$ \- dependent contributions of the $`t`$-channel and the $`s`$-channel diagrams, the selection efficiency also depends on $`\varphi _f`$. Using a MC technique to determine the selection efficiency for arbitrary $`\varphi _f`$ would not be practical because a large number of MC events would have to be simulated for different excited lepton masses. It turns out, however, that the selection efficiency can be written as:
$`ϵ(\varphi _f)={\displaystyle \frac{\sigma _{sel}}{\sigma _{gen}}}={\displaystyle \frac{e_1f^2+e_2ff^{}+e_3f_{}^{}{}_{}{}^{2}}{a_1f^2+a_2ff^{}+a_3f_{}^{}{}_{}{}^{2}}}={\displaystyle \frac{N(\varphi _f)}{A(\varphi _f)}},`$ (4)
with $`N(\varphi _f)=2(e_1\mathrm{sin}^2\varphi _f+e_2\mathrm{sin}\varphi _f\mathrm{cos}\varphi _f+e_3\mathrm{cos}^2\varphi _f)`$. The coefficients $`e_i`$ can be calculated by evaluating the selection efficiencies from three MC samples generated with different values of $`\varphi _f`$.
The selection efficiencies for the different decay topologies and the cross-section vary strongly with $`\varphi _f`$ in the case of single $`\mathrm{e}^{}`$ production. In order to avoid numerical errors, one of the generated MC points has been chosen for the coupling assumption $`f=f^{}`$, where the large $`t`$-channel contribution vanishes and the cross-section is smallest. After having calculated the coefficients describing the selection efficiencies, the error of this method has been tested by comparing the calculated selection efficiency for a given value of $`\varphi _f`$ with the selection efficiency determined from a MC sample generated with the same $`\varphi _f`$ value. The error is found to be small compared to the statistical error of the MC samples.
Finally, the results from different decay channels are combined:
$$L_{\mathrm{}^{}}\left(\frac{f_0^2}{\mathrm{\Lambda }^2}\right)=L_\mathrm{}^{}\mathrm{}\gamma \left(\frac{f_0^2}{\mathrm{\Lambda }^2}\right)L_\mathrm{}^{}\mathrm{}\mathrm{Z}\left(\frac{f_0^2}{\mathrm{\Lambda }^2}\right)L_{\mathrm{}^{}\nu _{\mathrm{}}\mathrm{W}}\left(\frac{f_0^2}{\mathrm{\Lambda }^2}\right).$$
(5)
The resulting likelihood functions from different decay topologies are combined and upper limits on $`\frac{f_0^2}{\mathrm{\Lambda }^2}`$ are inferred as a function of the excited lepton mass by using the most conservative limit as a function of $`\varphi _f`$. The limit on $`\frac{f_0^2}{\mathrm{\Lambda }^2}`$ for any value of $`\varphi _f`$ is determined and shown in Figure 8. Using a similar technique, from the searches for the pair-production of excited leptons, the mass limits independent of the values of $`f`$ and $`f^{}`$ are given in the last section of Tables 4 and 5 for charged and neutral excited leptons, respectively. Together these represent the first limits on the compositeness scale $`f_0/\mathrm{\Lambda }`$ which do not depend on the relative values of $`f`$ and $`f^{}`$.
## 5 Conclusion
We have searched for the production of unstable heavy and excited leptons in a data sample corresponding to an integrated luminosity of $`58\mathrm{pb}^1`$ at a centre-of-mass energy of 181-184 GeV, collected with the OPAL detector at LEP. No evidence for their existence was found. From the search for the pair-production of heavy and excited leptons, lower mass limits were determined. From the search for the single-production of excited leptons, upper limits on $`\sigma \times \mathrm{BR}`$ of $`\mathrm{e}^+\mathrm{e}^{}\mathrm{}^{}\overline{\mathrm{}},\mathrm{}^{}\mathrm{}V`$ and upper limits on the ratio of the coupling to the compositeness scale were derived. These limits supersede the results in . Limits on the masses of excited leptons and on the compositeness scale $`f_0/\mathrm{\Lambda }`$ are established independent of the relative values of the coupling constants $`f`$ and $`f^{}`$.
Acknowledgements
We particularly wish to thank the SL Division for the efficient operation of the LEP accelerator at all energies and for their continuing close cooperation with our experimental group. We thank our colleagues from CEA, DAPNIA/SPP, CE-Saclay for their efforts over the years on the time-of-flight and trigger systems which we continue to use. In addition to the support staff at our own institutions we are pleased to acknowledge the
Department of Energy, USA,
National Science Foundation, USA,
Particle Physics and Astronomy Research Council, UK,
Natural Sciences and Engineering Research Council, Canada,
Israel Science Foundation, administered by the Israel Academy of Science and Humanities,
Minerva Gesellschaft,
Benoziyo Center for High Energy Physics,
Japanese Ministry of Education, Science and Culture (the Monbusho) and a grant under the Monbusho International Science Research Program,
Japanese Society for the Promotion of Science (JSPS),
German Israeli Bi-national Science Foundation (GIF),
Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie, Germany,
National Research Council of Canada,
Research Corporation, USA,
Hungarian Foundation for Scientific Research, OTKA T-029328, T023793 and OTKA F-023259. |
warning/0001/hep-th0001142.html | ar5iv | text | # References
Structure of Low-Energy Effective Action in N=4 Supersymmetric Yang-Mills Theories
I.L. Buchbinder
Department of Theoretical Physics
Tomsk State Pedagogical University
Tomsk, 634041, Russia
## Abstract
We study a problem of low-energy effective action in N=4 super Yang-Mills theories. Using harmonic superspace approach we consider N=4 SYM in terms of unconstrained N=2 superfield and apply N=2 background field method to finding effective action for N=4 SU(n) SYM broken down to U(n)<sup>n-1</sup>. General structure of leading low-energy corrections to effective action is discussed and calculational procedure for their explicit finding is presented.
1. Introduction
Low-energy structure of quantum supersymmetric field theories is described by the effective lagrangians of two types: chiral and general or holomorphic and non-holomorphic. Non-holomorphic or general contributions to effective action are given by integrals over full superspace while holomorphic or chiral contributions are given by integrals over chiral subspace of superspace. As a result, the effective action in supersymmetric theories should have the form
$$\mathrm{\Gamma }=(\underset{\text{chiral subspace}}{}+c.c.)+\underset{\text{full superspace}}{}+\mathrm{}$$
$`(1.1)`$
where the dots mean the terms in effective action depending on covariant derivatives of the superfields. The complex chiral superfield $``$ is called holomorphic or chiral effective potential and real superfield $``$ is called non-holomorphic or general effective potential. Thus, the notions of holomorphic and non-holomorphic effective potentials are very generic and characterizing in principle any superfield model. We point out that a possibility of holomorphic corrections to effective action was firstly demonstrated in refs \[1-3\] ( see also ) for N=1 SUSY and in refs for N=2 SUSY.
The modern interest to structure of low-energy effective action in extended supersymmetric theories was inspired by the seminal papers where exact instanton contribution to holomorphic effective potential has been found for N=2 SU(2) super Yang-Mills theory. The models with gauge groups SU(n) and SO(n) were considered in refs . One can show that in generic N=2 SUSY models namely the holomorphic effective potential is leading low-energy contribution. Non-holomorphic potential is next to leading correction. A detailed investigation of structure of low-energy effective action for various N=2 SUSY theories has been undertaken in refs \[10-17\].
A further study of quantum aspects of supersymmetric field models leads to problem of effective action in N=4 SUSY theories. These theories being maximally extended global supersymmetric models posses the remarkable properties on quantum level. We list only two of them:
* N=4 super Yang-Mills model is finite quantum field theory,
* N=4 super Yang-Mills model is superconformal invariant theory and hence, its effective action can not depend on any scale. These properties allow to analyze a general form of low-energy effective action and see that it changes drastically in compare with generic N=2 super Yang-Mills theories.
Analysis of structure of low-energy effective action in N=4 SU(2) SYM model spontaneously broken down to U(1) has been fulfilled in recent paper by Dine and Seiberg (see also for the other gauge groups). They have investigated a part of effective action depending on N=2 superfield strengths $`W`$, $`\overline{W}`$ and shown
* Holomorphic quantum corrections vanish identically in N=4 SYM. Therefore, namely non-holomorphic effective potential is leading low-energy contribution to effective action.
* Non-holomorphic effective potential $`(W,\overline{W})`$ can be found on the base of the properties of quantum N=4 SYM theory up to a coefficient. All perturbative or non-perturbative corrections do not influence on functional form of $`(W,\overline{W})`$ and concern only this coefficient.
The approaches to direct calculation of non-holomorphic effective potential including the above coefficient have been developed in refs \[18-21\], extensions for gauge group SU(n) spontaneously broken to maximal torus have been given in refs \[23-25\] ( see also where some bosonic contributions to low-energy effective action have been found).
This paper is a brief review of our approach to derivation of non-holomorphic effective potential in N=4 SYM theories.
2. N=4 super Yang-Mills theory in terms of unconstrained harmonic superfields
As well known, the most powerful and adequate approach to investigate the quantum aspects of supersymmetric field theories is formulation of these theories in terms of unconstrained superfields carrying out a representation of supersymmetry. Unfortunately such a manifestly N=4 supersymmetric formulation for N=4 Yang-Mills theory is still unknown. A purpose of this paper is study a structure of low-energy effective action for N=4 SYM as a functional of N=2 superfield strengths. In this case it is sufficient to realize the N=4 SYM theory as a theory of N=2 unconstrained superfields. It is naturally achieved within harmonic superspace. The N=2 harmonic superspace is the only manifestly N=2 supersymmetric formalism allowing to describe general N=2 supersymmetric field theories in terms of unconstrained N=2 superfields. This approach has been successfully applied to problem of effective action in various N=2 models in recent works \[12, 13, 15-17, 21, 24\]. We discuss here the results of the papers .
From point of view of N=2 SUSY, the N=4 Yang-Mills theory describes interaction of N=2 vector multiplet with hypermultiplet in adjoint representation. Within harmonic superspace approach, the vector multiplet is realized by unconstrained analytic gauge superfield $`V^{++}`$. As to hypermultiplet, it can be described either by areal unconstrained superfield $`\omega `$ ($`\omega `$-hypermultiplet) or by a complex unconstrained analytic superfield $`q^+`$ and its conjugate ($`q`$-hypermultiplet).
In the $`\omega `$-hypermultiplet realization, the classical action of N=4 SYM model has the form
$$S[V^{++},\omega ]=\frac{1}{2g^2}\mathrm{tr}d^4xd^4\theta W^2\frac{1}{2g^2}\mathrm{tr}𝑑\zeta ^{(4)}^{++}\omega ^{++}\omega $$
$`(2.1)`$
The first terms here is pure N=2 SYM action and the second term is action $`\omega `$-hypermultiplet. In $`q`$-hypermultiplet realization, the action of the N=4 SYM model looks like this
$$S[V^{++},q^+,\stackrel{+}{\stackrel{}{q}}]=\frac{1}{2g^2}\mathrm{tr}d^4xd^4\theta W^2\frac{1}{2g^2}\mathrm{tr}𝑑\zeta ^{(4)}q^{+i}^{++}q_i^+$$
$`(2.2)`$
where
$$q_i^+=(q^+,\stackrel{+}{\stackrel{}{q}}),q^{i+}=\epsilon ^{ij}q_j^+=(\stackrel{+}{\stackrel{}{q}},q^+)$$
$`(2.3)`$
All other denotions are given in ref .
Both models (2.1, 2.2) are classically equivalent and manifestly N=2 supersymmetric by construction. However, as has been shown in refs , both these models posses hidden N=2 supersymmetry and as a result they actually are N=4 supersymmetric.
3. General form of non-holomorphic effective potential.
We study the effective action $`\mathrm{\Gamma }`$ (1.1) for N=4 SYM with gauge group SU(2) spontaneously broken down to U(1). This effective action is considered as a functional of N=2 superfield strengths $`W`$ and $`\overline{W}`$. Then holomorphic effective potential $``$ depends on chiral superfield $`W`$ and it is integrated over chiral subspace of N=2 superspace with the measure $`d^4xd^4\theta `$. Non-holomorphic effective potential $``$ depends on both $`W`$ and $`\overline{W}`$. It is integrated over full N=2 superspace with the measure $`d^4xd^8\theta `$.
Let us begin with dimensional analysis of low-energy effective action. Taking into account the mass dimensions of $`W`$, $`(W)`$, $`(W,\overline{W})`$ and the superspace measures $`d^4xd^4\theta `$ and $`d^4xd^8\theta `$ ones write
$$(W)=W^2f\left(\frac{W}{\mathrm{\Lambda }}\right),(W,\overline{W})=(\frac{W}{\mathrm{\Lambda }},\frac{\overline{W}}{\mathrm{\Lambda }})$$
$`(3.1)`$
where $`\mathrm{\Lambda }`$ is some scale and $`f(\frac{W}{\mathrm{\Lambda }})`$ and $`(\frac{W}{\mathrm{\Lambda }},\frac{\overline{W}}{\mathrm{\Lambda }})`$ are the dimensionless functions of their arguments.
Due to remarkable properties of N=4 SYM in quantum domain, the effective action is scale independent. Therefore
$$\mathrm{\Lambda }\frac{d}{d\mathrm{\Lambda }}d^4xd^4\theta W^2f\left(\frac{W}{\mathrm{\Lambda }}\right)=0$$
$`(3.2)`$
$$\mathrm{\Lambda }\frac{d}{d\mathrm{\Lambda }}d^4xd^8\theta (\frac{W}{\mathrm{\Lambda }},\frac{\overline{W}}{\mathrm{\Lambda }})=0$$
$`(3.3)`$
Eq. (3.2) leads to $`f(\frac{W}{\mathrm{\Lambda }})=const`$. Eq (3.3) reads
$$\mathrm{\Lambda }\frac{d}{d\mathrm{\Lambda }}=g\left(\frac{W}{\mathrm{\Lambda }}\right)+\overline{g}\left(\frac{\overline{W}}{\mathrm{\Lambda }}\right)$$
$`(3.4)`$
Here $`g`$ is arbitrary chiral function of chiral superfield $`\frac{W}{\mathrm{\Lambda }}`$ and $`\overline{g}`$ is conjugate function. The integral of $`g`$ and $`\overline{g}`$ over full N=2 superspace vanishes and eq (3.3) takes place for any $`g`$ and $`\overline{g}`$.
Since $`f(\frac{W}{\mathrm{\Lambda }})=const`$ the holomorphic effective potential $`(W)`$ has classical form $`W^2`$. General solution of Eq (3.4) is written as follows
$$(\frac{W}{\mathrm{\Lambda }},\frac{\overline{W}}{\mathrm{\Lambda }})=c\mathrm{log}\frac{W^2}{\mathrm{\Lambda }^2}\mathrm{log}\frac{\overline{W}^2}{\mathrm{\Lambda }^2}$$
$`(3.5)`$
with arbitrary coefficient $`c`$. As a result, holomorphic effective potential is trivial in N=4 SYM theory. Therefore, namely non-holomorphic effective potential is leading low-energy quantum contribution to effective action. Moreover, the non-holomorphic effective potential is found exactly up to coefficient and given by eq. (3.5) . Any perturbative or non-perturbative quantum corrections are included into a single constant $`c`$.
However, this result immediately face the problems:
* Is there exist a calculational procedure allowing to derive $`(W/\mathrm{\Lambda },\overline{W}/\mathrm{\Lambda })`$ in form (3.5) within a model?
* What is value of $`c`$? If $`c=0`$, the non-holomorphic effective potential vanishes and low-energy effective action in N=4 SYM is defined the next terms in expansion of effective action (1.1) in derivatives.
* What is structure of non-holomorphic effective potential for the other then SU(2) gauge groups?
The answers all these questions have been given in refs \[19-25\]. Further we are going to discuss a general manifestly N=2 supersymmetric and gauge invariant procedure of deriving the non-holomorphic effective potential in one-loop approximation . This procedure is based on the following points
* Formulation of N=4 SYM theory in terms of N=2 unconstrained superfields in harmonic superspace .
* N=2 background field method providing manifest gauge invariance on all steps of calculations.
* Identical transformation of path integral for effective action over N=2 superfields to path integral over some N=1 superfields. This point is nothing more then replacement of variables in path integral.
* Superfield proper-time technique which is manifestly covariant method for calculating effective action in superfield theories.
Next section is devoted to some details of calculating non-holomorphic effective potential.
4. Calculaltion of non-holomorphic effective potential
We study effective action for the classically equivalent theories (2.1, 2.2) within N=2 background field method \[15-17\]. We assume also that the gauge group of these theories is SU(n). In accordance with background field method \[15-17\], the one-loop effective action in both realizations of N=4 SYM is given by
$$\mathrm{\Gamma }^{(1)}[V^{++}]=\frac{i}{2}\mathrm{Tr}_{(2,2)}\mathrm{log}\stackrel{}{\mathrm{}}\frac{i}{2}\mathrm{Tr}_{(4,0)}\mathrm{log}\stackrel{}{\mathrm{}}$$
$`(4.1)`$
where $`\stackrel{}{\mathrm{}}`$ is the analytic d’Alambertian introduced in ref .
$$\begin{array}{ccc}\hfill \stackrel{}{\mathrm{}}& =& 𝒟^m𝒟_m+\frac{i}{2}(𝒟^{+\alpha }W)𝒟_\alpha ^{}+\frac{i}{2}(\overline{𝒟}_{\dot{\alpha }}^+\overline{W})\overline{𝒟}^{\dot{\alpha }}\hfill \\ & & \frac{i}{4}(𝒟^{+\alpha }𝒟_\alpha ^+W)𝒟^{}+\frac{i}{8}[𝒟^{+\alpha },𝒟_\alpha ^{}]W+\frac{i}{2}\{\overline{W},W\}\hfill \end{array}$$
$`(4.2)`$
The formal definitions of the $`\mathrm{Tr}_{(2,2)}\mathrm{log}\stackrel{}{}`$ and $`\mathrm{Tr}_{(4,0)}\mathrm{log}\stackrel{}{}`$ are given in ref .
Our purpose is finding of non-holomorphic effective potential $`(W,\overline{W})`$ where the constant superfields $`W`$ and $`\overline{W}`$ belong to Cartan subalgebra of the gauge group SU(n). Therefore, for calculation of $`(W,\overline{W})`$ it is sufficient to consider on-shell background
$$𝒟^{\alpha (i}𝒟_\alpha ^{j)}W=0$$
$`(4.3)`$
In this case the one-loop effective action (4.1) can be written in the form
$$\mathrm{exp}(i\mathrm{\Gamma }^{(1)})=\frac{𝒟^{++}\mathrm{exp}\left\{\frac{i}{2}\mathrm{tr}𝑑\zeta ^{(4)}^{++}\stackrel{}{\mathrm{}}^{(++)}\right\}}{𝒟^{++}\mathrm{exp}\left\{\frac{i}{2}\mathrm{tr}𝑑\zeta ^{(4)}^{++}^{++}\right\}}$$
$`(4.4)`$
The superfield $`^{++}`$ belonging to the adjoint representation looks like $`^{++}=^{ij}u_i^+u_j^+`$ with $`u_i^+`$ be the harmonics and $`^{ij}=^{ji}`$ satisfying the constraints
$$𝒟_\alpha ^{(i}^{jk)}=\overline{𝒟}_{\dot{\alpha }}^{(i}^{jk)}=0,\overline{}^{ij}=_{ij}$$
$`(4.5)`$
The next step is transformation of the path integral (4.4) to one over unconstrained N=1 superfields. This point is treated as replacement of variables in path integral (4.4). We introduce the N=1 projections of $`W`$ ( see the details in refs ). As a result one obtains
$$\mathrm{\Gamma }^{(1)}=\underset{k<l}{}\mathrm{\Gamma }_{kl},\mathrm{\Gamma }_{kl}=i\mathrm{Tr}\mathrm{log}\mathrm{\Delta }_{kl}$$
$`(4.6)`$
where
$$\mathrm{\Delta }_{kl}=𝒟^m𝒟_m(W^{k\alpha }W^{l\alpha })𝒟_\alpha +(\overline{W}_{\dot{\alpha }}^k\overline{W}_{\dot{\alpha }}^l)\overline{𝒟}^{\dot{\alpha }}+|\mathrm{\Phi }^k\mathrm{\Phi }^l|^2$$
$`(4.7)`$
and $`𝒟_m`$, $`𝒟_\alpha `$, $`\overline{𝒟}_{\dot{\alpha }}`$ are the supercovariant derivatives. Here
$$\mathrm{\Phi }=\mathrm{diag}(\mathrm{\Phi }^1,\mathrm{\Phi }^2,\mathrm{},\mathrm{\Phi }^n),\underset{k=1}{\overset{n}{}}\mathrm{\Phi }^k=0.$$
$`(4.8)`$
$$W_\alpha =\mathrm{diag}(W_\alpha ^1,\mathrm{},W_\alpha ^n),\underset{k=1}{\overset{n}{}}W_\alpha ^k=0$$
The operator (4.7) has been introduced in ref .
Thus, we get a problem of effective action associated with N=1 operator (4.7). Such a problem can be investigated within N=1 superfield proper-time technique . Application of this technique leads to lowest contribution to effective action in the form
$$\mathrm{\Gamma }_{kl}=\frac{1}{(4\pi )^2}d^8z\frac{W^{\alpha kl}W_\alpha ^{kl}\overline{W}_{\dot{\alpha }}^{kl}\overline{W}^{\dot{\alpha }kl}}{(\mathrm{\Phi }^{kl})^2(\overline{\mathrm{\Phi }}^{kl})^2}$$
$`(4.9)`$
where
$$\mathrm{\Phi }^{kl}=\mathrm{\Phi }^k\mathrm{\Phi }^l,W^{kl}=W^kW^l$$
$`(4.10)`$
Eqs (4.6, 4.9, 4.10) define the non-holomorphic effective potential of N=4 SYM theory in terms of N=1 projections of N=2 superfield strengths.
The last step is restoration of N=2 form of effective action (4.9). For this purpose we write contribution of non-holomorphic effective potential to effective action in terms of covariantly constant N=1 projections $`\mathrm{\Phi }`$ and, $`W_\alpha `$
$$d^4xd^8\theta (\overline{W},W)=d^8zW^\alpha W_\alpha \overline{W}_{\dot{\alpha }}\overline{W}^{\dot{\alpha }}\frac{^4(\overline{\mathrm{\Phi }},\mathrm{\Phi })}{\mathrm{\Phi }^2\overline{\mathrm{\Phi }}^2}+\mathrm{derivatives}$$
$`(4.11)`$
Comparison of eqs (4.6, 4.9) and (4.11) leads to
$$\mathrm{\Gamma }^{(1)}=d^4xd^8\theta (\overline{W},W)$$
$$(\overline{W},W)=\frac{1}{(8\pi )^2}\underset{k<l}{}\mathrm{log}\left(\frac{\overline{W}^k\overline{W}^l}{\mathrm{\Lambda }}\right)^2\mathrm{log}\left(\frac{W^kW^l}{\mathrm{\Lambda }}\right)^2$$
$`(4.12)`$
Eq (4.12) is our final result.
5. Discussion
Eq. (4.12) defines the non-holomorphic effective potential depending on N=2 superfield strengths for N=4 SU(n) super Yang-Mills theories. As a result we answered all the questions formulated in section 3. First, we have presented the calcualtional procedure allowing to find non-holomorphic effective potential. Second, we calculated the coefficient c in eq. (3.5) for SU(2) group. It is equal to $`1/(8\pi )^2`$. Third, a structure of non-holomorphic effective potential for the gauge group SU(n) has been established.
It is interesting to point out that the scale $`\mathrm{\Lambda }`$ is absent when the non-holomorphic effective potential (4.12) is written in terms of N=1 projections of $`W`$ and $`\overline{W}`$ (see eqs (4.6, 4.9)). Therefore, the $`\mathrm{\Lambda }`$ will be also absent if we write the non-holomorphic effective potential through the components fields. We need in $`\mathrm{\Lambda }`$ only to present the final result in manifestly N=2 supersymmetric form.
N=1 form of non-holomorphic effective potential (4.6, 4.9) allows very easy to get leading bosonic component contribution. Schematically it has the form $`F^4/|\varphi |^4`$, where $`F_{mn}`$ is abelian strength constructed from vector component and $`\varphi `$ is a scalar component of N=2 vector multiplet ( see also ref ). It means that non-zero expectation value of scalar field $`\varphi `$ plays a role of effective infrared regulator in N=4 SYM theories.
Generalization of low-energy effective action discussed here and in refs \[18-26\] has recently been constructed in ref .
Acknowledgements
I am very grateful to E.I. Buchbinder and S.M. Kuzenko for collaboration on problem of non-holomorphic effective potential and to E.A. Ivanov, B.A. Ovrut and A.A. Tseytlin for valuable discussions. The work was supported in part by the RFBR grant 99-02-16617, by the DFG-RFBR grant 99-02-04022, by the INTAS grant No 96-0308 and by GRACENAS grant 97-6.2-34. |
warning/0001/cond-mat0001109.html | ar5iv | text | # Quasiparticles and Phase Fluctuations in High Tc Superconductors
## 1 INTRODUCTION:
The superconducting (SC) state of the high Tc cuprates differs from conventional SCs in several ways: a d-wave gap with low energy quasiparticle excitations near the nodes, a small phase stiffness and a short coherence length. There is some controversy about the importance of quasiparticles versus phase fluctuations in determining the low temperature properties. In this paper, we discuss this problem focusing mainly on the doping and temperature dependence of the in-plane superfluid stiffness $`D_{_{}}`$ which is related to the penetration depth $`\lambda _{_{}}`$ through $`\lambda _{_{}}^2=4\pi e^2D_{_{}}/\mathrm{}^2c^2d_c`$ where $`d_c`$ is the mean interlayer spacing; we will set $`\mathrm{}=c=e=1`$ below.
We first review experimental evidence for quasiparticle excitations at optimal doping. Transport data in the SC state in YBCO shows a scattering rate decreasing sharply below $`T_c`$ implying long lived quasiparticle excitations for $`TT_c`$. Direct evidence from ARPES in Bi2212 shows the presence of sharp quasiparticle peaks over the entire Fermi surface for $`TT_c`$. Thermal conductivity data in YBCO and Bi2212 shows $`\kappa T`$ at low $`T`$. The slope predicted by quasiparticle theory is in good agreement with this $`\kappa `$ data on Bi2212, using ARPES estimates for the Fermi velocity $`v_F`$ and the gap slope $`v_\mathrm{\Delta }=(2\mathrm{}k__F)^1d\mathrm{\Delta }/d\varphi `$ at the node. Experimentally, it thus seems that quasiparticle excitations exist and are important at low temperature.
It then seems natural to interpret the linear $`T`$ dependence of $`\lambda _{_{}}(T)`$ as arising from nodal quasiparticles (QP). Ignoring QP interactions the layer stiffness $`D_{_{}}(T)=D_{_{}}(0)A_0T`$ with $`A_0=(k__B\mathrm{ln}2/\pi )v_F/v_\mathrm{\Delta }`$. ARPES estimates in Bi2212 for $`v_F`$ and $`v_\mathrm{\Delta }`$ give $`A_00.8meV/K`$, whereas experiments at optimality measure a slope $`0.30.4meV/K`$. Thus there is at least a factor of two discrepancy which needs to be understood.
Alternatively, it has been suggested that this linear $`T`$ behavior could arise entirely from thermal phase fluctuations without invoking nodal quasiparticles. However, there are two reasons to believe that thermal phase fluctuations are unimportant at low $`T`$ in the cuprates. (1) An effective action calculation for charged $`d`$wave SCs , summarized below, shows that thermal phase fluctuations become important only near $`T_c`$ for Bi2212 at optimality. On the other hand, quantum phase fluctuations are important at low $`T`$ and suppress both $`D_{_{}}(0)`$ and the slope. For Bi2212, the resulting renormalized $`D_{_{}}`$ is about $`30\%`$ smaller while the renormalized slope is about $`25\%`$ smaller than the bare values. (2) Further, with underdoping, $`D_{_{}}(0)`$ decreases and the slope of $`D_{_{}}(T)`$ also shows evidence of decreasing in Bi2212 and La214, although some YBCO data is consistent with a doping-independent slope (see the compilation in refs. ). Insofar as the data indicate a doping dependent slope for $`D_{_{}}(T)`$, they independently rule out classical thermal phase fluctuations as the explanation for the linear $`T`$ dependence, since the slope of $`D_{_{}}`$ in such theories is insensitive to doping. Finally there is another proposal for understanding $`D_{_{}}(x;T)`$ invoking incoherent pair excitations which, in our opinion, does not properly include the effect of Coulomb interactions.
We next summarize our phase action calculation, and then describe how quasiparticle interactions could account for the difference between the free QP value and the measured slope and its doping dependence.
## 2 PHASE FLUCTUATIONS:
We have recently derived, by appropriate coarse-graining, a quantum XY model describing phase fluctuations in charged, layered $`d`$wave SCs, starting with a lattice model of fermions. See ref. for details of this derivation and some of the discussion in this Section.
The phase action for layered SCs with in-plane lattice spacing $`a=1`$ takes the form
$`S[\theta ]={\displaystyle \frac{1}{8T}}{\displaystyle \underset{𝐪,\omega _n}{}^{}}{\displaystyle \frac{\omega _n^2\xi _0^2}{\stackrel{~}{V}_𝐪}}|\theta (𝐪,\omega _n)|^2`$
$`+`$ $`{\displaystyle \frac{1}{4}}{\displaystyle _0^{1/T}}𝑑\tau {\displaystyle \underset{𝐫,\widehat{\alpha }}{}}D_{_{}}^^F\left[1\mathrm{cos}(\theta _{𝐫,\tau }\theta _{𝐫+\widehat{\alpha },\tau })\right]`$
where $`\xi _0`$ is the in-plane coherence length, and $`D_{_{}}^^F`$ refers to the layer stiffness without phase fluctuation effects, but including possible renormalizations due to quasiparticle interactions. Here $`\stackrel{~}{V}_𝐪=V(𝐪_{_{}}/\xi _0,𝐪_{_{}})`$ with $`V(𝐪)=(2\pi e^2/q_{}ϵ_b)\mathrm{sinh}(q_{}d_c)/\left[\mathrm{cosh}(q_{}d_c)\mathrm{cos}(q_{}d_c)\right]`$ is the Coulomb interaction for layered systems, $`ϵ_b`$ is the background dielectric constant, $`d_c`$ the interlayer spacing, and $`𝐪_{_{}},𝐪_{_{}}`$ refer to in-plane and $`c`$-axis momentum components. The prime on the sum indicates a Matsubara frequency cutoff since the energy of the fluctuations should not exceed the condensation energy $`E_{\mathrm{cond}}=\frac{1}{8}D_{_{}}^^F(\pi /\xi _0)^2`$. The form of the first term of (1), which arises from coarse-graining up to a scale $`\xi _0`$, and the importance of cutoffs have not been appreciated earlier. The (typically very small) $`c`$-axis stiffness $`D_{_{}}^^F`$ can be ignored for in-plane properties since it was found not to lead to qualitative or quantitative changes.
We ignore vortices (transverse phase fluctuations) which are suppressed at low $`T`$ by their finite core energy. Analyzing longitudinal phase fluctuations for (1) within a self-consistent harmonic approximation (SCHA) leads to the renormalized stiffness $`D_{_{}}=D_{_{}}^^F\mathrm{exp}(\delta \theta ^2/2)`$ where $`\delta \theta ^2=(\theta _{𝐫,\tau }\theta _{𝐫+\alpha ,\tau })^2`$. Our numerical results can be simply understood as follows: $`\delta \theta ^2(T=0)\sqrt{(e^2/ϵ_b\xi _0)/D_{_{}}(0)}`$ is a measure of zero point quantum fluctuations, while classical thermal phase fluctuations become important near a crossover scale $`T_\times \mathrm{min}[T_c,T_\times ^0]`$ where $`T_\times ^0=\sqrt{D_{_{}}(0)(e^2/ϵ_b\xi _0)}`$ is the $`T=0`$ oscillator level spacing in the renormalized harmonic theory.
It is easy to see that phase fluctuation effects are negligible in the BCS limit, except very close to $`T_c`$. With $`e^2/ϵ_baD_{_{}}^^FE__F`$ and $`\xi _0v_F/\mathrm{\Delta }`$, one obtains the standard result $`E_{\mathrm{cond}}\mathrm{\Delta }^2/E__F`$ per unit cell, and $`\delta \theta ^2(T=0)\sqrt{\mathrm{\Delta }/E__F}1`$ and $`T_\times \mathrm{min}[T_c,\sqrt{E__F\mathrm{\Delta }}]=T_c`$.
For the cuprates, the small $`\xi _0`$ and small $`D_{_{}}^^F`$ act together to increase $`\delta \theta ^2`$, but they push $`T_\times `$ in opposite directions. For optimal Bi2212 we use $`e^2/ϵ_ba0.33eV`$ with $`ϵ_b10`$, $`\xi _0/a10`$, and $`d_c/a4`$. Assuming that the two layers within a bilayer are phase-locked, we get the bilayer stiffness $`D_{_{}}(0)80meV`$, from experimental data which shows $`\lambda _{_{}}(0)2000A`$. This leads to $`E_{\mathrm{cond}}6K/\mathrm{unitcell}`$ and $`T_\times ^0600K`$. Since the bare stiffness $`D_{_{}}^^F`$ actually decreases with temperature due to quasiparticle excitations, an estimate of the crossover scale $`T_\times `$ can be obtained from $`T_\times \sqrt{D_{_{}}(T_\times )(e^2/ϵ_b\xi _0)}`$. Assuming a linearly decreasing $`D_{_{}}(T)`$, this leads to $`T_\times T_c`$. Thus longitudinal thermal fluctuations are clearly unimportant at low temperatures. Quantum fluctuations are important at low temperatures since $`\delta \theta ^2(T=0)1`$ at optimality and detailed calculations lead to a $`30\%`$ decrease of $`D_{_{}}^^F(0)`$ and a $`25\%`$ decrease in the slope.
While it might appear that there could be a low $`T`$ crossover to thermal phase fluctuations due to the low energy $`c`$-axis plasmon ($`7K`$ for Bi2212) in the anisotropic layered SCs, the phase space for these low lying fluctuations is too small to lead to a linear $`T`$ behavior . Even in a purely 2D system with a low lying $`\sqrt{q_{_{}}}`$ plasmon, the decrease in the phase stiffness due to phase fluctuations only goes as a large power law ($`T^5`$). Quasiparticles are thus crucial in obtaining the observed linear temperature dependence.
We next turn to the doping dependence of phase fluctuations. The (amplitude) coherence length $`\xi _0`$ is crucial in determining the effect of phase fluctuations. Since $`\xi _0`$ is determined by the pairing gap which appears to remain finite as we underdope, we do not expect singular behavior in the phase fluctuations arising from the doping dependence of $`\xi _0`$. In this case, the dominant doping dependence to phase fluctuations arises only from the singular behavior of the bare parameters in the phase action on underdoping. This singular $`x`$-dependence in $`D_{_{}}^^F(x)`$ is most naturally explained by quasiparticle interaction effects as discussed below.
To estimate the doping dependence of phase fluctuations, we note that the core energy will lead to vortices being exponentially suppressed at low $`T`$, even as we underdope. Using the experimental input $`T_cD_{_{}}(x;T=0)x`$ and assuming a doping independent $`\xi _0`$, longitudinal fluctuations within the SCHA lead to $`T_\times ^0\sqrt{x}`$ and hence $`T_\times ^0T_c`$. Thus, $`T_\times \mathrm{min}[T_c,T_\times ^0]T_c`$, which implies that thermal phase fluctuations are unimportant at low $`T`$ as one underdopes. Further, within the SCHA, $`\delta \theta ^2x^{1/2}`$, which would lead to a destruction of superconductivity at small enough $`x`$. However, we do not expect the SCHA to be valid close to this transition.
## 3 QUASIPARTICLE INTERACTIONS:
The increasing importance of interactions with underdoping is evident: $`D_{_{}}(x;0)x`$ and the quasiparticle weight diminishes as one approaches the Mott insulator. We thus explore the possibility that residual interactions between the quasiparticles in the SC state can account for the value and doping dependence of the slope of $`D_{_{}}(x;T)`$. To this end we use a phenomenological superfluid Fermi liquid theory (SFLT) . All available experimental evidence on the ground state and low lying excitations suggests that the correlated SC state in the cuprates is adiabatically connected to a d-wave BCS state. We thus feel that SFLT may be a reasonable description of QP interactions in the SC state at low $`T`$, even though this formulation makes reference to a (hypothetical) $`T=0`$ normal Fermi liquid in which SC is induced by turning on a pairing interaction. (One could argue that approaching the superconducting phase from the overdoped side at $`T=0`$, one obtains a normal Fermi liquid to SC transition.)
The bilayer stiffness after including QP interaction effects is given by, $`D_{_{}}^^F(T)=\beta __FD_{_{}}^0(0)\alpha __F2(k__BT\mathrm{ln}2/\pi )v_F/v_\mathrm{\Delta }`$ where $`\alpha __F,\beta __F`$ are Fermi liquid renormalizations. We will constrain the Landau QP interaction function by demanding $`\beta __Fx`$, consistent with experiments and then determine the doping trends in $`\alpha __F`$.
To compute $`\alpha __F,\beta __F`$, using a standard Kubo formula in the quasiparticle basis, it is convenient to shift the origin of the Brillouin zone to the $`(\pi ,\pi )`$ point and describe the hole-like Fermi surface of Bi2212 in terms of an angle $`\varphi `$. The Landau $`f`$-function is denoted by $`f(\varphi ,\varphi ^{})`$. We define $`O_\varphi _0^{2\pi }𝑑\varphi k__F(\varphi )O(\varphi )/[2\pi |v_F(\varphi )|]`$. We get $`\beta __F=1+4\pi v_{Fx}(\varphi )v_{Fx}(\varphi ^{})f(\varphi ,\varphi ^{})_{\varphi \varphi }/v_{Fx}^2_\varphi `$ from the diamagnetic response of the free energy $`^2\delta F/𝐀_x^2`$ to an applied vector potential. The current carried by nodal quasiparticles is then renormalized by the factor $`\sqrt{\alpha __F}=1+v_{Fx}(\varphi )f(\varphi _n,\varphi )_\varphi /\left[\pi v_{Fx}(\varphi _n)\right]`$, relative to its non-interacting value, where the nodes are at $`\varphi _n=(2n1)\pi /4`$ with $`n=1\mathrm{}4`$.
We expand $`f(\varphi ,\varphi ^{})=_{mm^{}}F_{m,m^{}}[\mathrm{cos}(m\varphi +m^{}\varphi ^{})+\mathrm{cos}(m^{}\varphi +m^{}\varphi ^{})]`$ in a set of complete basis functions , where $`m,m^{}=0,\pm 1,\pm 2,\mathrm{}`$ with square lattice symmetry imposing $`m+m^{}=4p`$ with $`p=0,1,2,\mathrm{}`$. In an isotropic system only $`p=0`$ survives and $`k__F`$ and $`v__F`$ are $`\varphi `$-independent. However, as emphasized in ref. one then obtains $`\alpha __F=\beta __F^2x^2`$ in disagreement with experiments .
To illustrate how anisotropy can qualitatively change this scaling we keep only the leading $`p=0`$ term: $`f(\varphi ,\varphi ^{})=2F_{1,1}\mathrm{cos}(\varphi \varphi ^{})`$, but retain the full anisotropy of the dispersion seen in ARPES . We make a reasonable choice of $`F_{1,1}=P+Qx`$, with $`P`$ such that $`\beta __Fx`$ as $`x0`$, and $`Q`$ such that $`\beta __F=0.5`$ at $`x=0.2`$. This leads to $`\alpha __F(x)`$ shown in Fig.1, which is a weak function of doping. In general there are too many free parameters in the anisotropic case (an infinite set $`F_{m,m^{}}`$) for the theory to have predictive power; nevertheless the simple example above shows how the $`T=0`$ value and slope of $`D_{_{}}`$ can easily exhibit rather different $`x`$-dependences, and account for the experimentally observed $`D_{_{}}(x;T)`$.
We thus arrive at the following picture for the doping and temperature dependence of $`D_{_{}}`$. The bare stiffness arising from non-interacting quasiparticles is renormalized by both QP interactions and quantum phase fluctuations at low $`T`$ leading to the measured stiffness, $`D_{_{}}(x;T)`$. Its doping dependence, $`D_{_{}}(x;0)x`$, is determined by QP interactions while its linear $`T`$ behavior is governed by nodal QPs, with its slope renormalized by both QP interactions and quantum phase fluctuations.
Acknowledgments: We thank J.C. Campuzano, J. Mesot, M.R. Norman, C. Panagopoulos, T.V. Ramakrishnan and L. Taillefer for useful conversations. M.R. thanks the Indian DST for partial support through a Swarnajayanti fellowship. |
warning/0001/hep-th0001170.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Recent progress in the understanding of the role of tachyonic excitations in unstable brane systems has led to a new framework in which D-branes appear as topological defects in the worldvolume of these unstable systems. In this framework the tachyon is considered as a Higgs field which tends to develop a stable vacuum expectation value, and the solitonic configuration that appears after the condensation is a BPS D-brane (see and references therein). The process can be iterated by embedding the unstable tachyonic system onto a similar system of higher dimension, and one can derive in this manner brane descent relations.
The mechanism of tachyon condensation can be given a qualitative description by looking at the couplings in the Wess-Zumino term of the effective action corresponding to the unstable system. This analysis allows to describe as well NS-NS charged branes, in particular fundamental strings, as topological solitons in higher dimensional systems . In this case the tachyonic condensing charged object is extended and non-perturbative, and a more desirable quantitative description is far out of reach.
In the type IIB theory starting with an unstable system of D9, anti-D9 pairs of branes it is possible to classify the stable Dp-branes in the theory by analysing the homotopy groups of the vacuum manifold of the tachyonic field . Mathematically the Dp-brane charges are classified by the K-theory groups $`\stackrel{~}{K}(S^{9p})`$, which describe the equivalence classes of pairs of vector bundles that characterise the system of coincident D9, anti-D9 branes up to creation and annihilation . S-duality determines that, similarly, NS-NS branes can be classified by analysing the homotopy groups associated to the tachyonic field of systems of NS9, anti-NS9 pairs of branes . The K-theory description is identical to that of D-branes, but the equivalence classes of pairs of vector bundles characterise instead coincident NS9, anti-NS9 branes.
In the type IIA theory D-brane descent relations can be derived by analysing the homotopy groups of the vacuum manifold of the tachyonic field in a system of $`n`$ D8, anti-D8 pairs of branes and, recalling that the D8-brane can be obtained from a non-BPS D9-brane after tachyonic condensation , the Type IIA Dp-branes can finally be described as bound states of spacetime-filling branes, in this way preserving all the symmetries of the ten dimensional spacetime. Mathematically the D-brane charges are classified by $`K^1(S^{9p})`$ K-theory groups (see ).
In this paper we analyse the possibility of having brane descent relations in M-theory. We show that the brane descent relations that one expects to find by oxidation from Type IIA can be predicted through the analysis of the worldvolume effective action of a system of $`n`$ M9, anti-M9 pairs of branes. We propose as well the construction of a non-BPS M10-brane from which the BPS M9-brane is obtained after the tachyonic mode of an open M2-brane ending on it condenses. Therefore we are able to write the brane descent relations from M9, anti-M9 pairs in terms of M-theory spacetime-filling branes, thereby preserving the symmetries of the eleven dimensional theory.
The study of 9-branes and 10-branes in M-theory should be carried out in the context of massive eleven dimensional supergravity, given that the BPS M9-brane couples magnetically to the mass. It is well-known that a fully eleven dimensional Lorentz invariant massive supergravity cannot be constructed . Nevertheless, the massive Type IIA supergravity of Romans can be derived from eleven dimensions if the condition of eleven dimensional Lorentz invariance is relaxed and one assumes an isometric eleventh direction that is gauged in the supergravity action . In this paper, we will work in the framework of this massive eleven dimensional supergravity defined with a Killing isometry. This theory is therefore only invariant under ten dimensional Lorentz transformations. Accordingly, the branes living in this massive eleven dimensional background are described by worldvolume effective actions where the Killing direction is gauged<sup>1</sup><sup>1</sup>1Also additional terms proportional to the mass are added .. This is also the case for the non-BPS M10-brane, as we will see.
The Born-Infeld part of the M9-brane effective action has been constructed in , and the Wess-Zumino term, which is the relevant part for the study of brane solitons, has also been derived recently in . This construction provides a description of the M9-brane in terms of a gauged sigma-model<sup>2</sup><sup>2</sup>2In such a way that the right number of degrees of freedom of a vector multiplet is recovered , and the brane consistently propagates in a massive background.. The effective action of an M9, anti-M9 pair can be described by a similar but more complicated expression which contains as well explicit couplings to the tachyonic mode that must be present in this unstable system. This tachyonic excitation must appear in the spectrum of an open M2-brane stretched between the brane and the antibrane, as predicted by the duality with the type IIA theory, where the open string stretched between a D8, anti-D8 pair of branes contains this type of excitation. The effective action of the M9, anti-M9 pair can be derived by oxidating the effective action describing this D8, anti-D8 pair<sup>3</sup><sup>3</sup>3Effective actions describing type II Dp, anti-Dp pairs of branes have been constructed in .. Recalling that the brane antibrane system is characterised by two field strengths (one for the brane and one for the antibrane) and that the non-trivial character of the soliton can be carried by just one of the two, it is enough to look at the couplings in the worldvolume effective action of the brane, in our case of the M9-brane, in order to find out the topological defects that the system supports, taking into account as well that for this analysis the coupling of the tachyonic mode to the worldvolume effective action does not play any role.
Keeping this in mind we analyse in section 2 the topological defects that can occur in the worldvolume of $`n`$ M9, anti-M9 pairs of branes by studying the different terms that couple in the WZ part of the worldvolume effective action of a system of $`n`$ M9-branes. In this way we are able to derive brane descent relations in M-theory from pairs of M9, anti-M9 branes. Then in section 3 we propose the construction of a non-BPS, unstable, M10-brane from which the BPS M9-brane is obtained through tachyonic condensation. Again, this tachyonic mode must be associated to open M2-branes ending on the M10-brane, as required by the duality with type IIA. We thereby present the brane descent relations in terms of this spacetime-filling brane. In section 4 we discuss the existence of other non-BPS unstable branes in M-theory and the possibility of stabilising them within the Hor̆ava-Witten construction . In particular, in this picture the tachyonic mode of the non-BPS M10-brane is projected out. Finally section 5 contains some conclusions.
## 2 BPS M-branes as bound states
In this section we show how the M-theory BPS branes can be realised as bound states of $`n`$ (M9, anti-M9) pairs of branes. As we have mentioned in the introduction we can ignore the couplings of the tachyon in the worldvolume effective action of this system and set the field strength of the antibranes to zero, so that we can simply study the effective action of $`n`$ coincident M9-branes. We use the WZ term constructed recently in , but only consider explicitly those couplings that are relevant for the analysis of the topological defects. We also ignore all numerical prefactors and factors of $`\alpha ^{}`$ as well as the explicit couplings to the mass.
The WZ term of the M9-brane effective action contains the couplings<sup>4</sup><sup>4</sup>4Eleven dimensional fields are denoted with hats. :
$`S_{\mathrm{WZ}}^{(M9)}={\displaystyle _{R^{8+1}}}[i_{\widehat{k}}\widehat{B}^{(10)}+i_{\widehat{k}}\widehat{N}^{(8)}\widehat{}+i_{\widehat{k}}\widehat{\stackrel{~}{C}}\widehat{}\widehat{}+`$ (2.1)
$`+\widehat{C}_{\widehat{\mu }_1\widehat{\mu }_2\widehat{\mu }_3}D\widehat{X}^{\widehat{\mu }_1}D\widehat{X}^{\widehat{\mu }_2}D\widehat{X}^{\widehat{\mu }_3}\widehat{}\widehat{}\widehat{}+\widehat{A}\widehat{}\widehat{}\widehat{}\widehat{}+\mathrm{}],`$
where the dots indicate the couplings that we have omitted. This action is invariant under the local isometric transformations generated by the Killing vector $`\widehat{k}^{\widehat{\mu }}`$, since the pull-backs onto the worldvolume occur with covariant derivatives (defined below) or contracting with the Killing vector. This vector has to be taken along a worldvolume direction in order to obtain the D8-brane after double dimensional reduction (see , ). $`\widehat{B}^{(10)}`$ is the spacetime field electric-magnetic dual to the mass, $`\widehat{N}^{(8)}`$ is the dual of the Killing vector considered as a 1-form , and $`\widehat{C}`$, $`\widehat{\stackrel{~}{C}}`$ are, respectively, the 3- and 6-forms of eleven dimensional supergravity. $`\widehat{}`$ is the field strength of the worldvolume vector field describing an M2-brane, wrapped on the Killing direction, ending on the brane: $`\widehat{}=d\widehat{b}^{(1)}+i_{\widehat{k}}\widehat{C}\widehat{F}+i_{\widehat{k}}\widehat{C}`$. Finally, $`\widehat{A}\widehat{k}^2\widehat{k}_{\widehat{\mu }}\widehat{X}^{\widehat{\mu }}`$, and the covariant derivatives are defined as: $`D\widehat{X}^{\widehat{\mu }}=\widehat{X}^{\widehat{\mu }}+\widehat{A}\widehat{k}^{\widehat{\mu }}`$ (see and for more details).
The M6-brane, or M-theory Kaluza-Klein monopole, is realised as a bound state (M9, anti-M9) . This can be simply read from the term:
$$_{R^{8+1}}i_{\widehat{k}}\widehat{N}^{(8)}\widehat{F}$$
(2.2)
in the M9-brane effective action. As mentioned in the introduction, there must be a tachyonic mode in the spectrum of a wrapped M2-brane stretched between the M9 and the anti-M9, whose condensation will be accompanied by a non-trivial magnetic flux, so that a coupling:
$$_{R^{6+1}}i_{\widehat{k}}\widehat{N}^{(8)}$$
(2.3)
remains in the worldvolume. The M-theory Kaluza-Klein monopole is charged with respect to this field , so this is the topological defect that remains after the condensation.
The M5-brane is realised, in turn, as an instanton-like configuration<sup>5</sup><sup>5</sup>5Since a single M9-brane contains in its worldvolume a U(1) vector field a system of $`n`$ coincident M9-branes is described by a U($`n`$) worldvolume gauge theory. Therefore the corresponding effective action is given by the same expression (2.1) with $`F`$ U($`n`$) and where the trace is taken over the U($`n`$) indices.. The term:
$$_{R^{8+1}}(i_{\widehat{k}}\widehat{\stackrel{~}{C}})\mathrm{Tr}(\widehat{F}\widehat{F})$$
(2.4)
in the worldvolume effective action of $`n`$ M9-branes gives a coupling
$$_{R^{4+1}}(i_{\widehat{k}}\widehat{\stackrel{~}{C}})$$
(2.5)
when $`_{R^4}\mathrm{Tr}(\widehat{F}\widehat{F})=\mathrm{integer}`$, i.e. when the homotopy group $`\mathrm{\Pi }_3(U(n))=Z`$. This happens for all $`n>2`$, however in the particular case $`n=2^{k1}`$, where $`2k`$ is defined as the codimension of the topological defect (in this case $`k=2`$), it is possible to give a representation of the tachyon vortex configuration (the generator of $`\mathrm{\Pi }_{2k1}(U(n))`$) such that all higher and lower dimensional charges vanish . For our particular case $`n=2`$, and the M5-brane is realised as a bound state M5=2 (M9, anti-M9) <sup>6</sup><sup>6</sup>6This definition of the M5-brane seems to imply that a pile of coincident M5-branes would simply be described by a U($`m`$) gauge theory. This does not contradict the fact that the M5-brane field content must be that of the six dimensional antisymmetric tensor multiplet, whose non-abelian extension is not known, because in the bound state construction the M5-brane is wrapped on the Killing direction of the M9-brane and therefore its field content must be that of the five dimensional vector multiplet..
The term responsible for the realisation of the M2-brane as a topological defect is:
$$_{R^{8+1}}\widehat{C}_{\widehat{\mu }_1\widehat{\mu }_2\widehat{\mu }_3}D\widehat{X}^{\widehat{\mu }_1}D\widehat{X}^{\widehat{\mu }_2}D\widehat{X}^{\widehat{\mu }_3}\mathrm{Tr}(\widehat{F}\widehat{F}\widehat{F}).$$
(2.6)
Now we have $`\mathrm{\Pi }_5(U(4))=Z`$, and the topological defect is realised as M2=4 (M9, anti-M9), given that the non-trivial integration of the gauge field gives rise to a coupling:
$$_{R^{2+1}}\widehat{C}_{\widehat{\mu }_1\widehat{\mu }_2\widehat{\mu }_3}D\widehat{X}^{\widehat{\mu }_1}D\widehat{X}^{\widehat{\mu }_2}D\widehat{X}^{\widehat{\mu }_3},$$
(2.7)
which describes an M2-brane with an isometric transverse direction<sup>7</sup><sup>7</sup>7Since the M2-brane cannot move along this direction, which plays the role of the eleventh coordinate, it behaves effectively as a D2-brane, and a set of such coincident M2-branes can then be described by a U($`m`$) gauge theory..
Finally, the term
$$_{R^{8+1}}\widehat{A}\mathrm{Tr}(\widehat{F}\widehat{F}\widehat{F}\widehat{F})$$
(2.8)
describes an M-wave as a bound state of 8 (M9, anti-M9) pairs. In this case $`\mathrm{\Pi }_7(U(8))=Z`$, and the field that remains after the condensation: $`\widehat{A}=\widehat{k}^2\widehat{k}_{\widehat{\mu }}\widehat{X}^{\widehat{\mu }}`$ is the field to which the M-wave couples minimally .
These realisations of M-branes as bound states of M9, anti-M9 pairs are those that one would obtain by oxidising the brane descent relations in the Type IIA theory in terms of D8, anti-D8 pairs of branes. Here we have shown that the M9-brane worldvolume effective action correctly contains the couplings describing these realisations.
In the type IIA theory it is possible to write the brane descent relations in terms of non-BPS spacetime-filling D9-branes, and in this way all the symmetries of the spacetime are preserved . This D9-brane is unstable because the open strings ending on it contain tachyonic excitations. These modes however can condense in a kink (resp. anti-kink) configuration, giving rise to a D8 (resp. anti-D8) brane as the topological defect. Oxidation to M-theory predicts then an unstable M10-brane from which the M9-brane should be obtained after tachyonic condensation. In the next section we analyse this possibility.
## 3 The non-BPS M10-brane
Having the action of the M9-brane it is possible to construct the action of a non-BPS M10-brane which gives rise to this brane after tachyonic condensation. In this case, since the field strength $`\widehat{}`$ is given by: $`\widehat{}=d\widehat{b}^{(1)}+i_{\widehat{k}}\widehat{C}`$, the tachyonic mode must be associated to an M2-brane, wrapped on the Killing direction, ending on the M10-brane.
The clue for the construction of the effective action is to realise that the M10-brane should reproduce the Type IIA non-BPS D9-brane upon double dimensional reduction. The WZ term of the action of type II non-BPS D$`p`$-branes has been constructed in . In the particular case of the D9-brane it reads<sup>8</sup><sup>8</sup>8We ignore the contribution of the A-roof genus and, again, all numerical prefactors and factors of $`\alpha ^{}`$.:
$`S_{\mathrm{WZ}}^{(D9)}`$ $`=`$ $`{\displaystyle _{R^{9+1}}}[C^{(9)}+C^{(7)}+C^{(5)}+C^{(3)}+`$ (3.1)
$`+C^{(1)}]dT,`$
where $`C^{(p)}`$ denotes the $`p`$-form RR-potential, $`=F+B^{(2)}`$, with $`F=db^{(1)}`$ and $`B^{(2)}`$ the NS-NS 2-form, and $`T`$ stands for the, real, tachyon field induced in the worldvolume by the open strings ending on the brane. When the tachyon condenses to a non-trivial kink configuration, depending on a single coordinate $`x`$ and such that $`𝑑T(x)=\pm 2T_0`$, where $`\{T_0,T_0\}`$ are the two minima of the tachyon potential, we have, in the limit of zero size: $`dT(x)=2T_0\delta (xx_0)dx`$, and substituting in (3.1) the effective action of a BPS D8-brane localised in $`x_0`$ is obtained .
The WZ part of the effective action of the proposed non-BPS M10-brane is then given by<sup>9</sup><sup>9</sup>9We use dots to denote those terms that are not relevant for the analysis of the topological defects.:
$`S_{\mathrm{WZ}}^{(M10)}={\displaystyle _{R^{9+1}}}[i_{\widehat{k}}\widehat{B}^{(10)}+i_{\widehat{k}}\widehat{N}^{(8)}\widehat{}+i_{\widehat{k}}\widehat{\stackrel{~}{C}}\widehat{}\widehat{}+\widehat{C}_{\widehat{\mu }_1\widehat{\mu }_2\widehat{\mu }_3}.`$ (3.2)
$`.D\widehat{X}^{\widehat{\mu }_1}D\widehat{X}^{\widehat{\mu }_2}D\widehat{X}^{\widehat{\mu }_3}\widehat{}\widehat{}\widehat{}+\widehat{A}\widehat{}\widehat{}\widehat{}\widehat{}+\mathrm{}]d\widehat{T},`$
since this action reproduces (3.1) after double dimensional reduction along the Killing direction. Also, the condensation of the tachyon field into a non-trivial kink configuration, $`d\widehat{T}(x)=2\widehat{T}_0\delta (xx_0)dx`$, in the limit of zero size, gives the M9-brane effective action that we considered in the previous section. Like in the non-BPS D9-brane, the tachyon field living in the worldvolume of the M10-brane is a real scalar, in this case induced by open M2-branes wrapped on the Killing direction<sup>10</sup><sup>10</sup>10The presence of a tachyonic mode in the spectrum of open M2-branes ending on the M10-brane is inferred by duality with the type IIA theory, but being this a strong-weak coupling duality the open M2-branes are intrinsically non-perturbative and the presence of this instability cannot be tested by any perturbative methods..
Regarding the DBI part of the effective action, it has been argued in that non-BPS Dp-branes may be described by a DBI action<sup>11</sup><sup>11</sup>11Here we are only concerned with the bosonic part of the action. See for more details.:
$$S_{\mathrm{DBI}}^{(\mathrm{Dp})}=_{R^{p+1}}\sqrt{|\mathrm{det}(G+)|}R(T,T,\mathrm{}),$$
(3.3)
where $`R`$ is some function of the tachyon field vanishing at the minimum of the tachyon potential. In this way the worldvolume action vanishes identically, and can describe a configuration indistinguishable from the vacuum. On the other hand, for vanishing tachyon field $`R`$ gives a constant, and the action in for a non-BPS Dp-brane involving the massless fields is recovered. For a tachyonic kink configuration in the zero size limit it has been argued that $`R(x)C\delta (xx_0)`$ , and (3.3) reduces to the DBI part of the effective action of a BPS D(p-1)-brane.
Similarly, we can argue that the DBI part of the non-BPS M10-brane effective action is given by:
$$S_{\mathrm{DBI}}^{(\mathrm{M10})}=_{R^{9+1}}|\widehat{k}|^3\sqrt{|\mathrm{det}\left(\widehat{\mathrm{\Pi }}+|\widehat{k}|^1\widehat{}\right)|}\widehat{R}(\widehat{T},\widehat{T},\mathrm{}),$$
(3.4)
such that when the tachyon condenses to its minimum the DBI effective action of the M9-brane is recovered :
$$S_{\mathrm{DBI}}^{(\mathrm{M9})}=_{R^{8+1}}|\widehat{k}|^3\sqrt{|\mathrm{det}\left(\widehat{\mathrm{\Pi }}+|\widehat{k}|^1\widehat{}\right)|}.$$
(3.5)
Here $`\widehat{\mathrm{\Pi }}`$ is the pull-back of the spacetime metric:
$$\widehat{\mathrm{\Pi }}=D\widehat{X}^{\widehat{\mu }}D\widehat{X}^{\widehat{\nu }}\widehat{g}_{\widehat{\mu }\widehat{\nu }}=\widehat{X}^{\widehat{\mu }}\widehat{X}^{\widehat{\nu }}\left(\widehat{g}_{\widehat{\mu }\widehat{\nu }}+|\widehat{k}|^2\widehat{k}_{\widehat{\mu }}\widehat{k}_{\widehat{\nu }}\right).$$
(3.6)
The proposed M10-brane action, given by (3.2) and (3.4), contains a Killing direction in its worldvolume. Therefore it is only invariant under ten dimensional Lorentz transformations, consistently with the fact that it is defined in a massive eleven dimensional background. In this sense the M10-brane that we have constructed is spacetime-filling and preserves all the symmetries of the massive eleven dimensional spacetime, including the invariance under local isometric transformations.
Finally, since both the M9 and the anti-M9 branes can be obtained from the M10-brane when the tachyon condenses to a kink or an anti-kink configuration , we can write the brane descent relations that we derived in the previous section in terms of spacetime-filling branes as: M9=M10, M6=2 M10, M5=4 M10, M2=8 M10, M-wave=16 M10 <sup>12</sup><sup>12</sup>12The WZ part of the effective action describing the set of $`n`$ coincident M10-branes is of the same form (3.2) but now $`\widehat{F}`$ U($`n`$), the tachyon transforms in the adjoint representation and traces over the U($`n`$) indices are taken..
## 4 Other non-BPS branes in M-theory
We have seen how the duality between M-theory and type IIA predicts the existence of an unstable non-BPS M10-brane from which the type IIA D9-brane is obtained after double dimensional reduction. Similarly, it is possible to construct unstable non-BPS branes in M-theory giving rise to the whole spectrum of non-BPS D-branes in type IIA. One finds that apart from the M10-brane there are M(-1), M1, M4, M5 and M8 non-BPS branes<sup>13</sup><sup>13</sup>13Among these branes, the M(-1), M1, M4 and M5 can exist as well in massless M-theory.. These branes can be described as bound states of BPS M-branes as: Mp=(M(p+1), anti-M(p+1)). In this brane antibrane unstable system the complex tachyonic excitation in the open M2-branes stretched between the brane and the antibrane condenses through a kink configuration, giving rise to the non-BPS M-brane, which is also unstable, because the open M2-branes ending on it contain still a real tachyonic excitation.
The tachyonic excitation in the non-BPS Mp-brane can in turn condense, and give rise to a BPS M-brane. A careful analysis shows that the dimension of this resulting M-brane depends on whether the M2-brane ending on the non-BPS Mp-brane, whose tachyonic mode is condensing, is wrapped or unwrapped <sup>14</sup><sup>14</sup>14 As explained in certain realisations of branes as solitons in brane anti-brane systems require to choose one special direction.. For the M1, M5 and M8 branes the tachyonic mode is associated to a wrapped M2-brane ending on the brane, and one can see that the BPS object that remains after the condensation must be an M0, a wrapped M5 and an M6 brane respectively. In the case of the non-BPS M4-brane the tachyonic mode is associated to unwrapped M2-branes ending on the brane, and the object that remains after the condensation is an M2-brane.
Thus, the existence of these non-BPS branes in M-theory allows the construction of BPS branes from brane antibrane pairs in two steps. First, the complex tachyonic mode of the brane stretched between the brane and the antibrane condenses, giving rise to a non-BPS M-brane, with a real tachyonic mode associated to open M2-branes ending on it. Second, this tachyonic mode condenses in the same type of configuration and a stable BPS M-brane emerges. This generalises to M-theory the two-step construction of type II theories of .
It is interesting to note that the non-BPS M-branes that we have considered give rise, upon reduction, not only to non-BPS Dp-branes but also to some additional non-BPS branes in the type IIA theory. These branes predict in turn the existence of similar non-BPS branes in type IIB by T-duality. We have mentioned in the introduction that BPS NS-NS and gravitational branes in the type IIA and type IIB theories can be interpreted as solitons in brane antibrane systems , with tachyonic condensing charged objects that are extended and non-perturbative. In the realisations discussed in these references the complex tachyonic mode condenses in a vortex-like configuration, and the resulting solitonic object is stable and carries NS-NS or gravitational charge. One could instead consider a two-step construction of these BPS branes as in , and this would lead to new intermediate non-BPS “NS-NS” and “gravitational” branes<sup>15</sup><sup>15</sup>15NS-NS and gravitational in the sense that they are derived in the two-step construction of this kind of branes, but it is clear that they do not carry any charge. as remnants of tachyonic kink configurations. These are the additional non-BPS branes that are derived from M-theory.
### 4.1 Stable non-BPS M-branes in the Hor̆ava-Witten construction
Unstable non-BPS branes can become stable when the theory is orbifolded/ orientifolded by an appropriate symmetry (see ). Therefore, it is interesting to analyse whether the unstable non-BPS M-branes that we have discussed can become stable in the framework of the Hor̆ava-Witten construction . One can predict stable non-BPS branes in this construction by uplifting the stable non-BPS branes of the type I’ theory, which have in turn been analysed in by using the T-duality connection with type I.
The spectrum of stable non-BPS type I’ branes consists on a D(-1), D0, D1, D6, D7, D8 and D9 branes, where the D0, D1, D8 and D9 are stretched in the interval<sup>16</sup><sup>16</sup>16The D6, D7, D8 and D9 branes are however unstable due to the presence of a tachyonic mode in the open strings with one end on the brane and the other in one of the 16 + 16 D8-branes located on top of the orientifold fixed planes. This can be inferred from the type I case .. These branes can be constructed as bound states: D(-1)=(D0, anti-D0), D0=(D0, anti-D0) (in this case the D0’s are ‘stretched’), D1=(D2, anti-D2) (stretched), D6=(D6, anti-D6), D7=(D8, anti-D8), D8=(D8, anti-D8) (stretched). Uplifting these configurations to M-theory we find:
M(-1)=(M0, anti-M0), M0=(M0, anti-M0) (‘stretched’), M1=(M2, anti-M2) (stretched), M6=(M6, anti-M6), M8=(M9, anti-M9), M9=(M9, anti-M9) (stretched), together with a stretched M10-brane.
We can thus conclude that the non-BPS M(-1), M1, M8 and M10 branes of uncompactified M-theory are stabilised when one considers M-theory in the interval. Moreover, we find additional branes realised as BPS brane antibrane pairs in which the tachyonic mode is removed from the spectrum by the orbifold projection, and no condensation occurs. This happens for the M0, M6 and M9 non-BPS branes.
The Hor̆ava-Witten construction can be described as M-theory orientifolded by $`I_{10}\widehat{\mathrm{\Omega }}`$ (see ), where $`I_{10}:x^{10}x^{10}`$ and $`\widehat{\mathrm{\Omega }}`$ is the operation that reverses the orientation of the M2 and the M5 branes. In this description the end of the world branes of Hor̆ava-Witten are identified as the two fixed planes associated to the orientifold projection with 16 M9-branes on top of them. The connection with the type I’ theory through compactification on $`S^1`$ implies that the M6, M8, M9 and M10 branes must become unstable due to the presence of a tachyonic mode in the open M2-branes with one end on the brane and the other in one of the 16 + 16 M9-branes located on the top of the orientifold fixed planes.
## 5 Conclusions
We have shown how the brane descent relations that one expects to find in M-theory by oxidation from Type IIA are indeed predicted by the analysis of the topological defects that can occur in the worldvolume of $`n`$ M9, anti-M9 pairs of branes. We have proposed as well a non-BPS, unstable, M10-brane, from which the brane descent relations can be expressed in terms of spacetime-filling branes, preserving all the symmetries of the spacetime. Since massive eleven dimensional supergravity is at present only known explicitly when the spacetime contains a Killing isometry , the corresponding M10-brane contains as well this Killing isometry in its worldvolume, and therefore it is not fully eleven dimensional Lorentz invariant, consistently with the type of eleven dimensional spacetime in which it is defined.
Using the relation between ordinary and higher K-theory groups (see , ): $`\stackrel{~}{K}(S^{10p})=K^1(S^{9p})`$ , it is inferred that the branes that can be obtained in the descent construction in this massive M-theory are classified according to:
$$\stackrel{~}{K}(S^{10p})=\text{ }\{\begin{array}{ccccc}Z& \mathrm{for}& p& \mathrm{even}& \\ 0& \mathrm{for}& p& \mathrm{odd}& \end{array}\text{ .}$$
Therefore the K-theory groups predict the M-wave, M2 and M6 branes<sup>17</sup><sup>17</sup>17For the M-wave and the M6 the $`S^1`$ direction coincides with their special isometric direction., and an M4 and M8 branes, which are identified as wrapped M5 and M9 branes, identification that is supported by the bound state analysis that we have carried out in this paper.
We have discussed as well the existence of other non-BPS branes in M-theory, giving rise to type IIA non-BPS D-branes after dimensional reduction. These M-branes predict unstable non-BPS “NS-NS” and “gravitational” branes in both the type IIA and type IIB theories, which would arise naturally in a two-step construction, as in , of NS-NS and gravitational BPS branes.
Finally, we have shown that when M-theory is compactified in the interval some of the non-BPS M-branes become stable, in particular the spacetime-filling M10-brane, since they give rise to stable non-BPS branes in type I’ after reduction. The K-theory classification of stable M-branes in the Hor̆ava-Witten picture seems to be given by the K-theory group: $`KR(S^{9p}\times S^{1,1},S^{1,1})`$, which provides the M-theory interpretation of the type I’ K-theory group : $`KR^1(S^{8p}\times S^{1,1},S^{1,1})`$. Some of the non-BPS branes predicted by this group become however unstable due to the presence of background M9-branes in the Hor̆ava-Witten picture, which, as in the type I case , do not play any role in the K-theory classification.
### Acknowledgements
L. H. would like to acknowledge the support of the European Commission TMR programme grant ERBFMBICT-98-2872, and the Theory Division at CERN, where this work has been completed, for hospitality. The I.C. Theory Group is supported by PPARC under SPG grant 613. |
warning/0001/cond-mat0001353.html | ar5iv | text | # 1 The hidden collective factor of speculative bubbles
## 1 The hidden collective factor of speculative bubbles
In economic models, the price of a commodity is determined by its supply and demand functions. In the presence of unexpected events, “explanations” are forged by mechanisms that appear to be similar in economics and in history.
### 1.1 “Explanations” of unexpected events
When the king of Sweden Charles XII was defeated by the Russians at the battle of Poltava (1709), Europe was stunned for until then the Swedish army was considered invincible. Subsequently, the defeat was “explained” in various ways. For instance, several historians (e.g. Andersson 1973, Scott 1977) attributed the defeat to the fact that, having been wounded in the leg two days before, Charles XII could not lead his army. Even if this argument appears reasonable, one can doubt its soundness when one notes the huge numerical superiority of the Russians, a point which surprisingly enough is only seldom mentioned by historians: they were 45,000 against 26,000 Swedes and above all they had 72 guns while the Swedes had only four (Bodart 1908, p.159).
As an economic example, let us consider the almost simultaneous crashes in the diamond, gold, palladium, platinum and silver markets at the beginning of 1980 (Fig.1). Amazingly enough, different stories were invented in each of these cases to “explain” the crash. For the silver market, it was the attempt by the Hunt brothers to squeeze the market, i.e. to buy up all available silver in order to corner short sellers (see in this respect Fay 1982); it should be emphasized that no precise data are available to support and validate that story quantitatively. In the diamond market, depending upon the author, the reason invoked was the attempt by the cutters in Tel Aviv to hoard uncut diamonds for the purpose of speculation or the short-lived boom in diamond investment (Boyajian 1988, Epstein 1982). Other stories, which we omit here for the sake of brevity, were proposed to account for the crashes in the other markets. It can be noted that all these stories are (i) plausible (ii) purely qualitative and (iii) fairly picturesque; these features certainly explain why they were easily memorized and used over and over again by various authors. Such explanations are so to say self-fulfilling rationalizations. It is as if one would explain the fall of apples from from different trees in one case by a swing of the branch, in another by the fact that the apple had become too big, and so on, but without mentioning general causes such as the effect of a gust of wind and the role of the gravitational attraction.
### 1.2 The hypothesis of a collective factor due to imitation processes
Fig.1 suggests the existence of a common factor which may account for the simultaneous bubbles and crashes of these markets. The hypothesis of the existence of a common factor is a natural scientific endeavor: explain more data with less assumptions, in other words follow Occam’s razor principle of parsimony. The existence of a common factor would explain for instance why the speculative bubbles for commodities as different (in terms of demand and supply functions) as diamonds, gold, palladium, platinum and silver burst almost simultaneously in January 1980 as shown in figure 1. We propose that this common factor is an imitation process between traders and present quantitative empirical tests for this hypothesis.
### 1.3 Definition of a bubble
Speculative bubbles and crashes are among the most important tumultuous and extreme events that mark long time series of economic and financial data. Notwithstanding their apparent intuitive meaning and their fixation in the minds of the public, defining and identifying bubbles and crashes is not always obvious. With respect to crashes, a natural characterization is to define them as outliers, i.e. cumulative drops beyond the normality (Johansen and Sornette, 1998). Defining a bubble is even more trickly because acceptance or rejection of the existence of a bubble is contingent on the model of fundamentals one uses: indeed, the conclusion that a bubble exists in the data can emerge only from using a specific model of fundamental prices (Flood and Garber, 1994).
The usual conception is that a bubble can arise when the actual market price depends positively on its own expected rate of change. In the rational expectation framework, there is no systematic discrepancies between expectations and realized prices. Therefore, the actual rate of change is positively correlated to its expected value and the actual market price is also positively related to the actual rate of change. A simple example is
$$\frac{dp}{dt}=ap^b,\mathrm{with}b>0.$$
(1)
If $`b<1`$, the price $`p`$ exhibits a power law acceleration with an upward concavity proportional to $`(t+t_0)^{1/(1b)}`$, where $`t_0`$ depends on the initial price. If $`b1`$, the acceleration is so strong that the price diverges in finite time as in a critical point (Stanley, 1987) or “movable singularity” (Bender and Orszag, 1978) as $`(t_ct)^{1/(b1)}`$, where the critical time $`t_c`$ depends on the initial price. In such conditions, the expectation of price changes becomes self-fulfilling and drives actual price changes independently of market fundamentals. This is the definition of a price bubble (Flood and Garber, 1994).
### 1.4 How to detect a bubble?
Notwithstanding this deceivingly simple definition, as we said, the very notion of a bubble makes no sense without a precise model detailing the market behavior. Indeed, without a model, it is impossible to isolate the price trajectory characterizing a bubble: an apparent anomalous price behavior could actually result from the rational behavior of informed traders reacting to important news. A large literature has thus been devoted to the specification of reasonable models and to the identification of favorable time periods, like hyperinflation where it is believed that the work of disentangling the bubble trajectory from the fundamental evolution is easier (Flood and Garber, 1994).
A second approach is to try to extract “universal” results on the expected behavior of bubbles that are independent of the assumed underlying fundamentals. In this respect, using insights on the behavior of multiplicative stochastic processes and the no-arbitrage condition, Lux and Sornette (1999) have shown recently that rational bubbles à la Blanchard and Watson (1982) have a fat power tail for both the bubble price component, its price difference and its returns with an exponent $`\alpha `$ always smaller than $`1`$. In contrast, the usual empirical estimates give $`\alpha 34`$. It therefore appears that exogenous rational bubbles are hardly reconcilable with some of the stylized facts of financial data at a very elementary level.
Another approach developed in the last few year using analogies with physical phenomena is to characterize a highly specific signature that can be used as a fingerprint for a bubble. A notable example is the robust and apparently universal signature found in bubbles ending in large crashes or large corrections found in major financial stock markets as well as in emergent markets, namely accelerated price increase with a sharp peak (Roehner and Sornette, 1998) decorated by large scale log-periodic oscillations accelerating up to a kind of critical point (Johansen, Sornette and Ledoit, 1999; Johansen and Sornette, 1999 and references therein).
Here, we propose yet another method which bypasses all the difficulties of the previous approaches by monitoring external indicators which show an anomalously growing interest in the public at times of bubbles. Such indicators can be for instance the number of publications and books published on the topic and the size of their sales during bubble periods. From the definition of a bubble as a self-fulfilling reinforcing price change, we thus search for indicators of a possible self-reinforcing imitation between agents in the market. We will show that such tendency for traders to imitate their nearest “neighbors” is self-reinforcing and increases up to a certain point called the “critical” point, at which all traders may place the same order (sell) at the same time, thus causing the burst of the bubble (a crash, a large correction or more generally a change of regime). The main point of our paper is that, during a major speculative bubble, the fever does not remain confined to the economic or financial sphere but spreads to other segments of the society which can actually become actors themselves by buying the market, usually close to the end of the bullish mood. This “fever contamination” provides a probe of the imitation process which is at the source of the speculative bubble, and gives therefore a direct access to the fundamental mechanism of the bubble. We believe that this approach is better suited to qualify the existence of bubble than other methods relying on poorly constrained models of the fundamental prices.
### 1.5 Interaction channels
How can speculative fever spread? Sociologists distinguish three kinds of communication channels (Deutsch 1953): (i) inter-personal communication i.e. contacts between individuals by word of mouth, phone, email, etc; (ii) commercial communication channels such as newspapers, magazines, books, radio, television, films, etc.; (iii) institutional channels such as churches, political parties, etc. In terms of interaction range, the channel (i) is basically a short-range interaction in the (not necessarily spatial) sense that most of the persons that you know personally belong to the same segment of society while channel (ii) and also to some extent channel (iii) are “long range” interactions. In order to assess the role played by these channels, the crucial point is to find sources from which one can draw quantitative evidence. In this study, we restrict ourselves to one specific commercial channel namely the publishing business.
### 1.6 Content of the paper
In section 2, we propose a simple model of price evolution that shows that imitation may result from rational expectations. The resulting Ising model dynamics embodies the possible existence of a speculative bubble and its culmination at a critical point. In section 3, we show how the progress of the speculative frenzy can be estimated through the production of books on related topics. In section 4, we use the volume of transactions in a given market as a measure of its widening attraction power.
## 2 Imitation results from rational maximization of expected profit
### 2.1 Definition of the model
In this section, we follow and generalize the work of Bhamra (1999). Consider $`N`$ traders in a network, whose links represent the communication channels through which the traders communicate. The graph is not euclidean in general and is probably better described by the “small-world” type of topology (Watts, 1999) found to describe the short chain of intermediate acquaintances between any two people in the world (colloquially referred to as the “six degrees of separation”). We denote $`N(i)`$ the number of traders directly connected to $`i`$ on the graph. The traders buy or sell one asset at price $`p(t)`$ which evolves as a function of time assumed to be discrete and measured in units of the time step $`\mathrm{\Delta }t`$. In the simplest version of the model, each agent can either buy or sell only one unit of the asset. This is quantified by the buy state $`s_i=+1`$ or the sell state $`s_i=1`$. Each agent can trade at time $`t1`$ at the price $`p(t1)`$ based on all previous informations including that at $`t1`$. We assume that the asset price variation is determined by the following equation
$$\frac{p(t)p(t1)}{p(t1)}=\mathrm{\Delta }tF\left(\frac{\underset{i=1}{\overset{N}{}}s_i(t1)}{N}\right)+\sqrt{\mathrm{\Delta }t}\sigma \eta (t).$$
(2)
$`\sigma `$ is the price volatility per unit time, the factor $`\sqrt{\mathrm{\Delta }t}`$ represents the time factor of the volatility and $`\eta (t)`$ is a white Gaussian noise with unit variance. The first term in the r.h.s. of (2) is the systematic price drift resulting from the possible imbalance between buyers and sellers. We assume that the function $`F(x)`$ is such that $`F(0)=0`$ and is monotonically increasing with its argument: perfect balance between buyers and sellers does not move the price; a larger (resp. smaller) number of buyers than sellers drive the price up (resp. down). An often used dependence is simply a linear relationship $`F(x)=\mu x`$. Assuming that the time needed to complete a trade of size $`L`$ is proportional to $`L`$ and that the unobservable price fluctuations obey a diffusion process during that time, Zhang (1999) has proposed instead $`F(x)\mathrm{sign}(x)\sqrt{|x|}`$. The second stochastic term of the r.h.s. of (2) accounts for noisy sources of price fluctuations. Taken alone, it would give the usual log-normal random walk process (Cootner, 1967).
### 2.2 Rational expectation investment strategy
At time $`t1`$, just when the price $`p(t1)`$ has been announced, the trader $`i`$ defines his strategy $`s_i(t1)`$ that he will hold from $`t1`$ to $`t`$, thus realizing the profit $`(p(t)p(t1))s_i(t1)`$. To define $`s_i(t1)`$, the trader calculates his expected profit $`P_E`$, given the past information and his position, and then chooses $`s_i(t1)`$ such that $`P_E`$ is maximum. Within the rational expectation model, all traders have full knowledge of the fundamental equation (2) of their financial world. However, they cannot poll the positions $`s_j`$ that will take all other traders which will determine the price drift according to (2). The next best thing that trader $`i`$ can do is to poll his $`N(i)`$ “neighbors” and construct his prediction for the price drift from this information. The trader needs an additional information, namely the a priori probability $`P_+`$ and $`P_{}`$ for each trader to buy or sell. The probabilities $`P_+`$ and $`P_{}`$ are the only informations that he can use for all the traders that he does not poll directly. From this, he can form his expectation of the price change. The simplest case corresponds to a neutral market where $`P_+=P_{}=1/2`$. To allow for a simple discussion, we restrict the discussion to the linear case $`F(x)=\mu x`$. The general nonlinear case complicates the matter and a careful convexity analysis must be performed. In the linear case, the trader $`i`$ expects the following price change
$$\mathrm{\Delta }t\mu \left(\frac{\underset{i=1}{\overset{N(i)}{}}s_i(t1)}{N}\right)+\sqrt{\mathrm{\Delta }t}\sigma \eta (t).$$
(3)
Notice that the sum is now restricted to the $`N(i)`$ neighbors of trader $`i`$ because the sum over all other traders, whom he cannot poll directly, averages out. This restricted sum is represented by the star symbol. His expected profit is thus
$$\left(\mathrm{\Delta }t\mu \left(\frac{\underset{i=1}{\overset{N(i)}{}}s_i(t1)}{N}\right)+\sqrt{\mathrm{\Delta }t}\sigma \eta (t)\right)p(t1)s_i(t1).$$
(4)
The strategy that maximizes his profit is
$$s_i(t1)=sign\left(\frac{\mu }{N}\underset{i=1}{\overset{N(i)_{}}{}}s_i(t1)+\frac{\sigma }{\sqrt{\mathrm{\Delta }t}}\eta (t)\right).$$
(5)
### 2.3 Stochastic dynamical model of interacting particles
Equation (5) is exactly the evolution equation postulated by Johansen, Sornette and Ledoit (1999) and by Johansen, Ledoit and Sornette (2000) for the dynamics of imitation between agents. This evolution equation (5) belongs to the class of stochastic dynamical models of interacting particles (Liggett, 1985, 1997), which have been much studied mathematically in the context of physics and biology. In this model (5), the tendency towards imitation is governed by $`\mu /N`$, which is called the coupling strength; the tendency towards idiosyncratic behavior is governed by $`\sigma `$. Thus the value of $`\mu /N`$ relative to $`\sigma `$ determines the outcome of the battle between order (imitation process) and disorder, and the development of a bubble. More generally, the coupling strength $`\mu /N`$ could be heterogeneous across pairs of neighbors, without substantially affecting the properties of the model. Some of the $`\mu _{ij}`$’s could even be negative, as long as the average of all $`\mu _{ij}`$’s was strictly positive. The equation (5) only describes the state of an agent at a given point in time. In the next instant, new $`\eta `$’s are drawn, new influences propagate themselves to neighbors, and agents can change states. Thus, the best we can do is give a statistical description of the states.
The model does not assume instantaneous opinion interactions between neighbours. In real markets, opinions tend indeed not to be instantaneous but are formed over a period of time by a process involving family, friends, colleagues, newspapers, web sites, TV stations, etc. Decisions about trading activity of a given agent may occur when the consensus from all these sources reaches a trigger level. This is precisely this feature of a threshold reached by a consensus that expression (5) captures: the consensus is described by the sum over the $`N(i)`$ agents connected to agent $`i`$ and the threshold is provided by the sign function together with the idiosyncratic signal included in $`\eta `$. The delay in the formation of the opinion of a given trader as a function of other traders’opinion is captured in our model by the progressive spreading of information during successive updating steps (see for instance (Liggett, 1985, 1997)).
The simplest possible network is a two-dimensional grid in the Euclidean plane. Each agent has four nearest neighbors: one to the North, one to the South, the East and the West. The relevant parameter is $`K\mu /\sigma `$. It measures the tendency towards imitation relative to the tendency towards idiosyncratic behavior. In the context of the alignment of atomic spins to create magnetisation, this model is identical to the so-called two-dimensional Ising model which has been solved explicitly by Onsager (1944). Only its formulation is different from what is usually found in textbooks (Goldenfeld, 1992), as we emphasize a dynamical view point.
In the Ising model, there exists a critical point $`K_c`$ that determines the properties of the system. When $`K<K_c`$, disorder reigns: the sensitivity to a small global influence is small, the clusters of agents who are in agreement remain of small size, and imitation only propagates between close neighbors. Formally, in this case, the susceptibility $`\chi `$ of the system is finite. When $`K`$ increases and gets close to $`K_c`$, order starts to appear: the system becomes extremely sensitive to a small global perturbation, agents who agree with each other form large clusters, and imitation propagates over long distances. In the Natural Sciences, these are the characteristics of so-called critical phenomena. Formally, in this case the susceptibility $`\chi `$ of the system goes to infinity. The hallmark of criticality is the power law, and indeed the susceptibility goes to infinity according to a power law:
$$\chi A(K_cK)^\gamma .$$
(6)
where $`A`$ is a positive constant and $`\gamma >0`$ is called the critical exponent of the susceptibility (equal to $`7/4`$ for the 2-d Ising model). This kind of critical behavior is found in many other models of interacting elements (Liggett, 1985, 1997). In the case of a network where all traders are connected to all other traders, one obtains the so-called “mean-field” critical behavior, for which $`\gamma =1`$.
We view a bubble as resulting from imitation processes similar to that described in this class of models, in which the “coupling” strength $`K`$ is larger than the critical value so that a globally bullish mood is established independently of the fundamentals. Another scenario is that $`K`$ itself evolves as a function of time, progressively increasing up and beyond the critical value $`K_c`$. Both scenarii lead to a bubble, i.e. to a price increase which is decoupled from the fundamentals.
## 3 The production of books reflects the spread of the bubble
In this section, we use the number of books published on a given topic as an observational probe of the mood and crazes of the society at a given moment; more specifically, we list the number of books on a given subject in the catalogues of major libraries in order to assess to what extent this issue is a matter of concern or interest. As will turn out, this measure provides a surprisingly good “thermometer”. The use of the terminoly “thermometer” is not just a way of speaking. The model presented in section 2 actually defines the temperature as $`1/K`$ in proper units. As we have seen, this parameter controls the strength of imitation between agents, which in this class of models is indeed controlled by the temperature. Before we use this thermometer for our purpose, it is appropriate to test it by performing a preliminary experiment.
### 3.1 Preliminary test: assessing inflation from the number of published books
Let us estimate the number of books published about inflation every year. To this aim we use the catalogues of two university libraries, namely: (i) The Harvard library which is a major general library; (ii) The library of the National Foundation for Political Sciences (FNSP, Paris), which is a library specialized in economics and political sciences. Furthermore, in order to test whether the results depend upon the counting procedure, we used two different criteria. At Harvard, we counted all the books whose titles contain the word “inflation”; at the FNSP, we counted the books listed under the “inflation” heading of the subject classification; in spite of the fact that this second criterion depends to some extent on the judgment of the librarian in charge of the subject index, it will be seen that both criteria lead to similar results. Fig.2 shows the two curves for the number of books in each library along with the curve of the inflation rate in the United States (dotted line). There is obviously a close connection between the three curves; the three cross-correlations are over 0.82. For instance, the correlation between the Harvard book data and the inflation rate is 0.89; the correlation between the Harvard data and the FNSP data is 0.87.
These observations lead to the following conclusions: (i) when inflation becomes higher, the concern about price increase seems to pervade large sections of the society: it makes publishers more prone to publish books on that topic, authors more willing to write them, librarians more ready to buy them and the public more desirous to read them. In fact the correlation is so good that the level of inflation could be measured fairly accurately (albeit with a time lag of one or two years) through the number of books published on that topic. (ii) Estimates of the number of books published on an issue of worldwide interest are fairly robust with respect to the country where the library is located, the type of the library (general versus specialized for instance) or the method used (e.g. whether based on a title or subject search).
In order to emphasize the significance of the previous result, the following comparison may be helpful. When a new experiment in the field of particle physics is carried out, one would certainly expect an increased production of papers on that topic. What we have observed in the case of inflation is fairly different however for at least two reasons. (i) The book/inflation relationship is not restricted to major inflation peaks; even modest upsurges in times of low inflation (under 5%) brought about an increased book production. (ii) The study of the inflation phenomenon is by no means facilitated in times of high inflation. On the contrary, from a scientific point of view, one would be in a much better position 5 or 10 years after an inflation peak for one could then study both the upward and downward phase of the inflation episode.
In short, Fig.2 shows that the society reacted with great sensitivity (so to say as a resonance chamber) to fluctuations of the price index. It is important to note that, in this case, one cannot argue that the increased book production stimulated inflation. On the contrary, in the case of speculative trading to which we turn in the next section, an augmented book production has a positive feedback effect on the economic phenomenon it addresses: it feeds speculation by sustaining the public’s interest.
### 3.2 Connection between speculative trading and production of books
In this paragraph, we propose three examples which point to a close connection between a speculative bubble in a given sector and the production of new books on that topic. All these experiments have been conducted on the electronic catalogues of the Harvard Library; it can be accessed through the instruction: telnet hollis.harvard.edu.
The first curve (solid line) in Fig.3a shows the number of books published yearly whose titles contain one of the words: “stocks”, “stock market” or “speculation”. The second curve (broken line) gives the level of stock prices. The connection between both phenomena is apparent especially during the peak of 1925-1932. The book production lags behind the price index, the time lag being of the order of 1.5 years, which is approximately the time it takes to write and publish a book; with this time lag taken into account, the correlation is 0.57; of course it would be substantially higher if the series were restricted to the 1925-1932 peak.
Fig.3b tells a similar story for a more recent time period, namely the 1980-1996 bull market. The correlation is 0.70 (a zero- or a one year-lag lead almost to the same correlation).
The last example concerns the diamond bubble of 1975-1984 (Fig.3c). There is a close connection between the number of books whose titles contain the words “diamond” or “diamonds” and the actual price of diamonds. The correlation (without lag) is 0.66.
The augmented book production is of course but one manifestation of the society’s increased interest for speculative trading. By its very nature, because of the time it takes to write and publish a book, this indicator displays considerable inertia. Needless to say it would be of interest to find an indicator responding more quickly to a change in the society’s mood. The next paragraph provides such an example.
### 3.3 The coin bubble of 1965
This example is taken from a very stimulating study by Montroll and Badger (1974, p.200). In the early 1960s, there was a speculative bubble for a number of American coins. For instance, a roll of fifty 1960 Denver-minted pennies of the small date variety which sold for 4 dollars in 1961 fetched 21 dollars in 1964; similarly a roll of 1955 half dollars went from 20 dollars to 190 dollars in that period. One of the main journals in the field of coin collecting is “Coin World” which has appeared weekly since 1959. Undoubtedly a weekly publication can reflect the speculative mood of the market more swiftly and accurately than newly published books. The solid line curve in Fig.4 shows the number of paid copies of “Coin World” that have been distributed. The curve began to level off by October 1964 and the maximum was reached on March 1, 1965. On the same figure, we plot the price history of the 1960 D small date penny and of the 1955 half dollar; these curves are smoothed least square fits to the weekly price fluctuations given in “Coin World”.
Once again there is a conspicuous parallelism and this time the distribution curve does not lag behind the price series; the leveling off of distribution even precedes the turning point of the half dollar market. In this case, the distribution data could have been used as a warning signal. Of course, for the penny price, the leveling off occurred even earlier, but the price of one coin can hardly be used to predict the price of another coin; in contrast the fluctuations in “Coin World” distribution are a counter of the number of people who have a general interest in the coin market. Note that the sharp peak versus flat trough asymmetry (Roehner and Sornette, 1998) have been found in a simple percolation model of the stock market (Stauffer and Sornette, 1999) by letting the trading activity be dependent on the price,i.e. increasing prices causes more people to act than a decreasing price.
It is of interest to consider for a moment actual price increases. The price of the 1960 D penny increased by a factor 5.2 (from 4 to 21 dollars), while the price of the 1955 half dollar was multiplied by 9.5 (from 20 to 120 dollars). This provides a new confirmation of the price multiplier effect (Roehner 2000) which says that for a more expensive item the speculation will be stronger and lead to a larger price increase.
What is the rationale of the effect described in this section or, in other words, why does the interest shown by the public for a given item reflect actual price changes during a speculative bubble? Basically each new person showing an interest in the market is a potential customer. When the number of interested people levels off, no new customer appears. Observation shows that the high level attained by the market can be sustained only for a short time with “old customers” buying from and selling to each other. Theoretically, these customers could also decide to devote a larger share of their revenue to buying coins; in this case, the bubble could continue to grow even with a fixed pool of customers. However, observation shows that this does not happen, at least in the cases that we discussed. It can be added that for an item as the 1960 D penny that costs only 20 dollars even at its peak price, it would be fairly easy for many collectors to buy more; nevertheless this does not happen: as soon as the number of new customers entering the market begins to level off, the bubble begins to falter. This very same mechanism has been documented by Galbraith (1997) in his famous analysis of the 1929 crash and speculative bubble preceeding it. More recently, Krugman (1994, 1995, 1998) has shown that the same mechanism has been at work in the speculative bubble preceeding the Asian collapse a few years ago.
Can the same investigation procedure be applied to the stock market? The most serious difficulty is to find a market journal which is representative of the level of public interest. The “Wall Street Journal” has been the most important newspaper in the securities field for many years; however, it has also achieved recognition as a national newspaper so that it is read for general information as well as for investment news. As a matter of fact, its average yearly circulation does not closely follow the fluctuation of stock prices. For instance, during the bear market of 1968-1974, its circulation increased from 1.1 million copies to 1.3 million, i.e. a 18% increase. That increase was however much smaller than the one that occurred during the 1960-1968 bull market, namely 72%. In short, the Wall Street Journal to some extent shows the phenomenon described above but because of its broad spectrum it cannot serve as an accurate thermometer.
## 4 Volume of trade as an indicator of the public’s interest
For stock markets, there is a vast literature on the (possible) relationship between trading volume and prices: e.g. Crouch (1970), Rogalski (1978), Schneller (1978), Karpoff (1987); Crouch’s contribution is particularly interesting. However, most of these papers are concerned with short-term (daily or weekly) changes; since both prices and volumes are fairly erratic (the latter being even more erratic than the former), this turns out to be a difficult statistical question. In this section, we are concerned with long term (yearly) variations. Moreover, we use the volume of sales as an indicator of the public’s interest for a given speculative item; in that sense, the present section is a logical continuation of our analysis in the previous section.
### 4.1 The stock market
Wall Street has several adages about trading volumes, for instance: “It takes volumes to make prices move” (cited in Schneller 1978) or “Transaction volume \[i.e. the number of shares traded\] tends to be high in bull markets and low in bear markets” (cited in Karpoff 1987, p.112). More precisely it will be seen that transaction volumes are high and growing in bull markets and low and decreasing in bear markets. Moreover, this conclusion holds not only for stock markets but also for other speculative markets, for instance the land and property markets.
The volume of shares traded on the NYSE between 1895 and 1940 is plotted in Fig.5a (solid line) along with the (deflated) Standard and Poor stock price index. Between 1895 ad 1914, there is a clear parallelism between volume and price (the correlation is 0.60). During the bull market of the 1920s, the yearly volume of shares traded increased from 0.2 billion to one billion; this five fold increase matched a multiplication by three of the price index (the correlation is 0.76). Similarly, Fig.5b shows that, during the bull market of the 1980s and 1990s, there was a graphic connection between trading volume and share prices: the volume increased from about 10 billions to more than 100 billions, a ten fold increase which matches an eightfold increase of the price index (the correlation is 0.98). The huge increase in the volume of transactions is a direct proof of the fact that there has been a permanent inflow of new customers into the market; that inflow is itself a manifestation of a widening public interest for stock investment. In this case, this effect was probably accompanied by an increase of the funds each customer was prepared to invest in the stock market. This is in particular suggested by the series given in the following table.
Table 1 Distribution of transactions by size (NYSE)
$$\begin{array}{ccccccccccc}& & 1975& 1976& 1977& 1978& 1979& 1980& 1981& 1982& 1983\\ \\ \text{100-4900 shares}& (\%)& 73.5& 70.2& 66.0& 64.1& 61.2& 56.9& 52.1& 44.7& 41.3\\ >\text{5000 shares}& (\%)& 26.6& 29.8& 34.0& 35.9& 38.8& 43.1& 47.9& 55.3& 58.7\\ & & & & & & & & & & \\ & & 1984& 1985& 1986& 1987& 1988& 1989& 1990& 1991& 1992\\ \\ \text{100-4900 shares}& (\%)& 36.4& 34.7& 36.5& 34.7& 32.4& 34.1& 35.2& 34.8& 33.5\\ >\text{5000 shares}& (\%)& 63.6& 65.3& 63.5& 65.3& 67.6& 65.9& 64.8& 65.2& 66.5\end{array}$$
Source: Statistical Abstract of the United States (various years)
Thus, the number of transactions of more than 5,000 shares (i.e. more than 50,000 dollars if one takes the price of a share to be of the order of 10 dollars) has increased steadily from 1975 on, which was the first year for which this statistics was given. But the rate of growth which has been very rapid between 1975 and 1984 has become much slower afterwards. It would be of great interest to know if the progression has continued after 1992, unfortunately these data are no longer given after this date.
### 4.2 Real estate markets
When turning from stock markets to real estate markets, one is confronted with two difficulties. The first is to identify a genuine “bull market”, the second is to find adequate statistics, which is not easy especially for the number of transactions. Fig.6a and b provide two examples. The first one concerns sales of land in Paris; the solid line shows the number of transactions, while the broken line gives the price per square meter; during the 1982-1991 bull market, both curves were parallel (correlation is 0.95). The second example concerns the sales of apartments in Paris shown in Fig.6b. The solid line shows the volume of sales in thousands of square meters, while the broken line shows the price per square meter. The story is somewhat different for this case. During the bull market phase, sales were at a high level but experienced only a slow increase; then in 1987, almost three years before the turning point for the prices, sales began to plummet. The volume of sales could have be used as a useful indicator of an incoming change of regime similarly to the case of the distribution of World Coin preceding the drop of the 1960 D penny. What is not obvious is to estimate a priori the time lag, i.e. how low the high price of apartments could be sustained in the presence of decreasing demand.
## 5 Summary
This paper has proposed new indicators for identifying speculative bubbles, defined as self-fulfilling reinforcing price changes. This work started with the recognition that qualifying the existence of speculative bubbles is controversial due in part to the need for a model of fundamental prices. Tests of the existence of a bubble then also become tests of the model and rejection cannot be attributed uniquely to the absence of a bubble but is linked to the quality of the model. In this standard approach, there is no unique answer. Here, we have presented a novel empirical method which bypasses these difficulties by monitoring external indicators of an anomalously growing interest in the public at times of bubbles. We have shown the existence, during the growth of the bubble, of a growing interest in the public for the commodity in question, whether it consists in stocks, diamonds or coins. This interest can be quantified for instance by the increase in the number of books published on the topic or in the increase in the subscriptions to specialized journals. We have also presented a simple model of rational expectation which maps exactly onto the Ising model on a random graph. This model allows us to view a bubble as a result of the fight between disorder (idiosyncratic signal which may be different for each trader) and order (resulting from an imitation or influence process that tend to align the opinions or interests of people). The bubble is found to be a regime of low “disorder”, i.e. low “temperature” for which the imitation processes are strong. The indicators can then be interpreted as “thermometers” of the system. This model provides a theoretical basis for spin-like models of influences discussed earlier in this context.
Acknowledgment We are grateful to Ms Claire de Buttet from the Paris bureau of the Wall Street Journal for kindly sending us the journal’s circulation data. DS acknowledges stimulating discussions with A. Johansen.
References
ANDERSSON (I.) 1973: Histoire de la Suède des origines à nos jours. Editions Horvath. Paris.
BENDER (C.), ORSZAG (S.A.) 1978: Advanced Mathematical Methods for Scientists and Engineers. McGraw-Hill, New York.
BHAMRA (H.S.) 1999: Imitation in financial markets. Working paper.
BLANCHARD (O.J.), WATSON (M.W.) 1982: Bubbles, Rational Expectations and Speculative Markets, in: Wachtel, P. ,eds., Crisis in Economic and Financial Structure: Bubbles, Bursts, and Shocks. Lexington Books: Lexington.
BODART (G.) 1908: Militär-historisches Kriegs-Lexikon (1618-1905). C.W. Stern. Vienna.
BOYAJIAN (W.E.) 1988: An economic review of the past decade in diamonds. Gems and Gemology 24,134-153.
COOTNER (P.H.) ed. 1967: The random character of stock market prices. Cambridge, Mass., M.I.T. Press.
CROUCH (R.L.) 1970: The volumes of transactions and price changes on the New York Stock Exchange. Financial Analysts Journal 26,4,104-109.
DEUTSCH (K.W.) 1953: Nationalism and social communication. MIT Press. Cambridge (Ma).
EPSTEIN (E.J.) 1982: The diamond invention. An exposé of the international diamond monopoly. Hutchinson. London.
FAY (S.) 1982: The great silver bubble. Hodder and Stoughton. London.
FLOOD (R.P.), GARBER (P.M.) 1994: Speculative bubbles, speculative attacks, and policy switching. The MIT Press, Cambridge, Massachussets.
GALBRAITH (J.K.) 1997: The great crash, 1929. Boston : Houghton Mifflin Co.
GOLDENFELD (N.) 1992: Lectures on phase transitions and the renormalization group. Addison-Wesley Publishing Company, Reading, Massachussets.
JOHANSEN (A.), SORNETTE (D.) 1998: Stock market crashes are outliers. European Physical Journal B 1, 141-143.
JOHANSEN (A.), SORNETTE (D.) 1999: Financial “antibubbles”: log-periodicity in gold and Nikkei collapses. International Journal of Modern Physics C 10,4,1-13.
JOHANSEN (A.), SORNETTE (D.), LEDOIT (O.) 1999: Financial Crashes using discrete scale invariance. Journal of Risk 1 (4), 5-32.
JOHANSEN (A.), LEDOIT (O.), SORNETTE (D.), 2000: Crashes as critical points. Int. J. Theor. Applied Finance 3, No 1 (January issue).
JOHANSEN (A.), SORNETTE (D.) 1999: Log-periodic power law bubbles in Latin-American and Asian markets and correlated anti-bubbles in Western stock markets: An empirical study. submitted to the Journal of Empirical Finance (http://xxx.lanl.gov/abs/cond-mat/9907270
KARPOFF (J.M.) 1897: The relation between price changes and trading volume: a survey. Journal of Financial and Quantitative Analysis 22, 1, 109-126.
KRUGMAN (P.) 1994: The myth of Asia’s miracle. Foreign Affairs 73, n6 (Nov/Dec): 62-78.
KRUGMAN (P.) 1995: Dutch tulips and emerging markets. Foreign Affairs 74, n4 (Jul/Aug): 28-44.
KRUGMAN (P.) 1998: Asia: What went wrong. Fortune 137, n4 (Mar 2):32-34.
LIESNER (T.) 1989: One hundred years of economic statistics. Facts on File. New York.
LIGGETT (T.M.) 1985: Interacting particle systems. New York : Springer-Verlag.
LIGGETT (T.M.) 1997: Stochastic models of interacting systems. The Annals of Probability 25, 1-29.
LUX (T.), SORNETTE (D.) 1999: On Rational Bubbles and Fat Tails, submitted to the Journal of Monetary Economics. (http://xxx.lanl.gov/abs/cond-mat/9910141)
MONTROLL (E.W.), BADGER (W.W.) 1974: Introduction to quantitative aspects of social phenomena. Gordon and Breach. New York.
ONSAGER (L.) 1944: Physics Review 65, 117.
ROEHNER (B.M.) 1995: Theory of markets. Trade and space-time patterns of price fluctuations. A study in analytical economics. Springer-Verlag. Berlin.
ROEHNER (B.M.) 2000: Speculative trading: the price multiplier effect. European Physical Journal B (to appear).
ROEHNER (B.M.), SORNETTE (D.) 1998: The sharp peak-flat trough pattern and critical speculation. European Physical Journal B 4, 387-399.
ROGALSKI (R.J.) 1978: The relation between price changes and trading volumes: a survey. Journal of Financial and Quantitative Analysis 22,1,109-126.
SCHNELLER (M.I.) 1978: Security price changes and transaction volumes. Comment. American Economic Review 68,4,696-697.
SCOTT (F.D.) 1977: Sweden. The nation’s history. University of Minnesota Press. Minneapolis.
SORNETTE (D.), JOHANSEN (A.) 1997: Large financial crashes. Physica A 245,411.
STANLEY (H.E.) 1987: Introduction to Phase Transitions and Critical Phenomena. Oxford University Press, New York.
STAUFFER (D.), SORNETTE (D.) 1999: Self-Organized Percolation Model for Stock Market Fluctuations. Physica A 271, 496-506.
WATTS (D.J.) 1999: Small worlds: the dynamics of networks between order and randomness. Princeton, N.J.: Princeton University Press.
ZHANG (Y.C.) 1999: Toward a theory of marginally efficient markets. Physica A 269, 30-44.
Figure captions
Fig.1 Price bubble for diamonds, gold, platinum and silver. For gold, silver and platinum, the prices are in dollar per ounce. For diamonds, the price is for a one-carat G flawless diamond. The collapse occurred almost simultaneously in the four markets. Prices of palladium and cobalt showed a similar evolution, we did not represent them for the sake of clarity. Sources: Journal des Finances (26 October 1978); The Economist (5 April 1980); Diamond 1988: Special Report of the Economist
Fig.2 Comparison between the number of yearly published books about inflation and the inflation rate. Thick solid line: books at Harvard library whose titles contain the word “inflation”. Thin solid line: number of books under the “inflation” heading of a subject classification (library of the National Foundation for Political Science, Paris). Dotted line: Consumer price index in the United States. The inflation curve has been shifted by one year to the right. The correlation between the number of books published and the (shifted) inflation series is comprised between 0.82 (FNSP) and 0.89 (Harvard). Sources: Liesner (1989); Roehner (1995, p.381).
Fig.3a Comparison between the number of yearly published books about stock market speculation and the level of stock prices (1911-1940). Solid line: books at Harvard library whose titles contain one of the words “stocks”, “stock market” or “speculation”; broken line: Standard and Poor index of common stocks. The curve of published books lags behind the price curve with a time-lag of about 1.5 years. Source: The stock price index is taken from the Historical Abstract of the United States.
Fig.3b Comparison between the number of yearly published books about stock market speculation and the level of stock prices (1971-1996). Solid line: books at Harvard library whose titles contain one of the words “stocks”, “stock market” or “speculation”; broken line: Dow Jones index of industrial shares. Source: The stock price index is taken from Quid (1997).
Fig.3c Comparison between the number of yearly published books about diamonds and the price of diamonds (1975-1984). Solid line: books at Harvard library whose titles contain one of the words “diamond” or “diamonds”. Source: The diamond prices are taken from Diamond 1988 (Special Report of the Economist)
Fig.4 Comparison between the distribution of “Coin World” and the price of two American coins (1962-1970). The curve labeled “Penny” refers to the price of a roll of 1960 D small date pennies; the curve labeled “Half dollar” refers to the price of a roll of 1955 half dollars. Source: Montroll and Badger (1974)
Fig.5a Comparison between the number of shares traded on the NYSE and the level of stock prices (1897-1940). Solid line: number of shares traded; broken line: deflated Standard and Poor’s index of common stocks. Source: Historical Statistics of the United States
Fig.5b Comparison between the number of shares traded on the NYSE and the level of stock prices (1980-1997). Solid line: number of shares traded; broken line: Dow Jones index (industrials). Source: Statistical Abstract of the United States, various years
Fig.6a Comparison between the number of transactions and the price of land in Paris. Solid line: number of transactions; broken line: price per square meter in thousands of French francs. Source: Le marché immobilier français. Les chiffres et les sources (1993).
Fig.6b Comparison between sales and prices of apartments in Paris. Solid line: Sales in thousand of square meters; broken line: price of apartments per square meter in thousands of French francs (of 1990). These data refer to “old” apartments i.e. apartments which have already been sold at least once. Source: Le marché immobilier français. Les chiffres et les sources (1998). |
warning/0001/hep-th0001189.html | ar5iv | text | # Untitled Document
hep-th/0001189 DUK-CGTP-00-03, IASSNS–HEP–00/118
Invariance Theorems for Supersymmetric Yang-Mills Theories
Savdeep Sethi<sup>1</sup> sethi@sns.ias.edu and Mark Stern<sup>2</sup> stern@math.duke.edu
$``$ School of Natural Sciences, Institute for Advanced Study, Princeton, NJ 08540, USA
$``$ Department of Mathematics, Duke University, Durham, NC 27706, USA
We consider quantum mechanical Yang-Mills theories with eight supercharges and a $`Spin(5)\times SU(2)_R`$ flavor symmetry. We show that all normalizable ground states in these gauge theories are invariant under this flavor symmetry. This includes, as a special case, all bound states of D0-branes and D4-branes. As a consequence, all bound states of D0-branes are invariant under the $`Spin(9)`$ flavor symmetry. When combined with index results, this implies that the bound state of two D0-branes is unique.
1/00
1. Introduction
The existence of normalizable vacua in supersymetric Yang-Mills theories is a question that arises in many different contexts in string theory and field theory. Index arguments can be used to determine whether any vacua exist, but not exactly how many vacua. An index only counts the difference between the number of bosonic and fermionic vacua. To count the actual number of vacua, we need more information such as how the vacua transform under the global symmetries of the theory.
In this paper, we consider quantum mechanical Yang-Mills theories with eight supercharges and an $`Spin(5)\times SU(2)_R`$ symmetry. We take our theories to be dimensional reductions of $`d=6`$ N=1 Yang-Mills theories coupled to matter. The question of normalizable ground states in these models arises in the study of bound states of D0-branes and D4-branes \[1,,2\]; for example, a single D0-brane and a single D4-brane can be shown to bind using $`L^2`$ index arguments generalized to theories without a gap. Other examples from string theory involve D0-branes moving on orbifolds , and the question of counting H-monopoles in the heterotic string \[5,,3\].
In the following section, we describe the field content and symmetries of these gauge theories. We then show that all normalizable ground states in these theories must be invariant under the $`SU(2)_R`$ symmetry. The argument we give is suggested by recent work on the $`L^2`$-cohomology of hyperKähler spaces by Hitchin . Our result should have implications for defining and computing the $`L^2`$-cohomology of instanton moduli spaces. Certain instanton moduli spaces appear as Higgs branches in gauge theories of the kind under consideration. For example, the moduli space of $`U(N)`$ instantons in $`\mathrm{IR}^4`$ appears as the Higgs branch of the quantum mechanics describing D0-D4 systems. Although these spaces can be singular, their embedding into quantum mechanical gauge theory provides a natural regularization of the singularities. Heuristically, the wavefunction for a state corresponding to a form on the Higgs branch is smoothed out by leaking onto the Coulomb branch. It would be interesting to explore this connection further.
There is a second $`R`$-symmetry in these theories which comes from the dimensional reduction of the Lorentz group. For reductions of $`d=6`$ N=1 Yang-Mills theories, this is a $`Spin(5)`$ symmetry. Using basically the same argument as in the case of the $`SU(2)_R`$ symmetry, we show that all normalizable ground states in these theories are invariant under this $`Spin(5)`$ symmetry. For reductions of $`d=10`$ N=1 Yang-Mills theories , the $`R`$-symmetry group is $`Spin(9)`$. It is quite straightforward to argue that as a consequence of the $`SU(2)_R\times Spin(5)`$ invariance theorem, all ground states in these theories with sixteen supercharges must be invariant under the $`Spin(9)`$ symmetry.
We can couple these invariance theorems with results from $`L^2`$ index theory \[8,,9\]. The $`L^2`$ index for the non-Fredholm theory<sup>1</sup> By non-Fredholm, we mean a theory without a gap. of two D0-branes is proven to be one . We also know that the $`L^2`$ index for the theory of a single D0-brane and a single D4-brane is one . Our invariance results imply that all bound states in these theories are bosonic, and therefore unique. These results can also be combined with other interesting but heuristic attempts to study the $`L^2`$ index by either deforming the Yang-Mills theory \[10,,11\], or by using insights from string theory to compute the bulk and defect terms. The bulk terms for various Yang-Mills theories have been directly computed in \[13,,14,,15\]. There have also been a number of comments on the implications of invariance for the asymptotic form of particular bound state wavefunctions \[16,,17\].
2. The Field Content and Symmetries
2.1. The vector multiplet supercharge
The argument we wish to make requires reasonably little explicit knowledge of the gauge theory. There is a $`Spin(5)\times SU(2)_R`$ symmetry which commutes with the Hamiltonian $`H`$. Since we are considering a gauge theory, we must have at least one vector multiplet. It contains five scalars $`x^\mu `$ with $`\mu =1,\mathrm{},5`$ transforming in the $`(\mathrm{𝟓},\mathrm{𝟏})`$ of the symmetry group. These scalars transform in the adjoint representation of the gauge group $`G`$. Let $`p^\mu `$ be the associated canonical momenta obeying,
$$[x_A^\mu ,p_B^\nu ]=i\delta ^{\mu \nu }\delta _{AB},$$
$`(2.1)`$
where the subscript $`A`$ is a group index.
Associated to these bosons are eight real fermions $`\lambda _a`$ where $`a=1,\mathrm{},8`$ transforming in the $`(\mathrm{𝟒},\mathrm{𝟐})`$ representation of the symmetry group. These fermions are also in the adjoint representation of the gauge group. The eight supercharges also transform in the $`(\mathrm{𝟒},\mathrm{𝟐})`$ representation. These fermions obey the usual quantization relation,
$$\{\lambda _{aA},\lambda _{bB}\}=\delta _{ab}\delta _{AB}.$$
$`(2.2)`$
Let $`\gamma ^\mu `$ be hermitian real gamma matrices which obey,
$$\{\gamma ^\mu ,\gamma ^\nu \}=2\delta ^{\mu \nu }.$$
$`(2.3)`$
Appendix A includes an explicit basis for these gamma matrices along with a discussion of the symmetry group action.
The supercharge takes the form,
$$Q_a^v=\left(\gamma ^\mu p_A^\mu \lambda _A\right)_a+\frac{1}{2}f_{ABC}\left(\gamma ^{\mu \nu }\lambda _Ax_B^\mu x_C^\nu \right)_a+D_{abA}\lambda _{bA},$$
$`(2.4)`$
where $`f_{ABC}`$ are the structure constants and $`\gamma ^{\mu \nu }=(1/2)(\gamma ^\mu \gamma ^\nu \gamma ^\nu \gamma ^\mu ).`$ The real anti-symmetric matrix $`D`$ does not involve momenta. The $`D`$-term transforms in the $`(\mathrm{𝟏},\mathrm{𝟑})`$ representation of the symmetry group, and in the adjoint representation of the gauge group. The precise form of $`D`$ is not important for our argument. In general, there can be many vector multiplets. In that case, the terms in the supercharge (2.4) generalize in an obvious way.
2.2. The hypermultiplet supercharge
A hypermultiplet contains four real scalars which we can package into a quaternion $`q`$ with components $`q^i`$ where $`i=1,2,3,4`$. This field transforms as $`(\mathrm{𝟏},\mathrm{𝟐})`$ under the symmetry group, and in some representation $`T`$ of the gauge group. We again introduce canonical momenta $`p_i`$ satisfying the usual commutation relations. Now $`SU(2)_RSp(1)_R`$ is the group of unit quaternions. We choose $`SU(2)_R`$ to act on a hypermultiplet $`q`$ by right multiplication by a unit quaternion. The gauge symmetry commutes with the $`SU(2)_R`$ symmetry and acts by left multiplication on $`q`$. See Appendix A for a more detailed discussion.
The superpartner to $`q`$ is a real fermion $`\psi _a`$ with $`a=1,\mathrm{},8`$ satisfying,
$$\{\psi _a^R,\psi _{bS}\}=\delta _{ab}\delta _S^R.$$
$`(2.5)`$
These fermions transform in the $`(\mathrm{𝟒},\mathrm{𝟏})`$ representation, and the $`R,S`$ subscripts index the $`T`$ representation of $`G`$. For $`n`$ hypermultiplets, the gauge group $`G`$ acts via a subgroup of the $`Sp(n)_L`$ symmetry. In terms of the $`s^j`$ operators given in Appendix A, the hypermultiplet charge takes the form
$$Q_a^h=s_{ab}^j\psi _bp_j+I_{ab}\psi _b.$$
$`(2.6)`$
We have lumped all the interactions into the non-derivative operator $`I`$ which transforms in the $`\mathrm{𝟐}`$ of $`SU(2)_R`$. We also need to note that $`I`$ is proportional to $`x^\mu \gamma ^\mu `$ with a proportionality constant that commutes with the $`Spin(5)`$ generators. We have also suppressed gauge indices. Note that since the $`s^j`$ implement right multiplication by a quaternion, they commute with $`\gamma ^\mu `$. Again, there can be many hypermultiplets in different representations of the gauge group. In that case, the hypermultiplet supercharge (2.6) generalizes in a straightforward way. The full Hermitian supercharge is the sum of the vector and hypermultiplet supercharges,
$$Q_a=Q_a^v+Q_a^h.$$
2.3. The $`SU(2)_R`$ currents
The three generators of $`SU(2)_R`$ correspond to right multiplication by $`I,J,K`$ and are given in terms of the gauge invariant rotation generators,
$$W_{ij}=q_ip_jq_jp_i.$$
$`(2.7)`$
Again here and in the subsequent discussion, we generally suppress gauge indices. In accord with prior notation, we denote the three $`SU(2)_R`$ generators by $`\stackrel{~}{s}^i`$:
$$\begin{array}{cc}\hfill \stackrel{~}{s}^2& =W_{12}W_{34}+\frac{i}{2}\lambda s^2\lambda \hfill \\ \hfill \stackrel{~}{s}^3& =W_{13}+W_{24}+\frac{i}{2}\lambda s^3\lambda \hfill \\ \hfill \stackrel{~}{s}^4& =W_{14}W_{23}+\frac{i}{2}\lambda s^4\lambda .\hfill \end{array}$$
$`(2.8)`$
As they should, these generators act on the bosons of the hypermultiplet and the fermions of the vector multiplet. Adding either more vector multiplets or more hypermultiplets is straightforward: we simply need to sum the contributions to the three currents (2.8) from each multiplet.
2.4. The $`Spin(5)`$ currents
The ten generators of $`Spin(5)`$ act on the bosons of the vector multiplet and all fermions in the problem. The generators are given by:
$$T^{\mu \nu }=x^\mu p^\nu x^\nu p^\mu \frac{i}{4}\gamma _{ab}^{\mu \nu }\left(\lambda _a\lambda _b+\psi _a\psi _b\right).$$
$`(2.9)`$
Adding either more vector multiplets or more hypermultiplets is again straightforward.
3. An Invariance Argument for the $`SU(2)_R`$ Symmetry
3.1. Relating the $`SU(2)_R`$ currents to the supercharge
A key point in the argument is a relation between the supercharge and the $`SU(2)_R`$ currents. For some choice of $`v_a^i`$, we want to show that:
$$\stackrel{~}{s}^i=\underset{a}{}\{Q_a,v_a^i\}.$$
$`(3.1)`$
Let us start with the vector multiplet. We take a candidate gauge singlet,
$$(v_1)_a^i=\left(s^i\gamma ^\nu \lambda \right)_ax^\nu .$$
$`(3.2)`$
First note that this choice anti-commutes with $`Q^h`$ because $`\lambda `$ anti-commutes with $`\psi `$. It also anti-commutes with the $`D`$-term in (2.4). To see this, we compute:
$$\begin{array}{cc}\hfill \underset{a}{}\{D_{ab}\lambda _b,(v_1)_a^i\}& =x_A^\nu \mathrm{tr}\left(s^i\gamma ^\nu D_A^T\right),\hfill \end{array}$$
$`(3.3)`$
However, we can immediately see that (3.3) vanishes by noting that the operator $`s^i\gamma ^\nu D^T`$ does not contain a singlet under $`Spin(5)`$. The trace of the operator therefore vanishes. Our choice for $`v_1`$ anti-commutes with $`\frac{1}{2}f_{ABC}\left(\gamma ^{\mu \nu }\lambda _Ax_B^\mu x_C^\nu \right)_a`$ for the same reason: the resulting trace does not contain a singlet of $`Spin(5)`$.
What remains is the following anti-commutator which is not hard to compute,
$$\begin{array}{cc}\hfill \underset{a}{}\{\left(\gamma ^\mu p^\mu \lambda \right)_a,(v_1)_a^i\}& i\lambda s^i\lambda .\hfill \end{array}$$
$`(3.4)`$
The exact proportionality constant does not matter for this argument. The important point is that we can use (3.2) to generate the terms in the $`SU(2)_R`$ currents which act on vector multiplets.
For the hypermultiplet, we take the following candidate gauge singlet:
$$(v_2)_a^i=\left(s^is^l\psi \right)_aq^l.$$
$`(3.5)`$
Note that $`v_2`$ anti-commutes with $`Q^v`$ because $`\lambda `$ anti-commutes with $`\psi `$. It is also not too hard to argue that the anti-commutator of $`v_2`$ with the interaction term $`I`$ in (2.6) must vanish. We see that,
$$\begin{array}{cc}\hfill \underset{a}{}\{I_{ab}\psi _b,(v_2)_a^i\}& q^l\mathrm{tr}\left(s^is^lI\right),\hfill \end{array}$$
$`(3.6)`$
but $`s^is^lI`$ does not contain a singlet under the $`Spin(5)`$ action on fermions because $`I`$ is proportional to $`\gamma ^\mu `$ so the trace vanishes.
Again what remains is the anti-commutator,
$$\underset{a}{}\{s_{ab}^j\psi _bp_j,(v_2)_a^i\}.$$
$`(3.7)`$
It is easy to check that the $`\psi \psi `$ terms in the anti-commutator vanish because,
$$\underset{k}{}\psi \{s^k\}^Ts^is^k\psi =0.$$
With a little additional work, we find that (3.7) gives precisely the bosonic terms in (2.8) up to an overall non-vanishing constant. We therefore conclude that for appropriately chosen constants $`\alpha _1`$ and $`\alpha _2`$, the choice
$$v_a^i=\alpha _1(v_1)_a^i+\alpha _2(v_2)_a^i$$
$`(3.8)`$
satisfies (3.1).
3.2. Rotating a ground state
We assume there exists a normalizable ground state $`\mathrm{\Psi }`$ which is not a singlet under $`SU(2)_R`$. Under some $`SU(2)_R`$ rotation, we obtain another non-trivial $`L^2`$ zero-energy state. What does $`L^2`$ imply? Let us collectively denote all the bosonic coordinates $`x`$ and $`q`$ by $`y^i`$ where $`i=1,\mathrm{},D`$. Normalizability requires that,
$$<\mathrm{\Psi },\mathrm{\Psi }>=d^Dy\mathrm{\Psi }^{}(y^i)\mathrm{\Psi }(y^i)<\mathrm{}.$$
For some $`\stackrel{~}{s}^i`$, the state $`\stackrel{~}{s}^i\mathrm{\Psi }`$ is a non-trivial ground state. It satisfies the relation,
$$Q_a\left(\stackrel{~}{s}^i\mathrm{\Psi }\right)=Q_a\mathrm{\Psi }=0,$$
$`(3.9)`$
for each $`a`$ by definition of a ground state. Using (3.1), we find that
$$\begin{array}{cc}\hfill \stackrel{~}{s}^i\mathrm{\Psi }& =\underset{a}{}\{Q_a,v_a^i\}\mathrm{\Psi },\hfill \\ & =\underset{a}{}Q_a\left(v_a^i\mathrm{\Psi }\right).\hfill \end{array}$$
$`(3.10)`$
The new ground state looks $`Q`$-trivial. To show that it really is physically trivial, we need to check that it has zero norm. Since $`Q`$ is Hermitian and kills $`\stackrel{~}{s}^i\mathrm{\Psi }`$, the norm of $`\stackrel{~}{s}^i\mathrm{\Psi }`$ vanishes if we can integrate by parts. To integrate by parts, we argue as in Jost and Zuo \[18,,6\]: in terms of $`y=|y^i|`$, we can cutoff of the integral using a smooth bump function $`\rho _R(y)`$ which vanishes for $`y>2R`$, satisfies $`|d\rho _R|<4/R`$ and is one for $`y<R`$,
$$<\stackrel{~}{s}^i\mathrm{\Psi },\stackrel{~}{s}^i\mathrm{\Psi }>=\underset{R\mathrm{}}{lim}<\rho _R(y)\stackrel{~}{s}^i\mathrm{\Psi },\stackrel{~}{s}^i\mathrm{\Psi }>.$$
Using (3.9) and (3.10), we see that
$$\begin{array}{cc}\hfill <\stackrel{~}{s}^i\mathrm{\Psi },\stackrel{~}{s}^i\mathrm{\Psi }>& =\underset{R\mathrm{}}{lim}<\rho _R(y)\stackrel{~}{s}^i\mathrm{\Psi },\underset{a}{}\{Q_a,v_a^i\}\mathrm{\Psi }>,\hfill \\ & =\underset{R\mathrm{}}{lim}\underset{a}{}<[Q_a,\rho _R(y)]\stackrel{~}{s}^i\mathrm{\Psi },v_a^i\mathrm{\Psi }>.\hfill \end{array}$$
$`(3.11)`$
We see that $`[Q_a,\rho _R(y)]`$ is $`O(1/y)`$ and vanishes for $`y<R`$ and $`y>2R`$. Since $`v_i^a`$ is $`O(y)`$ at worst, the right hand side of (3.11) vanishes. The $`SU(2)_R`$ symmetry therefore acts trivially on all normalizable ground states.
4. Invariance Under the $`Spin(5)`$ Symmetry
4.1. Relating the $`Spin(5)`$ currents to the supercharge
We want to use essentially the same argument as in the $`SU(2)_R`$ case. For some choice of $`v_a^{\mu \nu }`$, we want to show that:
$$T^{\mu \nu }=\underset{a}{}\{Q_a,v_a^{\mu \nu }\}.$$
$`(4.1)`$
Let us start with the vector multiplet. We take a candidate gauge singlet,
$$(v_1)_a^{\mu \nu }=\left\{\gamma ^\mu x^\nu \gamma ^\nu x^\mu \right\}_{ab}\lambda _b.$$
$`(4.2)`$
Again this choice anti-commutes with $`Q^h`$ because $`\lambda `$ anti-commutes with $`\psi `$. The anti-commutator with $`\frac{1}{2}f_{ABC}\left(\gamma ^{\mu \nu }\lambda _Ax_B^\mu x_C^\nu \right)_a`$ results in a trace of three gamma matrices and so vanishes. It also anti-commutes with the $`D`$-term in (2.4). To see this, we compute:
$$\begin{array}{cc}\hfill \underset{a}{}\{D_{ab}\lambda _b,(v_1)_a^{\mu \nu }\}& =D_{ab}\left\{\gamma ^\mu x^\nu \gamma ^\nu x^\mu \right\}_{ab}.\hfill \end{array}$$
$`(4.3)`$
However, this combination does not contain a singlet under $`Spin(5)`$ so (4.3) vanishes.
We are left with the following anti-commutator which we need to compute quite carefully,
$$\begin{array}{cc}\hfill \underset{a}{}\{\left(\gamma ^\mu p^\mu \lambda \right)_a,(v_1)_a^{\mu \nu }\}& =8\left(x^\nu p^\mu x^\mu p^\nu \right)+2i\lambda \gamma ^{\mu \nu }\lambda .\hfill \end{array}$$
$`(4.4)`$
This computation is sensitive to the size of the $`\gamma `$ matrix. We obtain precisely the right ratio between the bosonic and fermion terms in (4.4) because the theory is reduced from six dimensions. We would not obtain the right ratio had we considered a theory reduced from ten dimensions with a $`Spin(9)`$ symmetry. Again, we can use (4.2) to generate the terms in the $`Spin(5)`$ currents which act on vector multiplets.
For the hypermultiplet, we take the following choice:
$$(v_2)_a^{\mu \nu }=\left(\gamma ^{\mu \nu }s^i\psi \right)_aq^i.$$
$`(4.5)`$
Again $`v_2`$ anti-commutes with $`Q^v`$ because $`\lambda `$ anti-commutes with $`\psi `$. In much the same way as before, we can argue that the anti-commutator of $`v_2`$ with the interaction term $`I`$ in (2.6) must vanish. We see that,
$$\begin{array}{cc}\hfill \underset{a}{}\{I_{ab}\psi _b,(v_2)_a^{\mu \nu }\}& q^i\mathrm{tr}\left(I\gamma ^{\mu \nu }s^i\right),\hfill \end{array}$$
$`(4.6)`$
but $`I\gamma ^{\mu \nu }s^i`$ again does not contain a singlet under $`Spin(5)`$ so the trace vanishes.
The remaining anti-commutator involves the kinetic term in the hypermultiplet charge,
$$\underset{a}{}\{s_{ab}^j\psi _bp_j,(v_2)_a^{\mu \nu }\}=i\psi \gamma ^{\mu \nu }\psi .$$
$`(4.7)`$
Again we conclude that for appropriately chosen constants $`\alpha _1`$ and $`\alpha _2`$, the choice
$$v_a^{\mu \nu }=\alpha _1(v_1)_a^{\mu \nu }+\alpha _2(v_2)_a^{\mu \nu }$$
$`(4.8)`$
satisfies (4.1). A straightforward repeat of the argument given in section 3.2 then implies that the $`Spin(5)`$ symmetry acts trivially on all normalizable ground states.
4.2. Theories with sixteen supercharges
For theories obtained by reduction from ten dimensions, the previous argument does not apply directly to the $`Spin(9)`$ symmetry for reasons mentioned earlier. These theories contain scalars $`y^i`$ where $`i=1,\mathrm{},9`$ transforming in the adjoint representation of the gauge group. The superpartners to these scalars are real fermions $`\eta _\alpha `$ where $`\alpha =1,\mathrm{},16`$ also in the adjoint representation.
However, we can always view these theories as special cases of theories with eight supercharges. We choose any $`5`$ of the $`9`$ scalars $`y^i`$ to be the vector multiplet, and the remaining $`4`$ scalars comprise an adjoint hypermultiplet. Of the original $`Spin(9)`$ symmetry, only a $`Spin(5)\times SU(2)_L\times SU(2)_R`$ subgroup is manifest. The scalars decompose in the following way,
$$\mathrm{𝟗}(\mathrm{𝟓},\mathrm{𝟏},\mathrm{𝟏})(\mathrm{𝟏},\mathrm{𝟐},\mathrm{𝟐}).$$
$`(4.9)`$
The fermions decompose according to,
$$\mathrm{𝟏𝟔}(\mathrm{𝟒},\mathrm{𝟏},\mathrm{𝟐})(\mathrm{𝟒},\mathrm{𝟐},\mathrm{𝟏}).$$
$`(4.10)`$
Our invariance argument implies that all normalizable ground states are invariant under the $`Spin(5)\times SU(2)_R`$ symmetry. However, this is true regardless of how we embed $`Spin(5)\times SU(2)_R`$ into $`Spin(9)`$. This is only possible if the full $`Spin(9)`$ symmetry acts trivially on all normalizable ground states.
Acknowledgements
The work of S.S. is supported by the William Keck Foundation and by NSF grant PHY–9513835; that of M.S. by NSF grant DMS–9870161.
Appendix A. Quaternions and Symplectic Groups
We will summarize some useful relations between quaternions and symplectic groups. Let us label a basis for our quaternions by $`\{\mathrm{𝟏},I,J,K\}`$ where,
$$I^2=J^2=K^2=\mathrm{𝟏},IJK=\mathrm{𝟏}.$$
A quaternion $`q`$ can then be expanded in components
$$q=q^1+Iq^2+Jq^3+Kq^4.$$
The conjugate quaternion $`\overline{q}`$ has an expansion
$$q=q^1Iq^2Jq^3Kq^4.$$
The symmetry group $`Sp(1)_RSU(2)_R`$ is the group of unit quaternions. Let $`\mathrm{\Lambda }`$ be a field transforming in the $`\mathrm{𝟐}`$ of $`Sp(1)_R`$. If we view $`Sp(1)_R`$ acting on $`\mathrm{\Lambda }`$ as right multiplication by a unit quaternion $`g`$ then,
$$\mathrm{\Lambda }\mathrm{\Lambda }g.$$
In this formalism, $`\mathrm{\Lambda }`$ is valued in the quaternions. Equivalently, we can expand $`\mathrm{\Lambda }`$ in components and express the action of $`g`$ in the following way,
$$\mathrm{\Lambda }_ag_{ab}\mathrm{\Lambda }_b,$$
where $`g_{ab}`$ implements right multiplication by the unit quaternion $`g`$. For example, right multiplication by $`I`$ on $`q`$ gives
$$\begin{array}{cc}\hfill q& qI\hfill \\ & q^1Iq^2q^3K+q^4J,\hfill \end{array}$$
which can be realized by the matrix
$$I^R=\left(\begin{array}{cccc}0& 1& 0& 0\\ 1& 0& 0& 0\\ 0& 0& 0& 1\\ 0& 0& 1& 0\end{array}\right)$$
$`(\text{A.}1)`$
acting on $`q`$ in the usual way $`q_aI_{ab}^Rq_b`$. The matrices $`J^R`$ and $`K^R`$ realize right multiplication by $`J,K`$ while $`\mathrm{𝟏}^R`$ is the identity matrix:
$$J^R=\left(\begin{array}{cccc}0& 0& 1& 0\\ 0& 0& 0& 1\\ 1& 0& 0& 0\\ 0& 1& 0& 0\end{array}\right),K^R=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 0\\ 0& 1& 0& 0\\ 1& 0& 0& 0\end{array}\right).$$
$`(\text{A.}2)`$
We define operators $`s^j`$ in terms of $`\{\mathrm{𝟏}^R,I^R,J^R,K^R\}`$
$$s^1=\left(\begin{array}{cc}\mathrm{𝟏}^R& 0\\ 0& \mathrm{𝟏}^R\end{array}\right),s^2=\left(\begin{array}{cc}I^R& 0\\ 0& I^R\end{array}\right),s^3=\left(\begin{array}{cc}J^R& 0\\ 0& J^R\end{array}\right),s^4=\left(\begin{array}{cc}K^R& 0\\ 0& K^R\end{array}\right).$$
In a similar way, the group $`Sp(2)Spin(5)`$ is the group of quaternion-valued $`2\times 2`$ matrices with unit determinant. We will view $`Sp(2)`$ as acting by left multiplication on a field $`\mathrm{\Psi }`$ in the defining representation. So an element $`USp(2)`$ acts in the following way:
$$\mathrm{\Psi }U\mathrm{\Psi }.$$
Equivalently, in terms of components
$$\mathrm{\Psi }_aU_{ab}\mathrm{\Psi }_b.$$
Lastly, we can give an explicit form for the gamma matrices (2.3) in terms of quaternions:
$$\gamma ^1=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right),\gamma ^2=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right),\gamma ^3=\left(\begin{array}{cc}0& I\\ I& 0\end{array}\right)$$
$$\gamma ^4=\left(\begin{array}{cc}0& J\\ J& 0\end{array}\right),\gamma ^5=\left(\begin{array}{cc}0& K\\ K& 0\end{array}\right).$$
In turn, $`\{I,J,K\}`$ can be expressed in terms of the Pauli matrices $`\sigma ^i`$
$$\sigma ^1=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right),\sigma ^2=\left(\begin{array}{cc}0& i\\ i& 0\end{array}\right),\sigma ^3=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)$$
as $`4\times 4`$ real anti-symmetric matrices:
$$I=\left(\begin{array}{cc}0& \sigma ^1\\ \sigma ^1& 0\end{array}\right),J=\left(\begin{array}{cc}i\sigma ^2& 0\\ 0& i\sigma ^2\end{array}\right),K=\left(\begin{array}{cc}0& \sigma ^3\\ \sigma ^3& 0\end{array}\right).$$
References
relax M. Berkooz and M. R. Douglas, hep-th/9610236, Phys. Lett. B395 (1997) 196. relax M. R. Douglas, D. Kabat, P. Pouliot and S. Shenker, hep-th/9608024, Nucl. Phys. B485 (1997), 85. relax S. Sethi and M. Stern, hep-th/9607145, Phys. Lett. B398 (1997), 47. relax M. R. Douglas, hep-th/9612126, J.H.E.P. 9707:004, 1997. relax E. Witten, hep-th/9511030, Nucl. Phys. B460 (1996) 541. relax N. Hitchin, math.DG/9909002. relax M. Claudson and M. Halpern, Nucl. Phys. B250 (1985) 689; R. Flume, Ann. Phys. 164 (1985) 189; M. Baake, P. Reinecke and V. Rittenberg, J. Math. Phys. 26 (1985) 1070. relax S. Sethi and M. Stern, hep-th/9705046, Comm. Math. Phys. 194 (1998) 675. relax P. Yi, hep-th/9704098, Nucl. Phys. B505 (1997) 307. relax M. Porrati and A. Rozenberg, hep-th/9708119, Nucl. Phys. B515 (1998) 184. relax V. G. Kac and A. Smilga, hep-th/9908096. relax M. B. Green and M. Gutperle, hep-th/9804123, Phys. Rev. D58 (1998) 46007. relax G. Moore, N. Nekrasov and S. Shatashvili, hep-th/9803265, Comm. Math. Phys. 209 (2000) 77. relax P. Vanhove, hep-th/9903050, Class. Quant. Grav. 16 (1999) 3147. relax W. Krauth, H. Nicolai and M. Staudacher, hep-th/9803117, Phys. Lett. B431 (1998) 31; W. Krauth and M. Staudacher, hep-th/9804199, Phys. Lett. B435 (1998) 350. relax M. Halpern and C. Schwartz, hep-th/9712133, Int. J. Mod. Phys. A13 (1998) 4367. relax J. Frohlich, G. M. Graf, D. Hasler, J. Hoppe and S.-T. Yau, hep-th/9904182; G. M. Graf and J. Hoppe, hep-th/9805080. relax J. Jost and K. Zuo, “Vanishing Theorems for $`L^2`$-cohomology on infinite coverings of compact Kähler manifolds and applications in algebraic geometry,” preprint no. 70, Max-Planck-Institut für Mathematik in den Naturwissenschaften, Leipzig (1998). |
warning/0001/cond-mat0001337.html | ar5iv | text | # Coarsening in surface growth models without slope selection88footnote 8Dedicated to the Peanuts on the occasion of their 50th birthday.
## References
## Figure captions |
warning/0001/gr-qc0001076.html | ar5iv | text | # 1 Intuitionistic theory of gravitation
## 1 Intuitionistic theory of gravitation
Intuitionistic theory of gravitation is based on Synthetic Differential Geometry of Kock-Lawvere (SDG) . SDG is built on the base of change the field of real numbers $`\mathrm{I}\mathrm{R}`$ on commutative ring $`𝐑`$, allowing to define on him differentiation, integrating and ”natural numbers”. It is assumed that there exists $`D`$ such that $`D=\{x𝐑|x^2=0\}`$ and that following the Kock-Lawvere axiom is held:
> for any $`g:D𝐑`$ it exist the only $`a,b𝐑`$ such that $`g(d)=a+db`$ for any $`dD`$.
This means that any function in given geometry is differentiable, but ”the law of excluded middle” is false. In other words, intuitionistic logic acts in SDG. But on this way one is possible building an intuitionistic theory of gravitation in analogy with the General theory of Relativity of Einstein . The elements of $`dD`$ are called infinitesimals, i.e. infinitesimal numbers. On the ring $`𝐑`$ we can look as on the field of real numbers $`\mathrm{I}\mathrm{R}`$ complemented by infinitesimals.
The vacuum Einstein equations in SDG in space-time $`𝐑^4`$ can be written with nonzero tensor of the energy. For instance,
$$R_{ik}\frac{1}{2}g_{ik}(R2\mathrm{\Lambda })=\frac{8\pi G}{c^2}du_iu_k,$$
$`(1)`$
where density of matter $`dD`$ is arbitrarily taken infinitesimal . For infinitesimals are holding relations which are impossible from standpoints of classical logic: $`d0`$,& $`d0`$ and formulas $`d=0`$, $`d0`$ are not valied. Such non-classical density of vacuum matter will consistent with zero in right part of the Einstein’s equations in the case of the vacuum in classical General theory of Relativity. For this one is sufficiently to consider SDG in so named well-adapted models, in which we can act within the framework of classical logic. For instance, in smooth topos $`\mathrm{𝐒𝐞𝐭}^{𝐋_{\mathrm{𝐨𝐩}}}`$, where $`𝐋`$ is category of loci, i.e. is the opposite category of category of finitely generated $`C^{\mathrm{}}`$-rings , the equations (1) at stage of locus $`\mathrm{}A=\mathrm{}(C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/I)`$, $`I`$ is a certain ideal of $`C^{\mathrm{}}`$-smooth functions from $`\mathrm{I}\mathrm{R}^n`$ to $`\mathrm{I}\mathrm{R}`$, have the form
$$R_{ik}(a)\frac{1}{2}g_{ik}(a)(R(u)2\mathrm{\Lambda }(a))=\frac{8\pi G}{c^2}d(a)u_i(a)u_k(a)\mathrm{mod}I,$$
$`(2)`$
where $`a\mathrm{I}\mathrm{R}^n`$ in parenthesises shows that we have functions, but at stage $`\mathrm{𝟏}=\mathrm{}(C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a\})`$, equations (2) take a classical form with null (on $`\mathrm{mod}\{a\}`$) tensor of the energy.
## 2 Spherically symmetrical vacuum field
We have the Einstein equations describing the gravitational field created by certain material system
$$R_{ik}\frac{1}{2}g_{ik}(R2\mathrm{\Lambda })=\kappa c^2\rho u_iu_k,$$
$`(3)`$
Here $`\kappa =8\pi G/c^4`$, $`\rho `$ is density of dust in the space which will consider further constant value. Suppose that dust is described in coordinate system in which $`u_i=(\mathrm{e}^{\frac{\nu }{2}},0,0,0),`$ $`u^k=g^{ik}u_i=(\mathrm{e}^{\frac{\nu }{2}},0,0,0).`$ Consider case, when gravitational field possesses a central symmetry. Central symmetry of field means that interval of space-time can be taken in the form
$$ds^2=\mathrm{e}^{\nu (r,t)}dt^2e^{\lambda (r,t)}dr^2r^2(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)$$
In our paper we found the following vacuum solution of the equations (3) for which
$$\rho \nu ^{}=0.$$
(1)
1) Well-known the classical Schwarzschild solution when $`\rho =0`$ and $`\nu ^{}0`$
$$ds^2=\left(1\frac{\mathrm{\Lambda }r^2}{3}+\frac{C}{r}\right)dt^2\frac{dr^2}{1{\displaystyle \frac{\mathrm{\Lambda }r^2}{3}}+{\displaystyle \frac{C}{r}}}$$
$$r^2(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2).$$
$`(4)`$
2) Non-classical case when both values $`\rho `$ and $`\nu ^{}`$ are simultaneously inseparable from the zero, i.e. are infinitesimals
$$ds^2=\left(1+\frac{(\kappa c^2\rho 2\mathrm{\Lambda })r^2}{6}+\frac{C}{r}\right)dt^2\frac{dr^2}{1{\displaystyle \frac{(\mathrm{\Lambda }+\kappa c^2\rho )r^2}{3}}+{\displaystyle \frac{C}{r}}}$$
$$r^2(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)$$
$`(5)`$
This metric can be called the non-classical Schwarzschild solution of the Einstein equations. But here $`\mathrm{\Lambda }`$ and $`C`$ are infinitesimals. So (5) is infinitesimal gravitational field, which is not weak in some sence (see below §5).
3) Suppose that gravitational field has no singularity in all space. This means that metric has no singularity in $`r=0`$. So we shall consider that $`C`$ is zero. But it is proved that
$$2\mathrm{\Lambda }\rho =\kappa c^2\rho ^2$$
$`(6)`$
and, besides, $`\mathrm{\Lambda }`$ is inconvertible value of ring $`𝐑.`$
In other words, matter has non-classical density, and its gravitational field has the form
$$ds^2=\left(1+\frac{(\kappa c^2\rho 2\mathrm{\Lambda })r^2}{6}\right)dt^2\frac{dr^2}{1{\displaystyle \frac{(\mathrm{\Lambda }+\kappa c^2\rho )r^2}{3}}}r^2(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)$$
$`(7)`$
Below we consider solutions (5), (7) in models which are different toposes, or, more exactly in categories of presheaves $`\mathrm{𝐒𝐞𝐭}^{𝒞^{op}},`$ where $`𝒞`$ is such subcategory of category $`𝐋`$ that ring $`𝐑=\mathrm{Hom}_𝐋(,\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}))`$ is local and Archimedean <sup>1</sup><sup>1</sup>1for example, categories of closed and germ-detrmined ideals are such subcategories .
## 3 Stagies
Synthetic Differential Geometry uses ”naive” style, i.e. contains term ”element”, or the set-theoretic formulas of the form $`aA`$. So it is necessary to be able to understand this naive writing as referering to cartesian closed categories, and to toposes in particular, because the Kock-Lawvere axiom has no models in the category of sets. The method for deciding this problem is introducing of generalized element $`b_XB`$, that is, a map $`XB`$, where $`X`$ is an arbitrary object, called the stages of definition, or the domain of variation of the element $`b`$. The ”classical” element is a map $`\mathrm{𝟏}B`$. As A.Kock writes ”when thinking in terms of physics (of which geometry of space forms a special case), reason for the name ”domain of variation” (instead of ”stage of definition”) becomes clear: for a non-atomistic point of view, a body $`B`$ is not described just in terms of its ”atoms” $`bB`$, that is $`\mathrm{𝟏}B`$, but in terms of ”particles” of varying size $`X`$, or in terms of motions that take place in $`B`$ and are parametrized by a temporal extent $`X`$; both of these situations being described by maps $`XB`$ for suitable domain of variation $`X`$ . In our case the role of ”body” $`B`$ will play a gravitational vacuum field $`g_{ik}`$, or geometry of space-time, which we will study at different stages, or with respect to different points of view concerning the possible geometric structure of the World.
In the case of topos $`\mathrm{𝐒𝐞𝐭}^𝐋`$ the concept of stage is realized with the help of the folowing method.
There exists the Yoneda embeding \[13, p.26\]
$$y:𝐋\mathrm{𝐒𝐞𝐭}^{𝐋^{op}},$$
$$y(\mathrm{}A)(\mathrm{}B)=\mathrm{Hom}_𝐋(\mathrm{}B,\mathrm{}A)$$
and for a morphism $`\alpha :\mathrm{}B\mathrm{}C`$
$$y(\mathrm{}A)(\alpha ):\mathrm{Hom}_𝐋(\mathrm{}C,\mathrm{}A)\mathrm{Hom}_𝐋(\mathrm{}B,\mathrm{}A),$$
$$y(\mathrm{}A)(\alpha )(u)=u\alpha ,u:\mathrm{}C\mathrm{}A.$$
Briefly for the Yoneda embedings we write
$$y(\mathrm{}A)=\mathrm{Hom}_𝐋(,\mathrm{}A).$$
Instead of $`y(\mathrm{}A)`$ we shall write simply $`\mathrm{}A`$. So if ring $`𝐑`$ is $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})`$ then
$$𝐑y(𝐑)=\mathrm{Hom}_𝐋(,\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})).$$
Hence element of ring $`𝐑`$, i.e. intuitionistic real number can be represented by arbitrary morphism of the form $`\mathrm{}A\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})`$. We say in such case that one have real at stage $`\mathrm{}A`$. It means that metric (5) must be considered at different stages. For example, at stage $`\mathrm{}A=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/I`$, where $`I`$ is some ideal of ring $`C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)`$.
Note that an event $`x`$ of the space-time $`𝐑^4`$ at stage $`\mathrm{}A`$ is the class of $`C^{\mathrm{}}`$-smooth vector functions $`(X^0(a),X^1(a),X^2(a),X^3(a)):\mathrm{I}\mathrm{R}^n\mathrm{I}\mathrm{R}^4`$, where each function $`X^i(a)`$ is taken by $`\mathrm{mod}I`$. The argument $`a\mathrm{I}\mathrm{R}^n`$ is some ”hidden” parameter corresponding to the stage $`\mathrm{}A`$. Hence it follows that at stage of real numbers $`𝐑=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})`$ of the topos under consideration an event $`x`$ is described by just a $`C^{\mathrm{}}`$-smooth vector function $`(X^0(a),X^1(a),X^2(a),X^3(a)),a\mathrm{I}\mathrm{R}`$. At stage of $`𝐑^2=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^2)`$ an event $`x`$ is 2-dimensional surface, i.e. a string. The classical four numbers $`(x^0,x^1,x^2,x^3)`$, the coordinates of the event $`x`$, are obtained at the stage $`\mathrm{𝟏}=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^0)=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a\}`$ (the ideal $`\{a\}`$ allows one to identify functions if their values at $`0`$ coincide), i.e., $`x^i=X^i(0),i=0,1,2,3`$.
There exists a number of types of infinitesimals: first-order infinitesimals $`D=\{x𝐑|x^2=0\}`$, $`k^{th}`$-order infinitesimals $`D_k=\{x𝐑|x^{k+1}=0\}`$, and the infinitesimals $`\mathrm{\Delta }\mathrm{\Delta }=\{x𝐑|f(x)=0,allfm_{\{0\}}^g\}`$, where $`m_{\{0\}}^g`$ is the ideal of functions having zero germ at $`0`$, i.e. vanishing in a neighborhood of $`0`$,
$$DD_2\mathrm{}D_k\mathrm{}\mathrm{\Delta }\mathrm{\Delta }.$$
Infinitesimals $`\rho ,\mathrm{\Lambda }\mathrm{\Delta }\mathrm{\Delta }`$ at stage $`\mathrm{}A=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/I`$ are the classes $`\rho (a)\mathrm{mod}I`$, $`\mathrm{\Lambda }(a)\mathrm{mod}I`$ such that for every $`\varphi m_{\{0\}}^g`$, $`\varphi \rho I`$, $`\varphi \mathrm{\Lambda }I`$, where $`\rho :\mathrm{I}\mathrm{R}^n\mathrm{I}\mathrm{R}`$ \[11, p.77\], or
$$(\varphi \rho )(a)=0\mathrm{mod}I,(\varphi \mathrm{\Lambda })(a)=0\mathrm{mod}I$$
The condition (6) has the form
$$2\mathrm{\Lambda }(a)\rho (a)\kappa c^2\rho ^2(a)=0\mathrm{mod}I.$$
$`(6^{})`$
Let us to consider the forms of metrics (5), (7) at different stages.
### 3.1 Stage $`\mathrm{𝟏}`$
For classical General Theory of Relativity we have stage $`\mathrm{𝟏}=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^0)=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a\}`$. At this stage metric (7) is metric of Minkowski space-time
$$g_{ik}(a)=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& r^2\mathrm{sin}^2\theta & 0\\ 0& 0& 0& r^2\end{array}\right),$$
i.e. $`\mathrm{\Lambda }`$ and $`\rho `$ are equal to zero. In fact,
$$\varphi (\rho (a))=\varphi (\rho (0))+\varphi ^{}(\rho (0))\rho ^{}(0)a+o(|a|).$$
$`(8)`$
Since $`\varphi (\rho (a))I=\{a\}`$ then $`\varphi (\rho (0))=0`$. Hence $`\rho (0)=0`$, because $`\varphi m_{\{0\}}^g`$. Then $`\rho \mathrm{mod}I=0`$. Simiraly $`\mathrm{\Lambda }\mathrm{mod}I=0`$, $`C\mathrm{mod}I=0`$
### 3.2 Stage $`D=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a^2\}`$
In this case
$$g_{ik}(a)=$$
$$\left(\begin{array}{cccc}1+\frac{(\kappa c^2\rho _12\mathrm{\Lambda }_1)}{6}ar^2+\frac{C_1a}{r}& 0& 0& 0\\ 0& \frac{1}{1{\displaystyle \frac{(\kappa c^2\rho _1+\mathrm{\Lambda }_1)}{3}}ar^2+{\displaystyle \frac{C_1a}{r}}}& 0& 0\\ 0& 0& r^2\mathrm{sin}^2\theta & 0\\ 0& 0& 0& r^2\end{array}\right)$$
In fact, it follows from (8) that $`\varphi (\rho (0))=\varphi ^{}(\rho (0))\rho ^{}(0)=0`$. Since $`\varphi |_U0,\varphi ^{}|_U0`$ for some neighborhood of $`0`$, then $`\rho (0)=0`$. So $`\rho \mathrm{mod}I=\rho _1a`$, $`\rho _1\mathrm{I}\mathrm{R}`$. Similaly $`\mathrm{\Lambda }\mathrm{mod}I=\mathrm{\Lambda }_1a`$, $`\mathrm{\Lambda }_1\mathrm{I}\mathrm{R}`$.
### 3.3 Stage $`D_p=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a^{p+1}\}`$
Here
$$g_{00}(a)=1+\underset{k=1}{\overset{p}{}}\left[\frac{(\kappa c^2\rho _k2\mathrm{\Lambda }_k)}{6}r^2+\frac{C_k}{r}\right]a^k$$
$$g_{11}(a)=\frac{1}{1\underset{k=1}{\overset{p}{}}\left[{\displaystyle \frac{(\kappa c^2\rho _k+\mathrm{\Lambda }_k)}{3}}r^2{\displaystyle \frac{C_k}{r}}\right]a^k}$$
and others $`g_{ik}`$ are classical.
### 3.4 Stage $`D_n(k)=\mathrm{}J_n^k=\mathrm{}C_0^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/m^{k+1}`$
Let $`m=f|f(0)=0`$ is maximal ideal of ring $`\mathrm{}C_0^{\mathrm{}}(\mathrm{I}\mathrm{R}^n).`$ Then
$$g_{00}(a,t,r,\phi ,\theta )=1+\underset{l=1}{\overset{k}{}}\underset{i_1,\mathrm{},i_l=1}{\overset{n}{}}\left[\frac{(\kappa c^2\rho _{i_1,\mathrm{},i_l}2\mathrm{\Lambda }_{i_1,\mathrm{},i_l})}{6}r^2+\frac{C_{i_1,\mathrm{},i_l}}{r}\right]a_{i_1}\mathrm{}a_{i_l}$$
$$g_{11}(a,t,r,\phi ,\theta )=\frac{1}{1\underset{l=1}{\overset{k}{}}\underset{i_1,\mathrm{},i_l=1}{\overset{n}{}}\left[{\displaystyle \frac{(\kappa c^2\rho _{i_1,\mathrm{},i_l}+\mathrm{\Lambda }_{i_1,\mathrm{},i_l})}{3}}r^2{\displaystyle \frac{C_{i_1,\mathrm{},i_l}}{r}}\right]a_{i_1}\mathrm{}a_{i_l}}$$
Here $`a=(a_1,\mathrm{},a_n)`$.
### 3.5 Stage $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^2)/\{a_1a_2\}`$
In this case functions $`\mathrm{\Lambda }`$, $`\rho ,C`$ depend of one variable, for example, $`a_2,`$ and vanishing at $`0`$. Then
$$g_{00}=1+\frac{\kappa c^2\rho (a_2)2\mathrm{\Lambda }(a_2)}{6}r^2+\frac{C(a_2)}{r}$$
$$g_{11}=\frac{1}{1{\displaystyle \frac{\kappa c^2\rho (a_2)+\mathrm{\Lambda }(a_2)}{3}}r^2+{\displaystyle \frac{C(a_2)}{r}}}$$
### 3.6 Stage $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{\mathrm{sin}\pi a,\mathrm{cos}\pi a\}`$
If $`\rho D`$ then
$$\rho (a)=\frac{\alpha _0}{2}+\underset{k=0}{\overset{\mathrm{}}{}}\alpha _k\mathrm{cos}k\pi a+\beta _k\mathrm{sin}k\pi a=A\mathrm{mod}I,A\mathrm{I}\mathrm{R},$$
$$\rho ^2(a)=A^2\mathrm{mod}IIA=0.$$
Hence $`\rho =0`$, $`\mathrm{\Lambda }=0`$ and $`g_{ik}`$ in this case coincides with Minkowski metric.
### 3.7 Stage $`\mathrm{}C^{\mathrm{}}(U)`$
Consider stage $`\mathrm{}C^{\mathrm{}}(U)`$, where $`U\mathrm{I}\mathrm{R}^n`$ is open set. Since
$$\mathrm{}C^{\mathrm{}}(U)\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^{n+1})/\{a_{n+1}\chi (a)1\},$$
$$U=\{a\mathrm{I}\mathrm{R}^n|\chi (a)0\},\chi C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n),$$
then with the help of transformation of variables
$$\{\begin{array}{ccc}a=a\hfill & & \\ a_{n+1}^{}=a_{n+1}\chi (a)1\hfill & & \end{array}$$
we can get, for example, that
$$\rho (a,a_{n+1})\mathrm{mod}I=\underset{|\alpha |=1}{\overset{\mathrm{}}{}}A_\alpha a^\alpha ,aU,A_\alpha \mathrm{I}\mathrm{R}$$
$$\rho (0,\frac{1}{a(0)})=0.$$
## 4 Transitions between stages
Change of stage $`\mathrm{}A`$ on stage $`\mathrm{}B`$ is morphism between two stages
$$\mathrm{}B\stackrel{\psi }{}\mathrm{}A$$
then transition between $`Hom_𝐋(\mathrm{}A,T)`$ $`Hom_𝐋(\mathrm{}B,T)`$ is realized by means of
$$Hom_𝐋(\mathrm{}A,T)\stackrel{\mathrm{\Psi }}{}Hom_𝐋(\mathrm{}B,T)$$
for any object $`T`$ of category $`𝐋`$, which each morphism $`h:\mathrm{}AT`$ puts in correspondence morphism $`h\mathrm{\Psi }:\mathrm{}BT`$.
Let, now, $`\mathrm{}A=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/I`$ and $`\mathrm{}B=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^m)/J.`$ Then transition between stages $`\mathrm{}A`$, $`\mathrm{}B`$ gives metric
$$g_{ik}(b)=$$
$$\left(\begin{array}{cccc}1+\frac{(\kappa c^2\rho (\mathrm{\Psi }(b))2\mathrm{\Lambda }(\mathrm{\Psi }(b)))}{6}r^2& 0& 0& 0\\ 0& \frac{1}{1{\displaystyle \frac{(\kappa c^2\rho (\mathrm{\Psi }(b))+\mathrm{\Lambda }(\mathrm{\Psi }(b)))}{3}}r^2}& 0& 0\\ 0& 0& r^2\mathrm{sin}^2\theta & 0\\ 0& 0& 0& r^2\end{array}\right)$$
modulo $`JC^{\mathrm{}}(\mathrm{I}\mathrm{R}^m\times \mathrm{I}\mathrm{R}^4)`$ instead of metric (7).
The condition of infinitesimallity for $`\mathrm{\Lambda }`$, $`\rho `$ and condition (6) will copied so
$$(\varphi \rho \mathrm{\Psi })(b)=0\mathrm{mod}J,(\varphi \mathrm{\Lambda }\mathrm{\Psi })(b))=0\mathrm{mod}J$$
and
$$2\mathrm{\Lambda }(\mathrm{\Psi })(b))\rho (\mathrm{\Psi })(b))\kappa c^2\rho ^2(\mathrm{\Psi })(b))=0\mathrm{mod}J.$$
## 5 Physical notes
At first note very interesting fact: at all considered stages signature of metric $`g_{ik}`$ depends of the form of functions $`\mathrm{\Lambda }`$, $`\rho `$ and $`C`$. For example, at stage $`D=\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a^2\}`$ $`\rho \mathrm{mod}I=\rho _1a`$, $`\mathrm{\Lambda }\mathrm{mod}I=\mathrm{\Lambda }_1a`$, $`C\mathrm{mod}I=C_1a`$, where $`\rho _1,\mathrm{\Lambda }_1,C_1\mathrm{I}\mathrm{R}`$ are arbitrary real numbers (under $`C=0`$ the condition (6) or (6’) is valid for all $`a\mathrm{I}\mathrm{R}`$). Hence field $`g_{ik}`$ is not weak with respect to five dimensions $`(t,r,\theta ,\varphi ,a)`$. More interesting situation can be observed at stages $`D_n(k)=\mathrm{}J_n^k=lC_0^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/m^{k+1}`$ and $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^2)/\{a_1a_2\}`$.
Note also that at stage $`\mathrm{}C^{\mathrm{}}(U)`$ if $`U`$ is bounded then functions $`\rho ,\mathrm{\Lambda },C`$ are can be taken small and signature of metric does not change.
What sence have the ”hidden” parameters $`a\mathrm{I}\mathrm{R}^n`$? We are thinking that they say us about existence of the additional dimensions the number of which can be changeable. For finding of coefficients $`\rho _1,\mathrm{\Lambda }_1,C_1`$ at stage $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R})/\{a^2\}`$, $`\rho _{i_1,\mathrm{},i_l},\mathrm{\Lambda }_{i_1,\mathrm{},i_l},C_{i_1,\mathrm{},i_l}`$ at stage $`\mathrm{}J_n^k=lC_0^{\mathrm{}}(\mathrm{I}\mathrm{R}^n)/m^{k+1}`$, functions $`\rho (a_2),\mathrm{\Lambda }(a_2),C_(a_2)`$ at stage $`\mathrm{}C^{\mathrm{}}(\mathrm{I}\mathrm{R}^2)/\{a_1a_2\}`$ must be possibly used the many-dimensional Einstein equations. In other words 4-dimensional intuitionistic theory contains uncountable number of many-dimensional theories. The infinitesimal field with respect to 4-dimensional universe can be found non-weak with respect to ”hidden” geometry. Intuitionistic logic implies new view about Nuture of World. |
warning/0001/cond-mat0001344.html | ar5iv | text | # Aging in a topological spin glass
## I Introduction
Much recent work has been focused on the nature of glassy magnetic states in the absence of quenched disorder. Simple experimental realizations of non-disordered systems in which glassy phases have been found are antiferromagnets (AFM) with kagomé and pyrochlore geometries : (H<sub>3</sub>O)Fe<sub>3</sub>(SO<sub>4</sub>)<sub>2</sub>(OH)<sub>6</sub>, Y<sub>2</sub>Mo<sub>2</sub>O<sub>7</sub>, Tb<sub>2</sub>Mo<sub>2</sub>O<sub>7</sub>, Y<sub>2</sub>Mn<sub>2</sub>O<sub>7</sub> and Yb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. In the case of the two dimensional (2d) kagomé AFM, it is believed that an xy anisotropy can induce a finite temperature glass transition . This glassy phase has been termed a topological spin glass, in order to distinguish it from conventional site-disordered systems. While the high degeneracy of these lattices is expected to be robust against the addition of small amounts of impurities, and so should not result in a spin glass state, in some cases conventional behaviour can be recovered if this dilution is sufficient. Examples of this are the 3d pyrochlore Y<sub>2</sub>Mn<sub>2</sub>O<sub>7</sub> where the effect of increasing disorder is to progressively restore conventional spin glass kinetics, and the 2d kagomé (D<sub>3</sub>O)Fe<sub>3-x</sub>Al<sub>y</sub>(SO<sub>4</sub>)<sub>2</sub>(OD)<sub>6</sub> where reduction of the Fe occupancy to $`90`$ % has been found to induce long-range magnetic order.
At present, very little is known about the glassy phases found in non-disordered frustrated systems. In this letter, we study the $`S=5/2`$ 2d kagomé AFM (H<sub>3</sub>O)Fe<sub>3</sub>(SO<sub>4</sub>)<sub>2</sub>(OH)<sub>6</sub>. Our ac susceptibility measurements indicate the critical character of the freezing process around $`T_g18`$ K. We find that while aging effects at a fixed temperature below $`T_g`$ are the same as in site-disordered spin glasses, there is a very weak sensitivity of these aging phenomena to temperature variations what is is in sharp contrast with conventional spin glasses. We propose to understand the transition found in this 2d system, and the slow dynamics observed, in terms of a theoretical picture of a spin glass state of the ordered kagomé AFM, based on topological excitations termed spin folds.
Our sample was that used in previous studies of the crystal structure, susceptibility and $`\mu SR`$ responses. Neutron diffraction measurements on the deuterated compound have shown that the glassy phase involves 2d antiferromagnetic correlations that saturate to a maximum at low temperature, with a spin-spin correlation length of $`\xi =19\pm 2`$ Å. Associated with these correlations is a $`T^2`$ specific heat, which contrasts with the linear dependence of conventional spin glasses and suggests the presence of 2d antiferromagnetic spin waves.
## II Nature of the freezing mechanism
We examined the nature of the freezing transition by measuring the frequency dependence of the ac susceptibility at 8 frequencies ranging from 0.04 to 80 Hz. The out-of-phase component $`\chi ^{\prime \prime }`$ was too weak to allow a precise study, but its overall shape compares well with spin glasses . We have found that the temperature of the $`\chi ^{}`$-peak shifts very slowly with frequency ($`T_f(80\text{ Hz})=18.5`$ K, $`T_f(0.04\text{ Hz})=18.0`$ K). Quantitatively, the shift amounts to $`\mathrm{\Delta }T_f/(T_f\mathrm{\Delta }\mathrm{log}\omega )0.008`$, which is in the same range as obtained for the critical behaviour of conventional 3d spin glasses . As a confirmation, a fit of the $`T_f(\omega )`$ data to an Arrhénius function $`\omega =\omega _0\mathrm{exp}(U/k_BT_f)`$ yields an unphysical value of $`\omega _010^{130}`$ Hz (in agreement with the previous interpretation of results from the deuterated sample in terms of a Vogel-Fulcher law ).
The critical character of the freezing transition can also be confirmed by a dynamic scaling analysis of the same data. The fit to the critical scaling law $`\omega =\omega _0[(T_f(\omega )T_f(0))/T_f(0)]^{z\nu }`$ yields, fixing $`z\nu =7`$, the plausible values of $`T_f(0)=17.8`$ K and $`\omega _0=\mathrm{5.9\; 10}^{11}`$ Hz. This transition temperature is far below the Curie-Weiss temperature, which is here $`\mathrm{\Theta }_{CW}=1200\pm 200`$ K, emphasizing a strong influence of frustration. In Ref., it has been proposed that, in the presence of some xy anisotropy, the kagomé AFM experiences a Kosterlitz-Thouless type transition to a glassy state. The prediction of an upper estimate of the transition temperature $`T_{KT}\mathrm{\Theta }_{CW}/48`$, obtained from the spin-wave stiffness in the xy limit, agrees well with the experimental transition temperature, as here $`\mathrm{\Theta }_{CW}/48=25`$ K $`17.8`$ K.
Zero-field cooled (ZFC) and field-cooled (FC) dc magnetization measurements show another aspect of the freezing transition (Fig.1). The ZFC-curve peaks at approximately $`T_{ZFC}=17.8`$ K ($`T_f(\omega 0`$), but slight irreversibilities are still visible as a ZFC-FC separation up to $`25`$ K, indicating that some freezing occurs well above the transition. Also, instead of the low-temperature FC plateau that is usual in conventional spin glasses , there is below $`T_{ZFC}`$ a rise in the FC magnetization towards low temperature. Both these features point towards a progressive freezing of some superparamagnetic entities over a wide temperature range . These species are presumed to be the cause of the slight influence of the thermal history between 18 and 25 K on the slow dynamics below $`T_g`$ (see Sect. III B), and their fluctuations are the probable origin of the dynamic component observed in the $`\mu SR`$ measurements at low temperature. This progressive freezing is superimposed on the critical behaviour at the transition, and does not prevent the transition from taking place.
In a study of other jarosite samples, which show stronger superparamagnetic components, the divergence of the non-linear susceptibility, normally associated with a spin-glass transition, could not be clearly characterized.
The origin of superparamagnetism at low temperatures in this kagomé system is an open question. Although analytic calculations have shown that dilution of the magnetic sites creates superparamagnetic components at high temperatures, it is not clear whether the presence of such superparamagnetic entities in the spin glass state is intrinsic to the (2d) kagomé lattice, or is the result of disorder, despite its small level in our sample (the occupancy of the magnetic sites, determined from neutron diffraction, is 97.3 (6) %).
## III Aging phenomena
### A Spin glass-like behavior at constant temperature
In site-disordered spin glasses, the out-of-equilibrium dynamics is characterized by aging effects: the response to a small magnetic excitation depends on the time (‘age’) spent after the quench from above the freezing temperature . As the ac signal of our present sample is small, we have explored the nonequilibrium dynamics by means of dc magnetization relaxation. In such measurements, aging results in the dependence of the response to a small field variation, applied after waiting a time $`t_w`$ from the quench, on two time variables: $`t`$, the observation time (i.e. the time from the field variation), and $`t_a=t_w+t`$, the total aging time . In our experiments the sample was first cooled from 25 to 12 K in a field of $`50`$ Oe. Then, after waiting a time $`t_w`$ at 12 K, the field was removed, and the slow decay of the ‘thermo-remanent’ magnetization (TRM) was recorded as a function of $`t`$.
The observed behaviour, shown in Fig.2, is precisely the same as that of site-disordered spin glasses: the longer $`t_w`$, the slower the decay of the TRM. The system has become ‘stiffer’ with time. The curves also present the commonly observed inflection point in the region of $`\mathrm{log}t\mathrm{log}t_w`$ . The $`t_w`$-dependence of these responses satisfies the same scaling properties as site-disordered spin glasses and glassy polymers, as shown in the inset of Fig. 2.
### B Weak sensitivity to temperature variations
In site-disordered spin glasses, temperature cycling protocols have been developed in order to characterize the influence on aging of small temperature variations (remaining below $`T_g`$). They have shown, in contrast with the expectations for thermally activated processes, that a small positive temperature cycle can erase previous aging (‘rejuvenation’ or ‘chaos-like’ effect) instead of simply speeding up the evolution .
We have investigated the effect of such temperature variations on the jarosite system. The procedure of positive temperature cycling is sketched in the inset of Fig. 3. After aging $`9700`$ s at $`T=12`$ K, the sample is heated to $`T_0+\mathrm{\Delta }T`$ during a short time ($`100`$ s). Following this cycle, aging at $`12`$ K is continued during $`300`$ s, and the field is cut off (at $`t=0`$). Fig. 3 shows the comparison of the resulting decay curves with the reference TRM curves obtained after an isothermal aging of $`300`$ s or $`10000`$ s at $`12`$ K.
All curves obtained after the positive cycling are lower than the $`t_w10000`$ s isothermal reference: cycling to $`T_0+\mathrm{\Delta }T`$ has not increased the age of the system. This result also differs strongly from the site-disordered spin glass case, in which (in an equivalent experiment) $`\mathrm{\Delta }T=2.5`$ K has been found to completely erase previous aging . Here, erasure only occurs when $`T_0+\mathrm{\Delta }T`$ crosses $`T_g`$ (for $`\mathrm{\Delta }T=6`$ K), and even then results in a curve with a weaker slope than that of the isothermal $`t_w300`$ s reference. We note that aging at temperatures $`TT_g`$ still results in some slowing down of the dynamics, as heating up to $`25`$ K is needed to fully reinitialize aging (at this time scale).
For intermediate $`\mathrm{\Delta }T`$ values ($`25`$ K), Fig.3 shows a partial erasure of previous aging (but the influence is again much weaker than in conventional spin glasses). The effect on the curves is the same as in site-disordered spin glasses , partial erasure corresponds to a stronger effect on the short-time part of the curves (which approaches the ‘younger’ $`t_w300`$ s reference), while the long-time part is less affected (and remains close to the ‘older’ $`t_w10000`$ s reference).
Negative temperature cycling experiments in site-disordered spin glasses have shown that the stage of aging before the temperature cycle can be memorized during the lower-temperature aging, in such a way that aging can restart from the same point when the system returns to temperature $`T`$. That is to say, a long aging at $`T_0\mathrm{\Delta }T`$ can be of no influence on the effective age at $`T_0`$, even for small $`\mathrm{\Delta }T`$’s ($`12`$ K) for which little thermal slowing down is expected. This memory effect is equivalent to a growth of free energy barriers as the temperature is lowered .
When applied the negative cycling protocol does not yield these simple memory effects and confirms the surprisingly weak influence of temperature changes on aging, already observed in the positive cyclings of Fig. 3. This data are shown in Fig. 4. Since the resulting curves have approximately the same shape as those obtained after isothermal aging, they can be ascribed an effective waiting time, by application of the scaling procedure evoked above (this is not the case for the ‘mixed age’ curves obtained after positive cyclings). Similar (but slightly lower) values of the effective age can be directly estimated from the position of the inflection point. Qualitatively we find that larger values of $`\mathrm{\Delta }T`$ lead to a younger effective age (less influence of aging at $`T_0\mathrm{\Delta }T`$). However, quantitatively, the freezing effect is much weaker than expected from thermal slowing down, and hence is minuscule when compared with the freezing processes of site-disordered spin glasses.
For $`\mathrm{\Delta }T=9`$ K, aging at $`T_0\mathrm{\Delta }T=3`$ K is still seen to add a contribution to aging at $`T=12`$ K (the curve is clearly ‘older’ than the $`t_w300`$ s reference). Simple arguments of thermal slowing down with fixed energy barriers for such a temperature change, would lead to the expectation of even astronomic waiting times having no effect. In other words, in terms of thermal activation, the observed influence at $`12`$ K of aging at $`3`$ K corresponds to that of a much higher effective temperature (of the order of $`11`$ K).
## IV Discussion
Thus, we have found that the slow dynamics and aging effects in this kagomé antiferromagnet are, at constant temperature, similar to those of site-disordered spin glasses, but the rôle played by temperature changes is markedly different. Positive and negative cycling experiments have shown that aging is neither reinitialized nor drastically frozen by temperature changes: it is mostly unaffected, as long as $`T_g`$ is not crossed. This behavior is in marked contrast with conventional site-disordered spin glasses .
In a ground state configuration of the kagomé AFM, the nearest-neighbour moments are related by 120 , and the local chirality of the triangular motifs is either uniform or staggered. The domain walls that separate regions of uniform and staggered chirality correspond to topological defects which cost zero internal energy. In Ref., the transition of the kagomé AFM to a glassy state is described in terms of a Kosterlitz-Thouless type mechanism involving the binding of these defects. It has been argued that the processes of defect propagation are ‘non-Abelian’, i.e. their formation does not commute. This creates a further hindrance to the evolution towards a given spin configuration and means that defect propagation is not simply a matter of thermally activated barrier crossing.
While at a given instant the kagomé system can be thought of as a mosaic of chiral domains, the equilibrium state itself is believed to involve fluctuating short-range chiral correlations. The existence of aging effects is in agreement with Monte Carlo studies and indicates that this state is approached by slow relaxation and a gradual evolution of the system. The equilibrium state may correspond to a temperature dependent distribution of domains. They will have different Boltzmann weights because those with uniform chirality develop extended defects, whereas those within domains of staggered chirality are localized . The weak effect of temperature changes suggests that this distribution of chiral ordering varies only slowly with temperature.
In conclusion, we believe that (H<sub>3</sub>O)Fe<sub>3</sub>(SO<sub>4</sub>)<sub>2</sub>(OH)<sub>6</sub> is representative of the new class of ‘topological’ spin glasses. There is good evidence from ac measurements that a critical transition to a glassy state occurs at $`T_g0`$. While its aging dynamics at a given temperature follows that of conventional spin glass systems, there is a remarkable insensitivity to temperature changes that demonstrates clearly the different natures of site-disordered and such topological spin glasses.
We acknowledge enlightening discussions with D. Grempel and L.F. Cugliandolo. |
warning/0001/astro-ph0001126.html | ar5iv | text | # MULTI-WAVELENGTH OBSERVATIONS OF DUSTY STAR FORMATION AT LOW AND HIGH REDSHIFT
## 1 INTRODUCTION
The universe of galaxies beyond $`z1`$ has finally become relatively easy to observe. While only a handful of galaxies at $`z\stackrel{>}{}1`$ were known five years ago, close to 2000 have spectroscopic redshifts today. Planned redshift surveys should increase the number by at least an order of magnitude in the next several years. Analyzing the large and rapidly growing samples of high-redshift galaxies will be a major challenge in the coming decade.
Much of this analysis will rely on estimates of star-formation rates for the detected galaxies. These estimates are required to answer the most basic questions that high-redshift surveys can address—when did the stars in the universe form? how were the sites of star formation related to the evolving perturbations in the underlying distribution of matter?—but unfortunately no standard and widely accepted techniques exist for estimating star-formation rates from the data usually available in high-redshift samples. Researchers analyzing data taken at a variety of wavelengths with a variety of techniques have reached wildly discrepant conclusions about the star-formation history of the universe (e.g. Madau et al. 1996; Smail et al. 1997; Barger et al. 1998; Blain et al. 1999a; Blain et al. 1999c; Eales et al. 1999; Pettini et al. 1999; Steidel et al. 1999; Tan, Silk, & Balland 1999; Trentham, Blain, & Goldader 1999; Barger, Cowie, & Richards 2000, hereafter BCR).
The central problem is that the energy produced by star formation in young galaxies is emitted across more than 6 decades of frequency, from the far-UV to the radio, while most high-redshift surveys are based on observations at only a narrow range of frequencies. Standard methods for deriving a galaxy’s star-formation rate require an estimate of the total energy emitted by its massive stars, but in existing surveys this is never directly observed. Radio surveys detect only the tiny percentage of the total energy that goes into accelerating electrons in supernova remnants, for example, while optical surveys detect the portion of massive stars’ emission that is not absorbed by dust and sub-mm surveys detect a fraction of the portion that is. Large and uncertain corrections are needed to estimate the total emitted energy from the energy detected at any of these wavelengths. Differences in the adopted corrections are largely responsible for current discrepancies in estimates of the star-formation history of the universe.
In principle one could obtain a robust estimate of the energy emitted by massive stars in a high redshift galaxy by observing it over a sufficiently wide range of wavelengths, but with current technology this is rarely possible. Star-forming galaxies at high redshift almost certainly emit the bulk of their luminosities at far-infrared wavelengths that are extremely difficult to detect. Although massive stars themselves radiate mainly in the far-UV, rapid star formation in the local universe occurs in dusty environments where most UV photons are quickly absorbed, and as a result most of the luminosity produced by massive stars tends to emerge in the far-IR where dust radiates. This is known to be true for a broad range of galaxies in the local universe, from the famous class of “Ultra Luminous Infrared Galaxies” (ULIRGs, galaxies with $`L_{\mathrm{FIR}}10^{12}L_{}`$) to the much fainter UV-selected starbursts contained in the IUE Atlas (e.g. Meurer, Heckman, & Calzetti 1999, hereafter MHC). The recent detection of a large extragalactic far-IR background (e.g. Fixsen et al. 1998) suggests that it is likely to have been true at high redshifts as well.
Because most of the energy produced by massive stars is likely radiated by dust in the far-IR, estimates of the bolometric dust luminosities of high redshift galaxies are required in order to estimate their star-formation rates. The bolometric dust luminosities usually cannot be directly observed—the earth’s atmosphere is opaque in the far-IR and most high-redshift galaxies are too faint to have been detected by any far-IR satellite that has flown—but they can be inferred from known correlations between local galaxies’ bolometric dust luminosities and their luminosities in the optical, mid-IR, sub-mm, and radio atmospheric windows.
The first part of this paper aims to present a unified framework for estimating the bolometric dust luminosities of high-redshift galaxies selected at different wavelengths. In §2 we summarize and re-examine locally established correlations between bolometric dust luminosities and luminosities through atmospheric windows, quantifying their scatter in cases where it has not been done explicitly by previous authors. In §3 we present the limited available evidence that high-redshift galaxies obey these correlations. Published estimates of the cosmic star-formation history derived from observations through one atmospheric window have often been accompanied by claims that most of the star formation could not have been detected with observations through other atmospheric windows. In the second part of this paper, §4, we use the results of §§2 and 3 to evaluate whether this is true. After reviewing the properties of star-forming galaxies in the local universe, we discuss the properties of high-redshift galaxies selected in surveys at different wavelengths and estimate the contribution to the far-IR background from galaxies similar to those detected in optical, mid-IR, and sub-mm surveys. Our conclusions are presented in §5.
Throughout this paper we shall denote luminosities by $`L`$ when they have units of energy time<sup>-1</sup>, by $`l_\nu `$ when they have units of energy time<sup>-1</sup> frequency<sup>-1</sup>, and by $`l_\lambda `$ when they have units of energy time<sup>-1</sup> wavelength<sup>-1</sup>; and observed fluxes by $`f_\nu `$ (units of energy time<sup>-1</sup> frequency$`1`$ area<sup>-1</sup>) or $`f_\lambda `$ (units of energy time<sup>-1</sup> wavelength<sup>-1</sup> area<sup>-1</sup>).
## 2 DUST-OBSCURED STAR FORMATION IN THE LOCAL UNIVERSE
The energy produced by massive stars emerges from galaxies over a wide range of wavelengths. Figure 1 shows typical spectral shapes of this emission for rapidly star-forming galaxies in the local universe. Massive stars radiate predominantly in the far-UV, but as galaxies become dustier an ever larger fraction of their luminosity is absorbed by dust and re-radiated in the far-IR. In typical cases most of the energy released by massive stars emerges in the far-IR. This section describes how the dust’s bolometric luminosity can be estimated from observations through the atmospheric windows indicated on the $`x`$-axis of Figure 1.
### 2.1 Sub-mm constraints on bolometric dust luminosity
Although a galaxy’s total dust luminosity at $`10\mu \mathrm{m}\stackrel{<}{}\lambda \stackrel{<}{}1000\mu \mathrm{m}`$ cannot be directly measured from the ground, narrow atmospheric windows in the sub-millimeter (Figure 1) allow a small fraction of it to reach the earth’s surface. If dust in all galaxies radiated with the same known spectral energy distribution (SED)—say a 50K blackbody—then a galaxy’s bolometric dust luminosity could be reliably estimated by simply scaling the flux measured through a sub-mm atmospheric window.<sup>1</sup><sup>1</sup>1provided, of course, that the galaxy’s redshift and luminosity distance are known. Unfortunately dust SEDs vary considerably from one galaxy to the next and are more complex than simple blackbodies. Obtaining a reasonable fit to a galaxy’s measured dust SED usually requires at least two “modified” blackbodies of the form
$$f_\nu \frac{\nu ^{3+ϵ}}{\mathrm{exp}(h\nu /kT)1},$$
$`(1)`$
where $`ϵ1.5`$, the emissivity index, is sometimes treated as a free parameter in addition to the temperatures and amplitudes of the modified blackbodies (e.g. Klaas et al. 1997). The required parameters in these fits vary among galaxies in a way that is not quantitatively understood.
Because of the significant variation in the shape of galaxies’ dust SEDs, measuring a galaxy’s flux in one of the narrow sub-mm atmospheric windows cannot precisely determine its bolometric dust luminosity. Nevertheless, sub-mm fluxes provide some useful information; galaxies with large fluxes through the sub-mm atmospheric windows tend also to have large bolometric dust luminosities. To quantify how well sub-mm flux measurements can constrain the bolometric dust luminosities of actively star-forming galaxies in the local universe, we have assembled from the literature a sample of these galaxies with published IRAS fluxes at $`25\mu `$m, $`60\mu `$m, and $`100\mu `$m, as well as at least one flux measurement at a wavelength $`200\mu \mathrm{m}<\lambda <1000\mu \mathrm{m}`$ to constrain the temperature of the cool dust component and the slope of its (modified) Rayleigh-Jeans tail. This sample consists of 27 galaxies: the ULIRGs IRAS05189-2524, IRAS08572+3915, UGC5101, IRAS12112+0305, Mrk231, Mrk273, IRAS14348-1447, and IRAS15250+3609 detected at $`800\mu `$m (and sometimes $`450\mu `$m) by Rigopoulou et al. (1996); the LIRGs NGC1614, NGC3110, NGC4194, NGC4418, NGC5135, NGC5256, NGC5653, NGC5936, Arp193, and Zw049 detected at $`850\mu `$m by Lisenfeld et al. 1999; the interacting galaxies NGC6240, Arp220, and Arp244 observed at $`25\mu \mathrm{m}<\lambda <200\mu \mathrm{m}`$ by Klaas et al. (1997); the UV-selected starbursts NGC6090, NGC7673, NGC5860, IC1586, and Tol1924-416 detected at $`150\mu `$m and $`205\mu `$m by Calzetti et al. (1999); and the nearby starburst M82 (Hughes, Robson, & Gear 1990 and references therein).
The dust SEDs of these galaxies are shown in Figure 2. For clarity we have connected measurements of individual galaxies at different wavelengths with spline fits (continuous lines). These fits were calculated in $`\mathrm{log}\lambda \mathrm{log}f_\lambda `$ space and should roughly capture the expected power-law shape of dust SEDs in their (modified) Rayleigh-Jeans tails. Measurements from all galaxies were normalized to have the same total luminosity under the fits.
This plot can be used to estimate a rapidly star-forming galaxy’s bolometric dust luminosity on the basis of its luminosity at a single observed wavelength:
$$L_{\mathrm{bol},\mathrm{dust}}\frac{L_{\mathrm{bol},\mathrm{dust}}}{\lambda l_\lambda }\lambda l_\lambda 𝒦(\lambda )\lambda l_\lambda $$
$`(2)`$
where the expectation value is taken over all dust SEDs in the figure. Table 1 lists the values $`𝒦`$ at various wavelengths. Ideally one would calculate $`𝒦`$ by fitting two modified blackbodies to each dust SED, as discussed above, but most of the galaxies in our sample have measured fluxes at only 4 relevant wavelengths ($`25\mu `$m, $`60\mu `$m, $`100\mu `$m, and $`850\mu `$m) and this fit has more than 4 free parameters. Instead, we estimated $`𝒦`$ from the log-log spline interpolations shown in Figure 2. This spline fitting does not take proper account of uncertainties in the measured fluxes, but the uncertainties are almost always far smaller than the intrinsic differences in dust SEDs among the galaxies in our sample. The observed spread among these galaxies in $`L_{\mathrm{bol},\mathrm{dust}}/\lambda l_\lambda `$ at a given wavelength provides a measure of how reliably a flux measurement at that wavelength can constrain a galaxy’s bolometric dust luminosity. The RMS spread in $`\mathrm{log}_{10}𝒦`$, also listed in Table 1, suggests sub-mm estimates of $`L_{\mathrm{bol},\mathrm{dust}}`$ will be accurate to about a factor of 2.
According to Sanders & Mirabel (1996) and Dunne et al. (2000), the shapes of galaxies’ dust SEDs are correlated with their bolometric luminosities, and so one might hope to estimate $`L_{\mathrm{bol},\mathrm{dust}}`$ more accurately by using different values of $`𝒦`$ for galaxies with different absolute sub-mm luminosities. Unfortunately this trend does not appear to be strong enough among rapidly star-forming galaxies to substantially reduce the scatter in $`𝒦`$ (i.e. to substantially improve estimates of $`L_{\mathrm{bol},\mathrm{dust}}`$ derived from photometry in a single sub-mm window). Illustrating this point, Figure 3 shows $`𝒦(200\mu \mathrm{m})`$ ($`L_{\mathrm{bol},\mathrm{dust}}/\lambda l_\lambda `$ at $`\lambda =200\mu `$m) versus $`L_{\mathrm{bol},\mathrm{dust}}`$ for each galaxy in our sample. Although LIRGs clearly have systematically different $`𝒦(200\mu \mathrm{m})`$ values than ULIRGs, the trend disappears at lower luminosities. Including an absolute luminosity dependence in $`𝒦(\lambda )`$ seems unlikely to significantly improve sub-mm constraints on $`L_{\mathrm{bol},\mathrm{dust}}`$.
Measuring the luminosity of a galaxy in two sub-mm bands instead of only one provides additional information about the shape of the dust SED that could in principle be used to estimate $`L_{\mathrm{bol},\mathrm{dust}}`$ more accurately. Unfortunately the major sub-mm atmospheric windows at $`450\mu `$m and $`850\mu `$m both lie in the Rayleigh-Jeans tail of typical dust SEDs at $`z\stackrel{<}{}3`$. Measuring both $`450\mu `$m and $`850\mu `$m fluxes of a galaxy at these redshifts therefore reveals little about the temperature of the cold dust component and almost nothing at all about the possible presence of other dust components at higher temperatures. At larger redshifts measuring both fluxes provides better constraints on the shape of the dust SED, but even then two flux measurements cannot uniquely determine the shape of dust SEDs that typically require more than four free parameters for a reasonable fit. In any case accurate fluxes in more than one sub-mm band are seldom available for the faint high-redshift galaxies that are our primary concern. Obtaining these additional sub-mm fluxes may ultimately prove useful for more accurate estimation of galaxies’ bolometric dust luminosities, but we will not consider this possibility further.
### 2.2 Mid-Infrared constraints on bolometric dust luminosity
At wavelengths $`\lambda \stackrel{<}{}15\mu `$m the dust emission from galaxies tends to rise significantly above the modified blackbody fits discussed in §2.1. This excess emission is thought to originate from very small grains transiently heated by single UV-photons to high non-equilibrium temperatures; much of it emerges in a few discrete polycyclic aromatic hydrocarbon (PAH) lines at 3.3, 6.2, 7.7, 8.6, and $`11.3\mu `$m (see, e.g., Xu et al. 1998 and references therein). $`7.7\mu `$m is often the strongest of these lines and the easiest to detect at high redshift. In this subsection we will attempt to quantify the constraints that observations through a 6–9$`\mu `$m filter can place on a galaxy’s bolometric dust luminosity. We will need this information in §3.3 below.
The sample we will use consists of 11 starbursts and 53 star-formation–dominated ULIRGs observed in the mid-IR by Rigopoulou et al. (2000) and the starburst NGC6090 observed in the mid-IR by Acosta-Pulido et al. (1996).
In analogy to §2.1 we will define a constant of proportionality $`𝒦_{\mathrm{MIR}}`$ between bolometric dust luminosity and luminosity at $`6\mu \mathrm{m}<\lambda <9\mu \mathrm{m}`$:
$$L_{\mathrm{bol},\mathrm{dust}}=𝒦_{\mathrm{MIR}}_{6\mu \mathrm{m}}^{9\mu \mathrm{m}}𝑑\lambda l_\lambda .$$
$`(3)`$
After excluding two outliers with anomalously low $`8\mu `$m luminosities, IRAS00199-7426 and IRAS20100-4156, the mean value of $`𝒦_{\mathrm{MIR}}`$ among the galaxies in this sample is 52, the mean value of $`1/𝒦_{\mathrm{MIR}}`$ is 0.025, and the rms dispersion in $`\mathrm{log}_{10}𝒦_{\mathrm{MIR}}`$ is 0.22 dex. Each galaxy’s bolometric dust luminosity was estimated from its measured IRAS $`60\mu `$m and $`100\mu `$m luminosities with the relationship $`L_{\mathrm{bol},\mathrm{dust}}1.47L_{FIR}`$, where $`L_{FIR}0.65L_{60}+0.42L_{100}`$, $`L_{60}\nu l_\nu `$ at $`\nu =c/60\mu `$m, and $`L_{100}\nu l_\nu `$ at $`\nu =c/100\mu `$m (e.g., Helou et al. 1988). The adopted relationship between $`L_{\mathrm{bol},\mathrm{dust}}`$ and $`L_{FIR}`$ is satisfied by the galaxies in the sample of §2.1 with an rms scatter of 7%. The estimated mean and rms of $`𝒦_{\mathrm{MIR}}`$ are not significantly altered by restricting our analysis either to galaxies with flux measurements in the sub-mm as well as at 60 and $`100\mu `$m or to the galaxies with the least noisy mid-IR spectra. Similar values of $`𝒦_{\mathrm{MIR}}`$ are observed in galaxies with a wide range of bolometric luminosities, as shown in Figure 4.
### 2.3 UV constraints on bolometric dust luminosity
The methods of estimating $`L_{\mathrm{bol},\mathrm{dust}}`$ discussed so far rely on fitting an average dust SED through the observed dust emission at a single wavelength. $`L_{\mathrm{bol},\mathrm{dust}}`$ can also be estimated without observing any dust emission at all. Many standard methods of estimating the dust content of galaxies rely on the absorption signature of dust in the UV/optical rather than its emission in the far-IR. We will focus our attention on one of these methods, which has been shown (see, e.g., MHC) to predict reasonably well the observed dust emission of local starburst galaxies.
The idea behind this technique is simple. Because dust absorbs shorter wavelengths more strongly than longer wavelengths, the UV continua of starbursts should become redder on average as the galaxies become dustier and the ratio of far-IR to far-UV luminosity increases. According to MHC, there is surprisingly little scatter in this trend among starburst galaxies in the local universe; the bolometric dust luminosities of the 57 starburst galaxies in their sample obey
$$L_{\mathrm{bol},\mathrm{dust}}=1.66(10^{0.4(4.43+1.99\beta )}1)L_{1600},$$
$`(4)`$
where $`L_{1600}\nu l_\nu `$ evaluated at $`\nu =c/1600`$Å and $`\beta `$ is the observed spectral slope ($`l_\lambda \lambda ^\beta `$) at $`1200\stackrel{<}{}\lambda \stackrel{<}{}2000`$Å, with a scatter of 0.3 dex. This correlation is illustrated in Figure 1.
### 2.4 Radio constraints on bolometric dust luminosity
Observations through the radio atmospheric window can also constrain $`L_{\mathrm{bol},\mathrm{dust}}`$. Increases in far-IR luminosity are accompanied, in local galaxies, by increases in radio emission with a typical spectral slope at $`\nu c/20\mathrm{c}\mathrm{m}`$ of $`f_\nu \nu ^{0.8}`$ (Condon 1992). The radio emission is thought to be synchrotron radiation from electrons in supernova remnants; it is proportional to $`L_{\mathrm{bol},\mathrm{dust}}`$ presumably because supernovae occur in the same population of massive stars responsible for heating the dust. According to Condon (1992), galaxies without active nuclei in the local universe satisfy
$$L_{\mathrm{bol},\mathrm{dust}}7.9\times 10^5L_{20},$$
$`(5)`$
where $`L_{20}\nu l_\nu `$ at $`\nu =c/20\mathrm{c}\mathrm{m}`$, with a scatter of 0.2 dex. This relationship is usually expressed in terms of the quantity $`L_{FIR}`$; the constant of proportionality in eq. 5 assumes $`L_{\mathrm{bol},\mathrm{dust}}1.47L_{FIR}`$, as discussed in §2.2.
The local radio/far-IR relation (eq. 5) is unlikely to hold to arbitrarily high redshifts. At some redshift the energy density of the cosmic microwave background, increasing as $`(1+z)^4`$, will become comparable to the magnetic energy density in star-forming galaxies, and relativistic electrons in supernova remnants will begin to cool through inverse Compton scattering rather than synchrotron emission. For magnetic energy densities comparable to those observed among rapidly star-forming galaxies in the local universe, this will occur at $`z\stackrel{>}{}`$ 3–6 (e.g. Carilli & Yun 1999).
We conclude this section with a brief digression. Carilli & Yun (1999) have recently proposed the ratio $`\eta f_\nu (850\mu \mathrm{m})/f_\nu (20\mathrm{c}\mathrm{m})`$ as a photometric redshift estimator. Results in this section allow us to quantify the estimator’s likely accuracy. Equations 2 and 5 imply
$$\eta \frac{3400(1+z)^{1\alpha }}{𝒦(850\mu \mathrm{m}/1+z)},$$
$`(6)`$
where $`\alpha 0.8`$ is the spectral index of the radio emission and $`𝒦(\lambda )`$ can be interpolated from table 1. Calculating the mean expected $`\eta `$ and its uncertainty as a function of redshift is complicated by the logarithmic uncertainties and by the presence of $`𝒦`$ in the denominator of eq. 6. To calculate the mean expected value of $`\eta `$ we used $`\overline{\eta }=3400(1+z)^{\alpha 1}1/𝒦(850\mu \mathrm{m}/1+z),`$ where $`1/𝒦`$ is the mean value of $`1/𝒦`$ among the galaxies in §2.1; to calculate the (asymmetrical) $`1\sigma `$ uncertainties in $`\eta `$ we assumed that $`\mathrm{log}_{10}(\eta )`$ was symmetrically distributed around $`\mu \mathrm{log}_{10}(\overline{\eta })\sigma _{\mathrm{dex}}^2\mathrm{log}_e10/2`$, where $`\sigma _{\mathrm{dex}}`$, the logarithmic spread in $`\eta `$ for galaxies at a given redshift, is the quadrature sum of the logarithmic uncertainty in $`𝒦(\lambda )`$ (from Table 1) and the 0.2 dex scatter in the radio/far-IR correlation. If the uncertainty in $`\mathrm{log}_{10}(\eta )`$ has a Gaussian distribution—as is roughly the case—this choice of $`\mu `$ correctly reproduces the desired mean $`\overline{\eta }`$. We will handle denominator terms and logarithmic uncertainties similarly throughout this paper unless stated otherwise.
Figure 5 shows the expected range of $`\eta `$ for star-forming galaxies as a function of redshift; redshift has been placed on the ordinate to allow redshift constraints to be easily read off this figure given a measured value of $`\eta `$. This figure can be used to estimate redshifts of a star-forming galaxies on the basis of their $`850\mu `$m and $`20`$cm fluxes, but the estimates will be accurate only if these galaxies obey the local radio to far-IR correlation (eq. 5) and have similar dust SEDs as the local sample of §2.1. Our next section is concerned, in part, with whether this is true (see also Carilli & Yun 2000).
## 3 OBSERVATIONS AT HIGH REDSHIFT
With very few exceptions, galaxies at high redshift are too faint to have been detected over the wavelength range $`20\mu \mathrm{m}\stackrel{<}{}\lambda \stackrel{<}{}200\mu \mathrm{m}`$ by any satellite that has flown. Their bolometric dust luminosities therefore cannot be directly measured; they can only be indirectly estimated with the correlations we have just described between $`L_{\mathrm{bol},\mathrm{dust}}`$ and luminosities at accessible wavelengths. Unfortunately the underlying physics responsible for these empirical correlations is poorly understood, and so there is no particular theoretical reason to expect them to exist in identical form at high-redshift. Nor can their existence be directly established empirically, for without measuring high-redshift galaxies’ fluxes at several wavelengths $`20\mu \mathrm{m}\stackrel{<}{}\lambda \stackrel{<}{}200\mu \mathrm{m}`$ it is impossible to show directly that their bolometric dust luminosities are correlated in the expected way with their fluxes through various atmospheric windows. It might therefore appear that the assumptions required to estimate high-redshift galaxies’ bolometric dust luminosities (and therefore star-formation rates) are not only questionable but also untestable; but this is not entirely true. If a galaxy’s fluxes in the rest-UV, mid-IR, sub-mm, and radio are each correlated with its bolometric dust luminosity, then they should also be correlated with each other in a way that is straightforward to calculate. Observing high-redshift galaxies at more than one of these wavelengths therefore provides an indirect test of standard methods for estimating $`L_{\mathrm{bol},\mathrm{dust}}`$. These indirect tests are the subject of this section.
Because our ultimate goal is to estimate the $`850\mu `$m fluxes of known UV-selected high-redshift populations from their UV colors, our emphasis in this section will be on testing whether high-redshift galaxies have the mid-IR, sub-mm, and radio fluxes that their UV-colors and the correlations in §2 would predict. This should not give the false impression that we are testing solely whether MHC’s $`\beta `$/far-IR relation is satisfied by high-redshift galaxies. Without directly measured bolometric dust luminosities, it is impossible to test for the existence of any one of the correlations of §2 independently of the others; they can only be tested in pairs. If $`z3`$ galaxies are found to have radio fluxes at odds with our predictions from their far-UV properties, for example, it might mean that they do not obey the local $`\beta `$/far-IR relation, but it could equally well mean that they do not obey the local radio/far-IR relation. What we are really testing in this section is whether high-redshift star-forming galaxies are sufficiently similar to their low redshift counterparts for their bolometric dust luminosities and star-formation rates to be reliably estimated in the absence of complete photometry at rest wavelengths $`20\mu \mathrm{m}\stackrel{<}{}\lambda \stackrel{<}{}200\mu \mathrm{m}`$.
The observations required for these tests are extremely difficult with current technology, and even though the data analyzed in this section are among the best available, much of the evidence presented is rather marginal. We have included even inconclusive evidence below when it illustrates what can (and cannot) be learned from currently feasible observations.
### 3.1 SMMJ14011
The lensed galaxy SMMJ14011+0252 at $`\alpha (\mathrm{J2000})=14^\mathrm{h}01^\mathrm{m}04^\mathrm{s}.97`$, $`\delta (\mathrm{J2000})=+02^o52^{}24^{\prime \prime }.6`$, $`z=2.565`$ (Smail et al. 1998, Frayer et al. 1999) is the only sub-mm–selected object robustly confirmed to be a high redshift galaxy without an obviously active nucleus (e.g. Ivison et al. 2000). $`450\mu `$m, $`850\mu `$m, and 20cm fluxes for this object were obtained by Ivison et al. (2000). We obtained rest-UV photometry through the $`G`$ and $``$ filters (Steidel & Hamilton 1993) with COSMIC (Kells et al. 1998) on the Palomar 200 inch Hale telescope in March 1999. Our optical photometry of SMMJ14011 is presented in Table 2.
Predicting SMMJ14011’s radio and sub-mm fluxes from its UV photometry requires an estimate of the slope of its UV continuum, $`\beta `$ (in $`f_\lambda \lambda ^\beta `$; see §2.3). Estimating $`\beta `$ from UV photometry is not entirely trivial for two reasons. First, the observed UV continua of high-redshift galaxies have been reddened by the intergalactic absorption of the Lyman-$`\alpha `$ forest, and this must be corrected before we can estimate their intrinsic continuum slopes $`\beta `$. Second, the UV continua of starburst galaxies are not exactly power laws, and so fitting their continua over different wavelength ranges can produce different estimates of the best fit slope $`\beta `$. MHC calculated $`\beta `$ for galaxies in their local sample by fitting these galaxies’ fluxes through ten “windows” spanning the wavelength range $`1260\mathrm{\AA }<\lambda <2600\mathrm{\AA }`$; nine of these windows were bluer than 1950Å. The available photometry of high-redshift galaxies rarely covers an identical wavelength range. At SMMJ14011’s redshift of $`z=2.565`$, for example, the $`G`$ and $``$ filters sample $`1350`$Å and $`1950`$Å rest. To estimate $`\beta `$ for SMMJ14011 in a way that took some account of these complications, we first subjected a GISSEL96 model of an actively star-forming galaxy (Bruzual & Charlot 1996) to the appropriate intergalactic absorption (Madau 1995), and then reddened the resulting spectrum with increasing dust extinction following a Calzetti (1997) law until this synthetic spectrum had the same $`G`$ and $``$ colors as SMMJ14011. Finally we fit this synthetic spectrum through the 10 windows of MHC with a power-law of the form $`l_\lambda \lambda ^\beta `$. The best fit $`\beta `$ we find—which is rather insensitive to the details of this procedure—is $`\beta =0.74`$, with a $`1\sigma `$ uncertainty due to photometric errors of $`\pm 0.25`$.
Equations 2–4 can now be used to predict SMMJ14011’s radio and sub-mm fluxes from its UV photometry. Defining the constants
$$𝒦_{850}(z)L_{\mathrm{bol},\mathrm{dust}}/\nu l_\nu \mathrm{at}\nu =c(1+z)/850\mu \mathrm{m},$$
$`(7a)`$
$$𝒦_{450}(z)L_{\mathrm{bol},\mathrm{dust}}/\nu l_\nu \mathrm{at}\nu =c(1+z)/450\mu \mathrm{m},$$
$`(7b)`$
$$𝒦_L(z,\alpha )L_{\mathrm{bol},\mathrm{dust}}/\nu l_\nu 7.9\times 10^5(1+z)^{(1+\alpha )}\mathrm{at}\nu =c(1+z)/20\mathrm{c}\mathrm{m},$$
$`(7c)`$
and
$$𝒦_{}(z,\beta )L_{1600}/\nu l_\nu (4.3/1+z)^{\beta +1}\mathrm{at}\nu =c(1+z)/6900\mathrm{\AA },$$
$`(7d)`$
where $`\alpha 0.85\pm 0.16`$ (Richards 2000) is the slope of the synchrotron emission (in $`f_\nu \nu ^\alpha `$), $`\beta `$ is the slope of the UV-continuum, as defined above, and 6900Å is approximately the central wavelength of the $``$ filter, and using $`f_\nu =10^{0.4(m_{\mathrm{AB}}+48.60)}`$ ergs/s/cm<sup>2</sup>/Hz, we find, after substituting into equations 2, 4, and 5, and changing from $`l_\nu `$ to $`f_\nu `$ with the equation $`l_{\nu _1}/l_{\nu _2}=f_{\nu _1}/f_{\nu _2}`$, the following relationships between the observed-frame optical, sub-mm, and radio fluxes of star-forming galaxies at $`z3`$:
$$f_{\nu ,\mathrm{obs}}(850\mu \mathrm{m})10^{0.4(22.17)}\left(\frac{𝒦_{}(z,\beta )}{1.}\right)\left(\frac{𝒦_{850}(z)}{10.}\right)^1(10^{0.4(4.43+1.99\beta )}1)\mathrm{mJy}$$
$`(8)`$
$$f_{\nu ,\mathrm{obs}}(450\mu \mathrm{m})10^{0.4(22.98)}\left(\frac{𝒦_{}(z,\beta )}{1.}\right)\left(\frac{𝒦_{450}(z)}{2.5}\right)^1(10^{0.4(4.43+1.99\beta )}1)\mathrm{mJy}$$
$`(9)`$
$$f_{\nu ,\mathrm{obs}}(1.4\mathrm{GHz})10^{0.4(23.86)}\left(\frac{𝒦_{}(z,\beta )}{1.}\right)\left(\frac{𝒦_L(z,\alpha )}{6.0\times 10^5}\right)^1(10^{0.4(4.43+1.99\beta )}1)\mu \mathrm{Jy}$$
$`(10)`$
with no dependence on cosmology. The redshift dependence is contained in the $`K`$-corrections $`𝒦_{450}`$, $`𝒦_{850}`$, $`𝒦_{}`$, and $`𝒦_L`$; the default values of $`𝒦_{850}=10`$, $`𝒦_{450}=2.5`$, $`𝒦_{}=1`$, $`𝒦_L=6.0\times 10^5`$ in this equation, derived from equation 7 and Table 1, are roughly appropriate for $`z3.0`$. The uncertainty in these equations is large. Eq. 4 is able to predict $`L_{\mathrm{bol},\mathrm{dust}}`$ from starbursts’ UV luminosities and spectral slopes with an uncertainty of about 0.3 dex. At fixed $`L_{\mathrm{bol},\mathrm{dust}}`$ there is (as argued in §2.1) an additional uncertainty of 0.1 dex in $`l_\nu `$ at $`100\mu \mathrm{m}450\mu \mathrm{m}/1+z`$, 0.3 dex at $`200\mu \mathrm{m}850\mu \mathrm{m}/1+z`$, and 0.2 dex at $`6\mathrm{c}\mathrm{m}20\mathrm{c}\mathrm{m}/1+z`$. If these logarithmic uncertainties add in quadrature, we would expect equations 8, 9, and 10 to be able to predict sub-mm and radio fluxes from UV fluxes to within 0.3–0.4 dex, if high-redshift galaxies resembled local starbursts exactly.
For a galaxy at $`z=2.565`$ with $`\beta =0.74`$, the appropriate $`K`$-correction constants for equations 8–10 are $`𝒦_{}=1.05`$, $`𝒦_L=6.13\times 10^5`$, and, interpolating between the values listed in Table 1, $`𝒦_{850}=15.3`$ and $`𝒦_{450}=2.6`$. These equations then predict $`850\mu \mathrm{m}`$, $`450\mu \mathrm{m}`$, 20cm fluxes for SMMJ14011 of $`23_9^{+14}`$ mJy, $`70_{28}^{+43}`$ mJy, $`150_{67}^{+95}\mu `$Jy. The quoted uncertainties reflect only the uncertainty in $`\beta `$ from photometric errors. An additional uncertainty of $`0.4`$ dex applies to each predicted flux, and so these predicted fluxes agree well with the measured fluxes (Ivison et al. 2000) $`f_\nu (850\mu \mathrm{m})=15\pm 2`$mJy, $`f_\nu (450\mu \mathrm{m})=42\pm 7`$mJy, $`f_\nu (20\mathrm{c}\mathrm{m})=115\pm 30\mu `$Jy (see Figure 6). SMMJ14011 evidently exhibits the same relationship between its UV, far-IR, and radio properties as rapidly star-forming galaxies in the local universe.
### 3.2 Lyman-break Galaxies
#### 3.2.1 Sub-mm
Although it is encouraging that one galaxy at $`z3`$, SMMJ14011+0252, appears similar to local galaxies in the correlations of its rest-frame UV, sub-mm, and radio fluxes, ideally we would like to establish that this holds true for entire populations of galaxies at $`z3`$. Unfortunately this is exceedingly difficult with current technology. None of the $`800`$ spectroscopically confirmed $`z3`$ galaxies in the UV-selected sample of Steidel et al. (2000), for example, has an $``$ magnitude as bright as SMMJ14011’s $`=21.25`$, and $``$ 95% of the galaxies in that sample are more than ten times fainter. Even if these galaxies were as heavily obscured as SMMJ14011, we would expect their $`850\mu `$m fluxes to be at least ten times lower than SMMJ14011’s, or $`\stackrel{<}{}1`$mJy, well below SCUBA’s detection limit. (SCUBA, on the James Clerk Maxwell Telescope, is the current state-of-the-art sub-mm bolometer array; see Holland et al. 1999). This suggests that (typical) individual Lyman-break galaxies are unlikely to be detected with SCUBA, and tests of whether these galaxies have the bolometric dust luminosities predicted from their UV spectral slopes will be at best statistical.
The faintness of typical Lyman-break galaxies makes their sub-mm and radio fluxes difficult to predict as well. Predicted bolometric dust luminosities depend sensitively on the UV spectral slope $`\beta `$, and this slope is difficult to estimate accurately for faint objects with relatively large photometric errors. For example, if a galaxy at $`z=2.6`$ had $`G=0.75`$, we would estimate (using the method of §3.1) $`\beta =0.60`$ and $`L_{\mathrm{bol},\mathrm{dust}}31\times L_{1600}`$, while if it had $`G=0.95`$ we would estimate $`\beta =0.07`$ and $`L_{\mathrm{bol},\mathrm{dust}}=85\times L_{1600}`$. A photometric error of 0.2 in $`G`$ is not unusual for Lyman-break galaxies, especially at fainter magnitudes (e.g. Adelberger et al. 2000), and so photometric errors can easily affect our predicted sub-mm and radio fluxes by a factor of $`3`$.
The situation is worsened by the fact that the only Lyman-break galaxies likely to be detectable in the sub-mm or radio are those with reddest UV slopes and highest dust obscurations. Sub-samples of Lyman-break galaxies selected for follow-up observations at these wavelengths will have to be chosen from among the reddest members of the population. Because highly reddened Lyman-break galaxies are much rarer than their less reddened counterparts (cf. Figure 12, discussed below), and because the typical photometric errors on Lyman-break galaxies are large, there is a reasonable chance that a Lyman-break galaxy which appears to be heavily reddened is actually a less reddened galaxy with photometric errors. These sub-samples will therefore suffer from significant Malmquist bias.
For these reasons, observing Lyman-break galaxies in the sub-mm does not appear to be an especially promising way to test whether these galaxies obey the $`\beta `$/far-IR correlation. Nevertheless Chapman et al. (2000) have attempted the observations. Their results are shown in Figure 7.
In order to calculate predicted sub-mm fluxes for these galaxies in a way that accounted for the significant photometric uncertainty and Malmquist bias discussed above, and for the expected intrinsic scatter in the UV/$`850\mu `$m relationship (eq. 8), we proceeded as follows. We first generated a large ensemble of simulated Lyman-break galaxies with $``$ magnitudes drawn at random from the $`z3`$ LBG luminosity function shown in Figure 12, UV slopes $`\beta `$ drawn at random from the distribution in Figure 12 (cf. Steidel et al. (1999), Adelberger et al. (2000)), and redshift $`z`$ drawn at random from the interval $`2.5<z<3.5`$. Each of these galaxies was assigned a $`G`$ magnitude by inverting the procedure of §3.1, and an $`850\mu `$m flux using eq. 8. The $`850\mu `$m fluxes were then changed randomly to reflect the scatter in the $`\beta `$/far-IR relation and in $`𝒦_{850}(z)`$. Finally, we placed each simulated galaxy at a random location in our $`G`$ and $``$ images and attempted to detect it and measure its photometry with the same software used to find real Lyman-break galaxies in these images. This produced a long list of the $``$ magnitudes and $`G`$ colors we would expect to measure for Lyman-break galaxies with known redshifts and $`850\mu `$m fluxes if the galaxies obeyed the local $`\beta `$/far-IR relation and had dust SEDs similar their local analogs. In order to predict the $`850\mu `$m fluxes shown in Figure 7, we binned together the $`850\mu `$m fluxes of simulated galaxies with magnitudes, colors, and redshifts similar to those of each galaxy in the sample of Chapman et al. (2000). As our best guess of the predicted $`850\mu `$m flux for each galaxy in the Chapman et al. sample, we took the sample mean of the $`850\mu `$m flux among simulated galaxies in its bin; for the uncertainty in the predicted flux we took the standard deviation of the $`850\mu `$m fluxes of these same simulated galaxies.
For simplicity we have indicated in Figure 7 the uncertainties in the predicted fluxes with the convention mean $`\pm `$ rms, but in fact these uncertainties are very skewed. The Monte Carlo simulations just described suggest that most Lyman-break galaxies should have $`850\mu `$m fluxes lower than the mean value and that some will have $`850\mu `$m fluxes significantly larger (how else could we have mean $``$ rms with the predicted fluxes bounded below by zero?), and so the observations are more consistent with the predictions than Figure 7 might suggest at first glance.
One way to quantify the agreement of the observations with our predictions is to calculate $`\chi ^2`$. Denoting the predicted fluxes by $`x`$ and the observed by $`y`$, the relevant $`\chi ^2`$ figure of merit,
$$\chi ^2\underset{i=1}{\overset{N}{}}\frac{(x_iy_i)^2}{\sigma _{x,i}^2+\sigma _{y,i}^2}$$
$`(11)`$
(e.g. Press et al. (1992, §15.3)) is equal to 14.5. There are 11 data points and no free parameters in this fit, and so the data are at least not grossly inconsistent with our expectations. It is difficult to make a more quantitative statement with this approach, however, because the error bars on the predicted flux are very skewed, and $`\chi ^2`$ is therefore not drawn from the usual distribution derived for Gaussian uncertainties.
An alternate approach is to ask what these data say about slope $`b`$ of the line $`f_{\mathrm{observed}}=bf_{\mathrm{predicted}}`$. If Lyman-break galaxies obeyed the $`\beta `$/far-IR correlation and had dust SEDs similar to those of local galaxies, we would expect the data to be consistent with $`b=1`$. A Bayesian approach can quantify the constraints on $`b`$ in a way that takes proper account of our skewed uncertainties. This approach requires us to calculate the likelihood of the data as a function of $`b`$. In the brief derivation below, $`f_m`$ and $`\sigma `$ denote measured $`850\mu `$m fluxes and uncertainties for the galaxies in Chapman et al. (2000), $`f_t`$ denotes these galaxies’ (unknown) true fluxes that would be measured if we had perfect data, $`f_t^{}f_t/b`$ denotes the true $`850\mu `$m fluxes of the simulated Lyman-break galaxies, calculated above under the assumption $`b=1`$, and $`I`$ is shorthand for the background knowledge (such as the luminosity and $`\beta `$ distributions of Lyman-break galaxies, and the scatter in the $`\beta `$/far-IR relation, and so on) that went into our Monte Carlo simulations. Our data set consists of the redshifts $`\{z\}`$, apparent magnitudes $`\{\}`$, colors $`\{G\}`$, and $`850\mu `$m fluxes $`\{f_m\}`$ of each galaxy the sample of Chapman et al. (2000). The likelihood of observing these data given $`b`$ is equal to the product of the likelihoods of observing each galaxy individually:
$$P(\{G\}\{\}\{z\}\{f_m\}|\{\sigma \}bI)=\underset{i}{\overset{N}{}}P((G)_i_iz_if_{m,i}|\sigma _ibI).$$
$`(12)`$
These individual likelihoods can be evaluated by integrating over the unknown true $`850\mu `$m flux of each galaxy:
$$P\left((G)zf_m|\sigma bI\right)=_0^{\mathrm{}}𝑑f_tP\left(f_t|bI\right)P\left((G)z|f_tI\right)P\left(f_m|f_t\sigma \right).$$
$`(13)`$
We have simplified this equation by omitting irrelevant variables from the conditional probabilities and by writing $`P(AB)`$ as $`P(A)P(B)`$ when $`A`$ and $`B`$ are independent. $`P\left(f_m|f_t\sigma \right)`$ can be calculated by assuming that the uncertainties in the measured fluxes are Gaussian, which is roughly true. The remaining probabilities in the integral can be estimated from the Monte Carlo simulations. Writing the total number of galaxies in the simulations as $`N_{\mathrm{tot}}`$, and the number of galaxies in the simulations with both properties $`A`$ and $`B`$ as $`n(A,B)`$, we have $`P\left(f_t|bI\right)n(f_t^{})/N_{\mathrm{tot}}`$ and $`P\left((G)z|f_tI\right)n(G,,z,f_t^{})/n(f_t^{})`$. Substituting into eq. 12 yields the likelihood, given $`b`$, of observing a single galaxy at redshift $`z`$ with UV photometry $``$ and $`G`$ and $`850\mu `$m flux $`f_m`$:
$$P\left((G)zf_m|\sigma bI\right)_0^{\mathrm{}}𝑑f_t^{}n(G,,z,f_t^{})\mathrm{exp}\left[\frac{1}{2}\left(\frac{bf_t^{}f_m}{\sigma }\right)^2\right].$$
$`(14)`$
The maximum likelihood value of $`b`$, calculated numerically using equations 12 and 14, is $`b=0.45`$. For a uniform prior, the 68% and 90% highest posterior density regions are $`0.22b0.87`$ and $`0.09b1.2`$. The sub-mm fluxes of Lyman-break galaxies therefore appear to be somewhat lower than we would have expected. This could indicate that Lyman-break galaxies do not quite satisfy MHC’s $`\beta `$/far-IR correlation, or that their dust SEDs are different from those of the local galaxies described in §2.1; but the inconsistency with $`b=1`$ is only marginally significant and we will not speculate further about its possible cause. All we can say with much confidence is that the sub-mm fluxes of Lyman-break galaxies are unlikely to be either much higher or more than an order of magnitude lower than our predictions.
Our interpretation of these data differs quantitatively from that of Chapman et al. (2000). This is because those authors ignored the uncertainties in the predicted fluxes in most of their analysis, and assumed that the dust SEDs of all Lyman-break galaxies would be well approximated by a single modified black-body (eq. 1) with $`T=50`$K and $`ϵ=1.5`$, rather than by the range of empirical SEDs estimated in §2.1.
#### 3.2.2 Radio
Radio observations of Lyman-break galaxies can in principle provide another indirect test of standard methods for estimating these galaxies’ bolometric dust luminosities. Extremely deep radio images are required, however, if we hope to put interesting limits on the radio fluxes of typical $`z3`$ galaxies: equation 10 implies that the vast majority of Lyman-break galaxies in the sample of Steidel et al. (2000) will have 20cm fluxes of only a few $`\mu `$Jy or less. One of the deepest existing radio images is Richards’ (2000) 20cm VLA map of the Hubble Deep and Flanking Fields, which reaches a $`1\sigma `$ noise limit of $`8\mu `$Jy/beam. Unfortunately even this image does not appear sufficient for a meaningful test. Of the 46 Lyman-break galaxies in our spectroscopic sample that lie within this 20cm image, only 2 are predicted (using the Monte-Carlo method of §3.2.1) to have marginally detectable fluxes $`f_\nu >10\mu `$Jy. These two, with predicted fluxes of $`12\pm 15\mu `$Jy and $`19\pm 17\mu `$Jy, have observed fluxes of $`4\pm 12\mu `$Jy and $`6\pm 12\mu `$Jy—not exactly in line with our predictions, but not damning either. (Our method of estimating the observed fluxes is described in §3.3.2 below.) Nine galaxies have marginally significant observed fluxes of $`f_\nu >10\mu `$Jy, and none of these nine were predicted to be especially bright; but six galaxies have measured fluxes $`f_\nu <10\mu `$Jy, and so even the marginally significant detections may simply reflect the noise characteristics of Richards’ (2000) image. Statistical tests based on the sample as a whole are as inconclusive as the object-by-object tests we have just described: the total observed and predicted 20cm fluxes from these 46 galaxies are $`105\pm 81\mu `$Jy and $`114\pm 36\mu `$Jy, respectively, and the formal 90% credible interval on $`bf_{\mathrm{observed}}/f_{\mathrm{predicted}}`$, estimated using the Monte Carlo simulations and Bayesian approach of §3.2.1, is $`0<b<2.1`$.
### 3.3 Balmer-break galaxies
Galaxies at redshifts as high as $`z3`$ are hardly ideal for testing whether the UV-properties of high-redshift galaxies can be used to predict their bolometric dust luminosities, because large luminosity distances and unfavorable $`K`$-corrections make them extremely faint in the optical, mid-IR, and radio (cf. Figure 1). A better redshift is $`z1`$. At this redshift galaxies are much brighter in these atmospheric windows, and their far-UV spectral slopes $`\beta `$ can still be measured with ground-based photometry. In this section we will use galaxies drawn from the $`z1`$ “Balmer-break” sample of Adelberger et al. (2000) to test whether actively star-forming galaxies at $`z1`$ follow the correlations between UV, mid-IR, and radio properties observed among galaxies in the local universe. This color-selected sample consists of $`700`$ star-forming galaxies with spectroscopic redshifts $`0.9\stackrel{<}{}z\stackrel{<}{}1.1`$ and apparent magnitudes $`\stackrel{<}{}25.5`$.
One warning is due before we begin. The $`\beta `$/far-IR correlation of MHC (eq. 4) has been shown to hold only for starburst galaxies in the local universe. It is not known whether less rapidly star-forming galaxies such as the Milky Way satisfy this correlation, and post-starburst galaxies, whose red UV slopes reflect age rather than dust content, almost certainly do not. At $`z3`$ galaxies are bright enough to satisfy the Lyman-break criteria of Steidel et al. (1999) only if they are forming stars rapidly ($`\stackrel{>}{}30M_{}`$/yr with typical dust obscurations, for $`h=0.7`$, $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$), and so these galaxies might reasonably be expected to obey the $`\beta `$/far-IR correlation (see MHC for further discussion of the Lyman-break galaxy/starburst connection). But at $`z1`$ post-starburst galaxies and slowly star-forming galaxies like the Milky Way can both satisfy our Balmer-break selection criteria, in addition to the starburst galaxies that are expected to obey the $`\beta `$/far-IR correlation. There is therefore good reason to suspect a priori that some galaxies in our Balmer-break sample will not obey this correlation.
In principle one might be able to select the starburst analogs from among a Balmer-break sample on the basis of their photometric or spectroscopic signatures, but a proper treatment is beyond the scope of this paper. Fortunately slowly star-forming galaxies will be far too faint in the radio and mid-IR to be detected in current data, and so their presence in our Balmer-break sample will not significantly affect the conclusions of this subsection. A crude way of partially excluding post-starbursts, which we shall adopt below, is to impose an additional selection criterion $`U_nG<0.5`$ on the Balmer-break sample. This limits the Balmer-break sample to the same range of $`\beta `$ observed among starburst galaxies in the local universe and among Lyman-break galaxies at $`z3`$ by excluding the galaxies that are reddest in the far-UV, whose slopes might reflect age rather than their dust content. We will discuss this further elsewhere (Adelberger et al. 2000).
#### 3.3.1 Mid-Infrared
Because of the strength of the $`7.7\mu `$m PAH feature, $`15\mu `$m observations provide a particularly promising way of observing dust emission from galaxies at $`z1`$. Seventy-one Balmer-break galaxies from our spectroscopic sample lie within the $`15\mu `$m (LW3) ISO image of the CFRS 1415+52 field obtained by Flores et al. (1999); six of them are listed as optical counterparts to $`15\mu `$m sources in these authors’ $`3\sigma `$ catalog.
Equations 3 and 4 imply the following relationship between the UV and mid-IR photometry of galaxies at $`z1`$:
$$f_{\nu ,\mathrm{obs}}(15\mu \mathrm{m})10^{0.4(U_n25.22)}\left(\frac{𝒦_U(z,\beta )}{1}\right)\left(\frac{𝒦_{15}}{20}\right)^1(10^{0.4(4.43+1.99\beta )}1)\mu \mathrm{Jy},$$
$`(15)`$
where
$$𝒦_{15}L_{\mathrm{bol},\mathrm{dust}}/\nu \overline{l}_\nu \mathrm{at}\nu =c(1+z)/15\mu \mathrm{m},$$
$`(16)`$
$$𝒦_U(z,\beta )L_{1600}/\nu l_\nu \left(\frac{1+z}{2.3}\right)^{\beta +1}\mathrm{at}\nu =c(1+z)/3700\mathrm{\AA },$$
$`(17)`$
$`\overline{l}_\nu `$ is the mean luminosity density at rest frequencies $`c(1+z)/18\mu \mathrm{m}<\nu <c(1+z)/12\mu \mathrm{m}`$, and 3700Å is roughly the central wavelength of the $`U_n`$ filter after accounting for atmospheric absorption and instrumental throughput. In most cases the observed $`15\mu `$m fluxes of $`z1`$ galaxies are dominated by emission from the 7.7 and $`8.6\mu `$m PAH lines, and $`𝒦_{15}`$ depends very weakly on redshift near $`z1`$. It is well approximated by $`𝒦_{\mathrm{MIR}}/2.422`$ for all the redshifts we will consider here ($`𝒦_{\mathrm{MIR}}`$ is defined in §2.2).
To estimate $`\beta `$ for these galaxies we used their $`U_nG`$ colors, which sample rest wavelengths $`1850`$Å and $`2350`$Å, and a procedure analogous to the one described in §3.1. Photometric errors are less of a concern for these optically bright galaxies than for the faint $`z3`$ galaxies of §3.2. Instead of taking the elaborate Monte-Carlo approach of §3.2, we estimated the predicted $`15\mu `$m fluxes for these galaxies by assuming that their true $`U_n`$ magnitudes and $`U_nG`$ colors were equal on average to the measured values; for the uncertainty in the predicted fluxes we simply propagated through the uncertainties in the $`U_nG`$ and $`U_n`$ fluxes determined from the Monte-Carlo simulations. The total uncertainty in the predicted fluxes, 0.5 dex, arises from the scatter in the $`\beta `$/far-IR correlation (0.3 dex), from the expected scatter in $`𝒦_{15}`$ (0.2 dex), from photometric uncertainties in $`U_n`$ magnitudes (0.1 dex), and from the uncertainty in $`\beta `$ due to photometric errors in $`U_nG`$ (0.3 dex).
Figure 8 shows the predicted and observed $`15\mu `$m fluxes of these 71 Balmer-break galaxies. The majority of these galaxies do not appear in the $`3\sigma `$ catalog of Flores et al. (2000), and so we know only that their fluxes were less than $`150\mu `$Jy; these sources are indicated with downward pointing arrows. The detections are indicated by points with error bars.
These data appear encouragingly consistent with our UV-based predictions: the large number of galaxies with predicted flux significantly lower than the detection limit are not detected, while the majority of those with predicted fluxes close to (or above) the detection limit are. The only apparently serious discrepancy between predicted and observed fluxes (ISO138 with $`f_{\mathrm{predicted}}4`$, $`f_{\mathrm{observed}}250`$), may result from a misidentification of the optical galaxy responsible for the $`15\mu `$m emission; according to Flores et al. (2000) there is a 27% chance of misidentification for this object, compared to a median 3% chance for the other five detections.
Readers who do not find Figure 8 persuasive may wish to consider the following. The seven galaxies that we predicted to be brightest have the following ranks in observed flux: 5, 6, $`>`$6, 3, $`>`$6, 1, 4, where undetected objects are assigned the rank “$`>`$6” because we know only that they are fainter than the six detected objects. Over 90% of the Balmer-break galaxies in the $`15\mu `$m image have the observed rank “$`>`$6”; if our UV-predicted $`15\mu `$m fluxes were unrelated to the observed fluxes, we would have expected $`90`$% of the seven objects predicted to be brightest to have the rank “$`>`$6”, but that is not the case. This type of reasoning can be quantified with a version of Kendall’s $`\tau `$ generalized for censored data sets by Oakes (1982): of 500 simulated data sets we have made by scrambling the observed rank assigned to each predicted rank, only 2 had a generalized $`\tau `$ as large or larger than the real data set. The predicted and observed fluxes of these galaxies evidently display a significant positive correlation; and moreover the fact that the handful of detections have observed fluxes broadly in line with their predicted fluxes suggests that this correlation is consistent with the expected linear correlation $`f_{\mathrm{predicted}}=f_{\mathrm{observed}}`$.
#### 3.3.2 Radio
Eighty-nine of the Balmer-break galaxies from the spectroscopic sample of Adelberger et al. (2000) lie within the deep 1.4GHz image of the Hubble Deep Field recently published by Richards (2000). We estimated rough 1.4GHz fluxes for this sample by summing the flux within a 2$`^{\prime \prime }.`$0 radius aperture surrounding the optical centroid of each object. The measured fluxes are shown in Figure 9. For the $`1\sigma `$ uncertainty we took $`12\mu `$Jy, the observed rms when we estimated fluxes in a similar manner for random locations in Richards’ (2000) image. The predicted fluxes were estimated using
$$f_{\nu ,\mathrm{obs}}(20\mathrm{c}\mathrm{m})10^{0.4(U_n24.19)}\left(\frac{𝒦_U(z,\beta )}{1}\right)\left(\frac{𝒦_L}{6.9\times 10^5}\right)^1(10^{0.4(4.43+1.99\beta )}1)\mu \mathrm{Jy},$$
$`(18)`$
which follows from equations 4, 5, 7c, and 17. Otherwise the predicted radio fluxes were derived in an identical manner to the predicted mid-IR fluxes in §3.3.1. The uncertainty in the predicted fluxes arises from the scatter in the $`\beta `$/far-IR correlation (0.3 dex), from the scatter in the far-IR/radio correlation (0.2 dex), from photometric uncertainties in $`U_n`$ magnitudes (0.1 dex), and from the uncertainty in $`\beta `$ due to photometric errors in $`U_nG`$ (0.4 dex). Added in quadrature these amount to a $`0.5`$ dex uncertainty in the predicted 20cm flux for each object.
Because of the large uncertainties in the predicted and observed fluxes, these data can hardly provide a rigorous test of whether $`z1`$ galaxies obey the correlations between far-UV, far-IR, and radio properties observed in rapidly star-forming galaxies in the local universe. However, despite the unimpressive appearance of Figure 9, the data do strongly suggest that the predicted and observed fluxes of Balmer-break galaxies are at least positively correlated: fewer than 1% of simulated data sets generated by randomly shuffling the observed and predicted fluxes had a Kendall’s $`\tau `$ as large as the observed data set. The trend of higher measured fluxes for objects with higher predicted fluxes is illustrated by the dotted crosses in Figure 9, which show the mean measured flux ($`\pm `$ the standard deviation of the mean) for objects in different bins of predicted flux. Although objects with higher predicted fluxes clearly tend to have higher observed fluxes, it appears that the measured 20cm fluxes may be somewhat lower than our predictions, especially for the brightest objects. The significance of this result is difficult to assess without running Monte-Carlo simulations similar to those discussed in §3.2. These simulations are not yet available.
### 3.4 HR10
The $`z=1.44`$ galaxy HR10 initially attracted the attention of Hu & Ridgway (1994) because of its unusually red optical-to-infrared color. Those authors suggested that HR10 was a high-redshift elliptical galaxy, but subsequent observations provided hints that HR10 was actively forming stars and suggested that its red colors might result from extreme dust reddening rather than from an aged stellar population (Graham & Dey 1996) This interpretation has received strong support from the recent detection of HR10 at $`450\mu `$m and $`850\mu `$m. HR10 is now considered the prototypical example of an extremely dusty star-forming galaxy at high redshift. Its photometry is presented in Figure 10 (Dey et al. 1999, Haynes et al. 2000).
Because of the large uncertainties in its rest-frame far-UV colors, we cannot reliably measure $`\beta `$ for HR10, but we can use its observed ratio of far-UV to far-IR luminosity to estimate the value of $`\beta `$ required for this object to satisfy MHC’s $`\beta `$/far-IR relation. This slope, shown in Figure 10, is not obviously inconsistent with the data.
Although it is unclear if HR10 satisfies the $`\beta `$/far-IR relation, we can say with confidence that its ratio of far-IR to UV luminosity far exceeds that of any other galaxy discussed in this section. Despite the large star-formation rate implied by its $`850\mu `$m flux, HR10 would not be included in most optically selected high-redshift surveys. Placed at $`z=3.0`$, for example, HR10 would be too faint and too red to satisfy commonly used “Lyman-break” photometric selection criteria (e.g. Steidel, Pettini, & Hamilton 1995). If most of the star formation in the universe occurred in objects like HR10, rather than in the relatively UV-bright Balmer-break and Lyman-break galaxies previously discussed, then UV-selected high-redshift surveys will provide a seriously incomplete view of star formation in the universe. But this seems an unlikely possibility to us, as we will now explain.
## 4 UV-SELECTED POPULATIONS, SUB-MM SOURCES, AND THE FAR-IR BACKGROUND
In §3 we tried to assess whether the star-formation rates of known high-redshift galaxies can be estimated from their fluxes through various atmospheric windows. Our conclusions were uncertain but hopeful: the meager available evidence is at least consistent with the idea that ground-based observations can be used together with locally calibrated correlations to provide rough constraints on high-redshift galaxies’ star-formation rates. But estimating moderately accurate star-formation rates is only one step towards the ultimate goal of high-redshift surveys, producing a reasonably complete census of star formation at high redshift that can be used to constrain our empirical and theoretical understanding of the early stages of galaxy formation. In this section we will set aside the object-by-object comparisons of star-formation rates estimated from photometry at different wavelengths, and turn our attention instead to the larger question of whether existing surveys are approaching the reasonably complete view of high-redshift star formation that we desire.
There is good reason to be concerned that the view of high-redshift star formation provided by these surveys is far from complete. Few star-forming galaxies in the local universe would be included in existing surveys if they were placed at high redshift. This is illustrated by Figure 11a, which shows the ratio of dust to ultraviolet luminosity for local galaxies with various star-formation rates. For this plot we have adopted $`L_{1600}+L_{\mathrm{bol},\mathrm{dust}}`$ as a crude measure of star-formation rate, though in some cases 1600Å luminosities and dust luminosities can be powered by processes other than star formation. Because of the different constants of proportionality in the conversions of $`L_{1600}`$ and $`L_{\mathrm{bol},\mathrm{dust}}`$ to star-formation rates, the quantity $`1.66L_{1600}+L_{\mathrm{bol},\mathrm{dust}}`$ (cf. eq. 4; see also MHC) would likely provide a better measure of star-formation rate, but in almost all cases $`L_{1600}`$ is small compared to $`L_{\mathrm{bol},\mathrm{dust}}`$ and our adoption of the simpler $`L_{1600}+L_{\mathrm{bol},\mathrm{dust}}`$ will not appreciably affect the subsequent discussion.
Solid circles on Figure 11a denote UV-selected starbursts from the sample of MHC, discussed earlier in the text; § symbols denote nearby spirals<sup>2</sup><sup>2</sup>2NGC247, NGC1232, NGC1313, NGC1398, NGC2403, NGC2683, NGC2976, NGC3556, NGC3623, NGC3726, NGC3877, NGC4178, NGC4216, NGC4254, NGC4303, NGC4307, NGC4321, NGC4522, NGC4559, NGC4565, NGC4592, NGC4595, NGC4653, NGC5055, NGC5248, NGC5457, NGC5806, NGC5879, NGC5907, NGC6946, selected randomly from among the objects present in both the Carnegie Atlas of Sandage & Bedke (1994) and the far-UV catalog of Rifatto, Lango, & Capaccioli (1995); open squares denote three LIRGs (NGC4793, NGC5256, and NGC6090) from the catalog of Sanders, Scoville, & Soifer (1991) which had UV fluxes estimated by Kinney et al. (1993) or Rifatto et al. (1995); and open stars denote the ULIRGs VII Zw 031, IRAS F12112+0305, and IRAS F22691-1808 observed in the far-UV by Trentham, Kormendy, & Sanders (1999). Bolometric dust luminosities were estimated from these galaxies’ $`60\mu `$m and $`100\mu `$m IRAS fluxes as discussed in §2.4. The adopted conversion from IRAS fluxes to bolometric dust luminosities is more appropriate for the starbursts, LIRGs, and ULIRGs in Figure 11a than for the spirals, but the resulting errors in the spirals’ estimated dust luminosities will not affect our conclusions significantly.
Taken together the galaxies in this plot are representative of those that host the majority of star formation in the local universe: according to Heckman (1998), starbursts and spirals account for roughly equal shares of the bulk of star formation in the local universe, while LIRGs and ULIRGs (galaxies with $`L_{FIR}\stackrel{>}{}10^{11}L_{}`$) account for perhaps 6% (Sanders & Mirabel 1996).
The curved lines on this plot show the $`z=3.0`$ completeness limits ($`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$) for existing surveys at optical (solid line), sub-mm (dashed line), and radio (dotted line) wavelengths. These lines assume an optical flux limit of $`f_\nu (1600\mathrm{\AA }\times (1+z))\stackrel{>}{}0.2\mu `$Jy ($`_{\mathrm{AB}}\stackrel{<}{}25.5`$), roughly appropriate for the Lyman-break survey of Steidel et al. (1999), a sub-mm flux limit of $`f_\nu (850\mu \mathrm{m})\stackrel{>}{}`$ 2mJy, the confusion limit of SCUBA on the JCMT, and a radio flux limit of $`f_\nu (20\mathrm{c}\mathrm{m})\stackrel{>}{}50\mu `$Jy, the completeness limit of Richards’ (2000) catalog of HDF sources. Observations with lower flux limits have been obtained in each of these bands, especially the optical—the flux limit in the Hubble Deep Field is nearly an order of magnitude fainter than the $`0.2\mu `$Jy optical limit we have adopted—but the flux limits above are roughly appropriate for the bulk of existing “deep” observations at each wavelength. Sub-mm and radio flux limits were converted into dust luminosity limits under the assumption that high-redshift galaxies follow the empirical correlations described in §2. The minimum luminosity for sub-mm detection is roughly independent of redshift, because $`K`$-corrections largely cancel out changes in luminosity distance, but the minimum luminosity for optical and radio detection would be (respectively) 7.4 and 13 times fainter at $`z1`$ (cf. Figures 11b and 11c), and 2.2 and 3.1 times brighter at $`z5`$. Evidently only a small fraction of star-forming galaxies in the local universe would be detected in existing surveys if they were placed at $`1\stackrel{<}{}z\stackrel{<}{}5`$.
Nevertheless large numbers of high-redshift galaxies have been found. Figure 11b shows the location on the star-formation rate versus dust obscuration plot of the majority of known galaxies at $`z1`$. Solid circles represent optically selected Balmer-break galaxies at $`0.8<z<1.2`$ from the sample of Adelberger et al. (2000), open stars represent $`15\mu \mathrm{m}`$ ISO sources at $`0.8<z<1.2`$ from the HDF sample discussed in §4.2 below, open squares (triangles) represent radio and sub-mm sources with estimated redshifts $`z<2`$ from the optically detected (undetected) sample of BCR, and the open circle represents HR10. Each object’s dust luminosity was estimated from its available ground-based photometry through the empirical correlations described in §2. In cases where an object’s dust luminosity could have been estimated with more than one of the empirical correlations, we took an estimate from the object’s flux in only one wavelength range, with the following order of preference: $`850\mu \mathrm{m}`$, $`15\mu \mathrm{m}`$, $`20\mathrm{c}\mathrm{m}`$, UV. The dust luminosities of ISO sources in the HDF, for example, were estimated from their $`15\mu \mathrm{m}`$ fluxes (§2.2) rather than from their UV spectral slopes (§2.3) or radio fluxes (§2.4). The curved lines in Figure 11b show typical $`z=1.0`$ detection thresholds for observations at different wavelengths. The optical, sub-mm, and radio thresholds assume the flux limits discussed above; the mid-IR limit assumes a $`15\mu \mathrm{m}`$ flux limit of $`75\mu `$Jy, which is roughly appropriate for the HDF catalog of Aussel et al. (1999). Because the objects in this plot do not lie at precisely $`z=1.0`$, and because small amounts of data exist with flux limits deeper than the typical values adopted here, some detected objects have star-formation luminosities $`L_{1800}+L_{\mathrm{bol},\mathrm{dust}}`$ lower than the indicated thresholds.
Figure 11c is a similar plot for galaxies at $`z3`$. The solid circles represent optically selected Lyman-break galaxies from the sample of Steidel et al. (2000), open circles represent the sub-mm sources SMMJ14011+0252 and west MMD11, the open squares (triangles) represent radio/sub-mm sources with estimated redshifts $`z>2`$ from the optically detected (undetected) sample of BCR, and the solid triangle is the optically undetected sub-mm source SMMJ00266+1708 with an estimated redshift $`z3.5`$ (Frayer et al. 2000). In this plot (and in Figure 17 below) the gravitational lensing magnification of SMMJ14011+0252 and SMMJ00266+1708 was roughly corrected by dividing their observed luminosities by 2.75 (Frayer et al. 1999) and 2.4 (Frayer et al. 2000) respectively.
At each of these redshifts, $`z0`$, $`z1`$, and $`z3`$, star-forming galaxies appear to have a similar range of dust obscuration $`0.1\stackrel{<}{}L_{\mathrm{bol},\mathrm{dust}}/L_{\mathrm{UV}}\stackrel{<}{}1000`$, and to follow a similar correlation of dust obscuration and star-formation rate, but the bolometric star-formation luminosities of detected galaxies increase steadily from $`z0`$ to $`z1`$ to $`z3`$. This is partly a selection effect, but it also reflects a genuine increase in the number density of rapidly star-forming galaxies at higher redshifts (e.g. Lilly et al. 1996, Steidel et al. 1999, BCR). Although existing high-redshift surveys are not deep enough (for the most part) to detect star-forming galaxies similar to those in the local universe, the observed increase in bolometric star-formation luminosity with redshift suggests that these surveys may nevertheless have detected a substantial fraction of high-redshift star formation.
The goal of this section is to estimate how large a fraction of high-redshift star formation they have detected. Have these surveys detected the majority of high-redshift star formation, or only the relatively small fraction that comparison to the local universe might make us expect? Although we will be primarily concerned with the possibility that dust obscuration has hidden large amounts of star formation from existing surveys, other physical effects—surface brightness dimming (e.g. Lanzetta et al. 1999) is an obvious example—can make high-redshift star formation undetectable as well. The results in this section will provide a rough limit on the amount of high-redshift star formation that is be undetected in existing surveys.
This limit can be estimated because the brightness of the $`850\mu `$m background provides an upper limit on the total amount of star formation at high redshift. Since $`K`$-corrections at $`850\mu `$m largely cancel out changes in luminosity distance, a galaxy with given bolometric dust luminosity will appear almost equally bright at $`850\mu `$m for any redshift in the range $`1\stackrel{<}{}z\stackrel{<}{}5`$ (e.g. Blain & Longair 1993). Taken together with the observation that star-forming galaxies emit the majority of their luminosities in the far-IR (cf. Figures 11a–c), this implies that a flux-limited $`850\mu `$m survey is nearly equivalent to star-formation limited survey at $`1\stackrel{<}{}z\stackrel{<}{}5`$, and that the integrated $`850\mu `$m background provides upper limit to the total amount of star formation at high redshift. It is an upper limit, rather than a measurement, because some fraction of the $`850\mu `$m background is known to be produced by objects other than star-forming high-redshift galaxies (e.g., AGN and low-redshift galaxies: Edge et al. 1999; Ivison et al. 2000).
Our strategy for placing limits on the amount of undetected star formation at high redshift will be to calculate the contribution to the $`850\mu `$m background from the various known high-redshift populations represented in Figures 11b and 11c, and see if there is significant shortfall between the background contribution from these galaxies and the total background that is observed.
### 4.1 Sub-mm sources
We begin by reviewing the contribution to the $`850\mu `$m background from resolved sub-mm sources. According to Barger, Cowie, & Sanders (1999), 20–30% of the $`850\mu `$m background is produced by objects brighter than SCUBA’s 2mJy confusion limit. Understanding the nature of these objects is difficult, because the large diffraction disk of SCUBA on the JCMT means that most have several plausible optical counterparts, but multi-wavelength observational programs (e.g. Frayer et al. 1999, Dey et al. 1999, BCR, Chapman et al. 2000, Frayer et al. 2000, Ivison et al. 2000) have established robust optical counterparts for a handful of them. This work has shown that approximately half of the detected $`850\mu `$m sources are associated with AGN (Ivison et al. 2000). Of the sources which do not appear to be AGN, three—SMMJ14011+0252 (Frayer et al. 1999), HR10 (Dey et al. 1999), and west MMD11 (Chapman et al. 2000)—are robustly associated with high-redshift star-forming galaxies. These galaxies, shown in Figures 11b and 11c, have relatively large dust obscurations $`30<L_{\mathrm{bol},\mathrm{dust}}/L_{\mathrm{UV}}<1000`$ and large implied star-formation rates, but their UV luminosities are comparable to those of typical optically selected galaxies at similar redshifts. BCR have shown that the brightest $`850\mu `$m sources tend to be even more heavily obscured: their analysis suggests that $``$ 75% (6/8) of $`850\mu `$m sources brighter than 6mJy have extreme dust obscurations $`300<L_{\mathrm{bol},\mathrm{dust}}/L_{\mathrm{UV}}<3000`$, (cf. Figures 11b and 11c). The ratio of dust to UV luminosity for these sources is very uncertain because their redshifts (and consequently luminosity distances) are unknown, but BCR’s radio/sub-mm photometric redshifts (cf. §2.4) would have to be seriously in error to bring their estimated dust obscurations into a less extreme range. Despite the large star-formation rates implied by their sub-mm fluxes, objects similar to those in the sample of BCR account for only a tiny fraction, $``$ 5% (Barger, Cowie, & Sanders 1999), of the measured $`850\mu `$m background (Fixsen et al. 1998).
The large dust obscurations observed in $`850\mu `$m sources have excited the interest of numerous authors (e.g. Smail et al. 1998; Hughes et al. 1998; Eales et al. 1999; Barger, Cowie, & Sanders 1999; Sanders 1999; BCR), who have rightly noted that if fainter high-redshift galaxies had similar levels of dust obscuration, the majority of high-redshift star formation would have occurred in objects that could not be detected in any existing survey (cf. Figures 11b and 11c). This popular argument for “hidden” star formation at high redshift is based on analysis of galaxies responsible for only a small fraction of the $`850\mu `$m background, however, and it ignores the strong correlation of dust obscuration and luminosity that is observed at low and high redshift alike (cf. Figure 11). Compelling arguments for or against the existence of hidden high-redshift star formation will require observations of galaxies with typical star-formation rates, not merely those with the extreme star-formation rates characteristic of the objects discussed in this subsection. Although some progress can be made at $`850\mu `$m through observations of lensed sources (e.g. Smail et al. 1998; Ivison et al. 2000), the best constraints currently come from observations at other wavelengths.
### 4.2 Mid-IR sources
At $`z1`$, $`15\mu \mathrm{m}`$ observations can provide a deeper view of dusty star formation than the sub-mm observations we have just discussed (cf. Figure 11b). The deepest existing large-area $`15\mu `$m image, a $``$ 20 square arcmin region centered on the Hubble Deep Field, was obtained by the ISO satellite in 1997. In this subsection we will review the properties of the dusty galaxies at $`z1`$ detected in the PRETI reduction (Aussel et al. 1999) of this data.
The primary (high significance) catalog of Aussel et al. (1999) contains 46 $`15\mu `$m sources. As discussed by Aussel et al. (1999) and Cohen et al. (2000), the majority of these sources ($`\stackrel{>}{}`$ 80%) can be reasonably associated with relatively bright optical counterparts ($`R\stackrel{<}{}23`$). Although some of the ISO sources may lie close to optically bright galaxies by chance, the fact that so many lie close to optically bright galaxies cannot be a coincidence. Spectroscopic redshifts for 40 of these optical counterparts are provided by Aussel et al. (1999) and Cohen et al. (2000); an additional redshift, $`z0.94`$ for HDF-PM3-22, was obtained by Adelberger et al. (2000). 16 of the optical counterparts<sup>3</sup><sup>3</sup>3HDF-PM3-1, HDF-PM3-6, HDF-PM3-8, HDF-PM3-9, HDF-PM3-10, HDF-PM3-11, HDF-PM3-19, HDF-PM3-20, HDF-PM3-21, HDF-PM3-22, HDF-PM3-25, HDF-PM3-31, HDF-PM3-34, HDF-PM3-37, HDF-PM3-39 and HDF-PM3-45 lie at redshifts $`0.8<z<1.2`$.
These 16 galaxies provide a reasonably complete sample of the $`z1`$ galaxies in this part of the sky that have the largest dust luminosities. The limiting dust luminosity for this sample is roughly a factor of three lower than the limit of blank-field SCUBA surveys. The sample is not exactly dust-luminosity limited, because of the scatter in the $`L_{MIR}/L_{\mathrm{bol},\mathrm{dust}}`$ relationship (§2.2), and it is not 100% complete, because redshifts have not been obtained for some of the $`15\mu `$m sources in this field<sup>4</sup><sup>4</sup>4If the $`15\mu `$m sources without redshifts have the same redshift distribution as those with redshifts, then perhaps 2 additional $`15\mu `$m sources in this field lie at the redshifts of interest $`0.8<z<1.2`$., but it is as close an approximation a complete, dust-luminosity sample at high redshift as currently exists.
The bolometric luminosities and dust obscurations of the galaxies in this sample are indicated in Figure 11b. Their typical dust obscurations, $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}100`$, are lower than those of the brighter sub-mm sources at similar redshifts. This trend of lower dust obscuration in fainter sources is also observed in the local universe (Figure 11a).
For the most part, the $`z1`$ ISO sources in the HDF appear to be drawn from among the same population of $`z1`$ galaxies that is routinely studied in the optical. Most of the sources are included in the (optical) magnitude-limited HDF survey of Cohen et al. (2000), for example, and 15 of 16 satisfy the Balmer-break photometric selection criteria of Adelberger et al. (2000) and are included in that $`z1`$ optical survey as well. They are not typical of the galaxies in optical populations, as the non-detection at $`15\mu `$m of most optically selected HDF galaxies shows, but Figure 11b suggests that the $`z1`$ ISO population can be naturally interpreted as the high luminosity, high obscuration tail of known $`z1`$ optical populations. Optically selected galaxies are thought to possess a wide range of dust obscurations, and the analysis of §4.3.1 (below) suggests that $`10_4^{+7}`$% have the obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}\stackrel{>}{}30`$ characteristic of ISO sources in the HDF. For $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$ the comoving number density of Balmer-break galaxies to $`U_n(AB)25.5`$ (roughly the faintest $`U_n`$ magnitude for HDF ISO sources) is $`3\times 10^2h^3`$ Mpc<sup>-3</sup>, and so optical observations would lead us to expect that $`3.0_{1.2}^{+2.0}\times 10^3`$ galaxies per comoving $`h^3`$ Mpc<sup>3</sup> would have dust obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}\stackrel{>}{}30`$ similar to those observed among ISO sources. This is roughly equal to the observed comoving number density of $`z1`$ ISO sources in the HDF (16 galaxies with $`0.8<z<1.2`$ in $`20`$ square arcmin, or $`2.4\times 10^3h^3`$ Mpc<sup>-3</sup> for $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$), showing that the $`z1`$ ISO population can be consistently interpreted as the dustiest part of the optical population.
A worrisome aspect of Figure 11b is that 10 of the $`z1`$ ISO sources are predicted to have detectable radio fluxes $`f_\nu (20\mathrm{c}\mathrm{m})\stackrel{>}{}50\mu \mathrm{Jy}`$, while only 2 were detected at this level by Richards (2000). Other $`z1`$ ISO sources are detected in Richards’ 20cm image at lower flux levels, but the total observed 20cm flux for the $`z1`$ ISO sources, after excluding HDF-PM3-6 and HDF-PM3-20 which have optical spectra suggesting the presence of an AGN (Cohen et al. 2000), is $`360\mu `$Jy, roughly a factor of 4 lower than the total predicted from the local correlations described in §2. The implication is that the bolometric dust luminosities of the ISO sources may be on average a factor of $`4`$ lower than indicated in Figure 11b, presumably because the ratio of mid-IR luminosity to total dust luminosity is systematically higher at $`z1`$ than in the local universe. This possibility only strengthens the main point we aimed to make in this section, that ISO sources have lower dust luminosities $`L_{\mathrm{bol},\mathrm{dust}}`$ and lower dust obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$ than the brighter SCUBA sources, and does not significantly affect the mid-IR/far-UV comparison of §3.3.1 which exploited the ranks of observed $`15\mu `$m fluxes rather than their absolute values; but it does suggest that estimates of the $`850\mu `$m background contribution of ISO sources will be subject to significant uncertainty.
The $`850\mu `$m background contribution of galaxies similar to ISO sources can be crudely estimated as follows. If we adopt a version of the mid-IR/$`L_{\mathrm{bol},\mathrm{dust}}`$ relationship (§2.2) that correctly predicts the radio fluxes of $`z1`$ ISO sources in the HDF, and use the sub-mm/$`L_{\mathrm{bol},\mathrm{dust}}`$ relationship of §2.1, then the total predicted $`850\mu `$m flux from the 16 $`z1`$ ISO sources in the HDF is $`8`$mJy, or $`0.4`$mJy / arcmin<sup>2</sup>, roughly 3% of the background measured by Fixsen et al. (1998). This number applies only to galaxies in the redshift range $`0.8<z<1.2`$, where ISO $`15\mu `$m observations are particularly sensitive due to the PAH features. Because little is known about the number density evolution of bright and highly obscured galaxies similar to ISO sources, it is difficult to estimate the contribution from galaxies outside this redshift interval, but a similar population of galaxies distributed uniformly between $`z=1`$ and $`z=5`$ would contribute $`40`$% ($`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$) of the total $`850\mu `$m background.
Because of the substantial overlap between ISO sources and galaxies in optical surveys, much of the $`850\mu `$m background contribution attributed above to ISO sources will be included as well in our calculation of the background contribution from optically selected galaxies, which we now describe.
### 4.3 Optical sources
Relatively few high-redshift galaxies have been detected by their dust emission, and the few that have apparently account for less than half of the $`850\mu `$m background. The vast majority of known high-redshift galaxies have been found in optical surveys. Can the large numbers of optically selected galaxies account for the remainder of the $`850\mu `$m background?
Figures 11b and 11c illustrate the relationship between optically selected galaxies and the $`850\mu `$m and $`15\mu `$m sources we have just discussed. The solid circles in Figure 11b represent $`z1`$ galaxies from the spectroscopic Balmer-break sample of Adelberger et al. (2000). Except in cases where their dust emission has been directly measured, the handful of Balmer-break galaxies with estimated dust obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}>100`$ were omitted from this plot due to the large uncertainties in their dust luminosities (cf. §3.3). Balmer-break galaxies with $`U_n(AB)>25.5`$ were also omitted, because at these faint magnitudes the Balmer-break sample suffers from severe selection effects. The solid circles in Figure 11c represent Lyman-break galaxies from the spectroscopic sample of Steidel et al. (2000). As described in the introduction to §4, the dust luminosities of the optically selected galaxies in Figures 11b and 11c were estimated from their dust emission in the few cases where it was measured, and from their UV photometry in the typical cases where it was not.
These figures support the common view (e.g. Eales et al. 1999; BCR; Frayer et al. 2000) that galaxies detected in the UV are typically far less luminous and less obscured than those detected in the IR. In the next sections we will see if UV-selected populations can make up in number what they lack in luminosity and still produce a significant contribution to the $`850\mu `$m background.
#### 4.3.1 The $`850\mu `$m luminosity function of Lyman-break galaxies
As a first step we will estimate the contribution to the $`850\mu `$m background from the best understood population of UV-selected high-redshift galaxies: the Lyman-break galaxies at $`z3`$. Figure 12 shows these galaxies’ apparent magnitude and $`\beta `$ distributions (Steidel et al. 1999, Adelberger et al. 2000). Adelberger et al. (2000) find no significant correlation between apparent magnitude and $`\beta `$ for Lyman-break galaxies in their ground-based sample ($`25.5`$; a scatter plot of UV-luminosity versus dust obscuration in $`z3`$ galaxies is presented in Figure 17 and discussed in §4.4 below) and so it is relatively simple to turn these two distributions into an $`850\mu `$m luminosity function. The result is shown in Figure 13. To produce this plot in a way that accounted for the various uncertainties, we first generated a large number of random realizations of Lyman-break galaxies’ apparent magnitude and $`\beta `$ distributions, consistent to within the errors with the distributions of Figure 12. We then picked at random a $`\beta `$ distribution and an apparent magnitude distribution from among these realizations, generated a long list of $``$, $`\beta `$ pairs, with $``$ drawn randomly from the best Schechter-function fit to the apparent magnitude distribution and $`\beta `$ from the $`\beta `$ distribution, assigned to each $``$, $`\beta `$ pair in the list an $`850\mu `$m flux with eq. 8, and adjusted each of these $`850\mu `$m fluxes at random by an amount reflecting the uncertainty in their predicted values (see §3). Binning this list and dividing by the appropriate volume produced one realization of the $`850\mu `$m luminosity function shown in Figure 13. We repeated the process, choosing a different realization of the $`\beta `$ and apparent magnitude distributions each time, until we had many binned realizations of the expected $`850\mu `$m luminosity function. The points shown in Figure 13 are the mean values among these many binned realizations; the uncertainties are the standard deviations. Also shown in Figure 13 are two related luminosity functions estimated in a similar manner: the expected $`60\mu `$m luminosity function of Lyman-break galaxies (similar in shape to their bolometric dust luminosity function) and the “dust corrected” UV luminosity function (derived by taking $`A_{1600}=4.43+1.99\beta `$ for the extinction in magnitudes at 1600Å, as suggested by MHC). The shapes of these two luminosity distributions, both essentially equivalent to the star-formation rate distribution, are similar to what is predicted by the simplest hierarchical models of galaxy formation (Adelberger 2000, Adelberger et al. 2000).
The systematic uncertainties in these luminosity functions are far larger than the random uncertainties. The most obvious systematic uncertainty is in the validity of our underlying assumption that the $`\beta `$/far-IR correlation will hold at high redshift. This uncertainty is difficult to treat quantitatively. If the $`\beta `$/far-IR correlation does not hold these results are meaningless; the data in §3 give us hope that it may hold; there is little else to say.
A second source of systematic uncertainty is the unknown shape of the apparent magnitude distribution at magnitudes fainter than $`27`$. Galaxies so faint in the UV may still have large star-formation rates if they are sufficiently dusty, and so our estimate of the number density of high star-formation rate galaxies depends to some degree on the unknown number density of these faint galaxies. In our calculation above we assumed that the $`\alpha 1.6`$ faint-end slope of the apparent magnitude distribution would continue to arbitrarily faint magnitudes. This extrapolation has a significant effect on our conclusions only at the faintest luminosities; Figure 13 shows the luminosities below which most of the estimated luminosity density comes from objects with $`>27`$. These parts of the luminosity functions should be viewed with considerable scepticism.
A related source of systematic uncertainty is our assumption that the $`\beta `$ distribution of Lyman-break galaxies is the same at all apparent magnitudes. This is known to be true only for $`<25.5`$ (Adelberger et al. 2000), and so the faintest points in our inferred luminosity functions are subject to further systematic uncertainties which are difficult to quantify. This source of systematic uncertainty is negligible only at the brightest luminosities, where most of the estimated luminosity density comes from galaxies with $`<25.5`$; see Figure 13.
The final source of significant systematic uncertainty is the shape of the $`\beta `$ distribution. Deriving this distribution from our data is difficult. It requires (among other things) a quantitative understanding of the selection biases that result from our photometric selection criteria and of the effects of photometric errors and Lyman-$`\alpha `$ emission/absorption on our measured broadband colors. Our current best estimate of the $`\beta `$ distribution is shown in Figure 12 (Adelberger et al. 2000). To give some idea of the systematic uncertainty in this distribution, we also show our previously published estimate (adapted from Steidel et al. 1999) and the $`\beta `$ distribution we would derive if we neglected photometric errors and incompleteness corrections altogether and simply corrected the measured broadband colors for the spectroscopically observed Lyman-$`\alpha `$ equivalent width. The most important difference between these three distributions (for our present purposes) is the shape of their red tails. The true shape almost certainly falls within the range they span; it is unlikely to be redder than the distribution of Steidel et al. (1999), and, since all known selection biases cause the Lyman-break technique to miss the reddest objects at $`z3`$, it cannot be bluer than the distribution that neglects completeness corrections. The luminosity functions that we would infer given these two limiting $`\beta `$ distributions are show in Figure 13 by curved lines. The systematic uncertainty in the $`\beta `$ distribution means that the mean extinction at 1600Å for Lyman-break galaxies, a factor of 6 in our best estimate, could lie between a factor of 5 and a factor of 9.
#### 4.3.2 UV-selected populations and the $`850\mu `$m background
In this section we will extend the preceding calculation to derive a crude estimate of the total contribution to the $`850\mu `$m background from known UV-selected populations at $`1<z<5`$. §4.3.1 described the uncertainties affecting our calculation of $`z3`$ Lyman-break galaxies’ contribution to the $`850\mu `$m background. Because relatively little is known about UV-selected populations at redshifts other than $`z3`$, the uncertainties in the present calculation will be much larger still. As a first approximation we can assume that these galaxies will have the same $`1800\mathrm{\AA }`$ luminosity distribution, the same $`\beta `$ distribution, and the same lack of correlation between $`\beta `$ and $`1800\mathrm{\AA }`$ luminosity as Lyman-break galaxies at $`z3`$. This is clearly an over-simplification, but the available data suggest that it might not be seriously incorrect. For example, Lyman-break galaxies at $`z4`$ appear to have approximately the same luminosity distribution and (possibly) $`\beta `$ distribution as Lyman-break galaxies at $`z3`$ (Steidel et al. 1999); Balmer-break galaxies at $`z1`$ have far-UV spectral slopes $`\beta `$ that lie mainly in the range $`2.2<\beta <0.5`$ observed in Lyman-break galaxies at $`z3`$ (Adelberger et al. 2000); and, like Lyman-break galaxies at $`z3`$, Balmer-break galaxies at $`z1`$ exhibit no significant correlation of $`\beta `$ with apparent magnitude at rest-frame $`1800\mathrm{\AA }`$ (Adelberger et al. 2000; cf. Figure 17, which we discuss below). The available data suggest further that the comoving star-formation density in UV-selected populations is roughly constant for $`1<z<5`$ (Steidel et al. 1999), and so the contribution to the $`850\mu `$m background from UV-selected high-redshift populations can probably be crudely approximated as the contribution that would arise from Lyman-break–like populations distributed with a constant comoving number density at $`1<z<5`$.
To estimate this contribution we simulated an ensemble of $`10^6`$ Lyman-break–like galaxies at $`1<z<5`$. Each galaxy was assigned a UV slope $`\beta `$ drawn randomly from the (shaded) $`\beta `$ distribution shown in Figure 12; a redshift $`z_i`$ drawn randomly from the interval $`1<z<5`$ with probability proportional to $`dV/dz`$ (where $`dV/dz`$ is the comoving volume per arcmin<sup>2</sup> per unit redshift in an $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$ cosmology); an apparent magnitude at rest-frame $`1600\mathrm{\AA }`$, $`m_{1600}`$, drawn randomly from the apparent magnitude distribution in Figure 12 shifted by the ($`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$) distance modulus between $`z_i`$ and $`z=3.0`$ and truncated at $`m_{1600}=27`$; and an observed-frame $`850\mu `$m flux with a version of eq. 8 appropriate to redshift $`z_i`$. Finally each $`850\mu `$m flux was adjusted at random by an amount reflecting the uncertainty in its predicted value (§3).
Figure 14 was produced by placing these simulated galaxies into bins of $`850\mu `$m flux and dividing by the appropriate area. The top panel shows the estimated contribution to the $`850\mu `$m background per logarithmic interval in $`f_\nu `$, $`f_\nu ^2n(f_\nu )`$, where $`n(f_\nu )`$ is the number of sources per mJy per square degree. The overall background expected from UV-selected populations in this crude calculation, $`4.1\times 10^4\mathrm{mJy}/\mathrm{deg}^2`$, is close to the measured background of $`4.4\times 10^4`$ (Fixsen et al. 1998). The closeness of the agreement is not especially meaningful, because the uncertainties in the calculation are so large, but nevertheless it is clear that the known UV-selected galaxies at high redshift could easily have produced the $``$ 75% of the $`850\mu `$m background that has not currently been resolved into discrete sources by SCUBA. There is little need to invoke large amounts of hidden high-redshift star formation in order to account for the brightness of the $`850\mu `$m background.
If the bulk of the $`850\mu `$m background is produced by galaxies similar to those in UV-selected surveys, as we have argued, then the shape of the $`850\mu `$m number counts should be similar to what our calculation suggests UV selected populations would produce. This appears to be the case. The bottom panel of Figure 14 compares the cumulative $`850\mu `$m number counts $`N(>f_\nu )`$ ($``$ the number of objects per square degree brighter than $`f_\nu `$) measured by Blain et al. (1999b) to the number counts that would be produced by the UV-selected high-redshift galaxies in our calculation. Although there are clear differences at the brightest end, where the observed number counts receive contributions from AGN, from low-redshift galaxies, and from extremely luminous high-redshift galaxies that may differ from the more normal galaxies detected in the UV, the observed number counts are remarkably close to what UV-selected populations would produce in the range $`0.3<f_\nu <3`$mJy that dominates the $`850\mu `$m background.
Figure 15 shows the expected contribution to the $`850\mu `$m background from objects with different apparent magnitudes at rest-frame 1600Å. This gives a rough idea of the apparent magnitudes we would expect to measure for sub-mm sources with various fluxes if rest-frame UV-selected populations obeying the local correlations between far-UV and far-IR fluxes produced the bulk of the $`850\mu `$m background. At any given 850$`\mu `$m flux density, objects are expected to have a wide range of redshifts and dust obscurations, and consequently of optical magnitudes, but even at the brightest $`850\mu `$m fluxes most objects are expected to be relatively faint ($`m_{1600}>24`$) in the optical. The observed faintness of optical counterparts to $`850\mu `$m sources is sometimes thought to show that the $`850\mu `$m background is produced by objects not present in rest-UV-selected surveys, but this argument has to be made carefully: UV-selected high-redshift galaxies are themselves quite faint in the optical, and as a result it is not sufficient to show (for example) that most optical counterparts are fainter than $`24^{\mathrm{th}}`$ magnitude; it must be shown instead that they are significantly fainter than Figure 15 predicts. The available data suggest that the sources responsible for brightest $`850\mu `$m number counts, $`f_\nu \stackrel{>}{}6`$mJy, are in fact significantly fainter than the predictions of Figure 15 (e.g. BCR; Frayer et al. 2000), and this is hardly surprising since our calculation shows that UV-selected populations cannot by themselves account for the brightest $`850\mu `$m number counts. Objects with larger obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}1000`$ are required and have been found. Measuring the optical magnitudes of $`1`$mJy sources would provide a better test of whether UV-selected populations contribute significantly to the far-IR background, but with current technology this is possible only for lensed sources and adequate data do not yet exist.
Also shown in Figure 15 is the contribution to the $`850\mu `$m background that would be produced by galaxies with $`27<m_{1600}<29`$ if the $`\alpha =1.6`$ faint-end slope of the luminosity function continued past the faintest observed magnitude of $`m_{1600}27`$ and if dust obscuration and reddening remained uncorrelated at the faintest magnitudes. Due to the steep faint-end slope, objects with $`m_{1600}>27`$ might be expected to contribute significantly to the $`850\mu `$m background. MHC have presented evidence, however, from Lyman-break galaxies with $`25.5\stackrel{<}{}m_{1600}<27`$ in the HDF, that galaxies fainter in the UV tend to have lower dust extinctions. Because apparent magnitude and dust obscuration are not correlated among $`z3`$ Lyman-break galaxies with $`m_{1600}<25.5`$, we assumed when producing Figures 14 and 15 that they would remain uncorrelated at fainter magnitudes as well. MHC’s analysis suggests that this assumption is incorrect: galaxies with $`25.5<m_{1600}<27`$ are probably somewhat less dusty than we have assumed and galaxies with $`m_{1600}>27`$ are probably significantly less dusty. When estimating the total background contribution from UV-selected populations (Figure 14), we assumed as a compromise that galaxies with $`25.5<m_{1600}<27`$ would be as dusty as galaxies with $`m_{1600}<25.5`$, and that galaxies with $`m_{1600}>27`$ would contain negligible amounts of dust. This is obviously a very crude solution to a complicated problem; the background contribution from objects with $`m_{1600}>25.5`$ is a significant source of systematic uncertainty in Figures 14 and 15.
Figure 15 would be more useful for observational tests if we showed the predicted magnitudes at fixed observed-frame wavelengths, say the $`i`$ band, rather than a fixed rest-frame wavelength, but this requires us to estimate $`K`$-corrections for these galaxies and we have not found an adequate model for their UV/optical SEDs. The problem is illustrated by Figure 16, which shows the measured rest-frame UV/optical SEDs of two Lyman-break galaxies detected at $`850\mu `$m, SMMJ14011+0252 (Smail et al. 1998) and west MMD11 (Chapman et al. 2000). The optical data were obtained as described in §3.1 and in Steidel et al. (2000); the near-IR photometry was obtained with NIRC (Matthews & Soifer 1994) on Keck I with excellent seeing ($`\stackrel{<}{}0{}_{}{}^{\prime \prime }.5`$) in May 1999. Neither SED is fit well by a model of a continuously star-forming galaxy subjected to varying amounts of dust following a standard extinction law. Moreover, even though SMMJ14011+0252 and west MMD11 have essentially indistinguishable SEDs in the UV—both are consistent with the same value of $`\beta `$—their optical SEDs are significantly different. The shapes of galaxies’ UV/optical SEDs are strongly dependent on their star-formation histories, and perhaps the simplest conclusion to draw from Figure 16 is that the star-formation histories of high-redshift galaxies are complicated and diverse. We have found no obvious way to predict their optical photometry from their UV photometry. See Shapley et al. (2000) for a more complete discussion.
This is a good place to remind readers that our UV-based estimates of high-redshift galaxies’ bolometric dust luminosities are not derived from an assumed intrinsic SED and dust reddening law; they are derived instead from the empirical (and somewhat mysterious) correlation between $`\beta `$ and $`L_{\mathrm{bol},\mathrm{dust}}/L_{1600}`$ described in §2.3. The diversity of observed SEDs makes it difficult to understand why this correlation should exist, but local (§2) and (less convincingly) high redshift (§3) observations suggest that it does.
### 4.4 Discussion
The major result of this section is our calculation showing that galaxies similar to those detected in UV-selected surveys probably produced a large fraction of the measured $`850\mu `$m background. Recent analyses of $`850\mu `$m surveys have reached the opposite conclusion (e.g. Smail et al. 1998, Hughes et al. 1998, Eales et al. 1999, Barger, Cowie, & Sanders 1999, Sanders 1999, BCR). Skeptical readers may wonder if our calculation is robust given its apparently heavy reliance on the poorly tested assumption (§3) that high-redshift galaxies’ dust luminosities can be estimated from their UV fluxes with the correlations of §2.
Although the details of our calculation certainly depend upon the assumption that high-redshift galaxies obey MHC’s $`\beta `$/far-IR relationship, the overall background contribution from UV-selected galaxies depends primarily upon the adopted mean value of $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$. MHC’s relationship implies that $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}1.66\times 10^{0.4A_{1600}}18`$ for galaxies at $`z3`$, and our calculation in §4.3 showed that this factor of 8 is sufficient for UV-selected galaxies to have produced the bulk of the $`850\mu `$m background. High-redshift galaxies may not satisfy MHC’s relationship, but the adopted $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}8`$ is still a plausible estimate of their mean dust obscuration. This mean obscuration is hardly unrealistically high; it is roughly equal to mean obscuration among the Lyman-break galaxies observed in the sub-mm by Chapman et al. (2000), for example, and it is lower than the mean obscuration ($`15`$) in the local UV-selected starburst sample of MHC. Even the local spirals in the sample discussed before §4.1 have $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}5`$; given the correlation of dust obscuration and star-formation rate that appears to exist at all redshifts (cf. Figures 11a–c), it is hard to imagine that a rapidly star-forming population at any redshift could have values of $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$ significantly lower than spirals in the local universe.
But the fundamental difference between our analysis and the analyses of previous authors who reached opposite conclusions is not simply the value of $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$ adopted for UV-selected galaxies; it is instead the assumed dependence of dust obscuration on luminosity among high-redshift galaxies. Arguments for large amounts of hidden star-formation at high redshift generally assume that dust obscuration and luminosity are independent. Data from low and high redshifts lead us to believe that $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$$`\stackrel{}{}`$$`L_{\mathrm{bol},\mathrm{dust}}`$ is a better approximation (cf. Figures 11a–c and 17). This correlation between dust obscuration and luminosity means that the $`f_\nu (850\mu \mathrm{m})1`$mJy sources that produce most of the $`850\mu `$m background are likely to be significantly less obscured than the brighter sources currently detectable in the sub-mm. For example, if the $`10`$mJy sources studied by BCR have $`200\stackrel{<}{}L_{\mathrm{bol},\mathrm{dust}}/L_{UV}\stackrel{<}{}2000`$ (cf. Figures 11b,c), then the correlation of dust obscuration and luminosity would suggest the 1mJy sources responsible for most of the $`850\mu `$m background have $`20\stackrel{<}{}L_{\mathrm{bol},\mathrm{dust}}/L_{UV}\stackrel{<}{}200`$. These sources would be dustier than typical optically selected galaxies, but they would nevertheless be as easy to detect in the optical as in the sub-mm. More generally, the correlation of dust obscuration with luminosity means that heavily obscured sources are not necessarily any harder to detect in the rest-frame UV than less obscured sources. Their higher UV obscurations are largely cancelled out by their higher intrinsic luminosities. This is illustrated by Figure 17, which shows the observed UV luminosities of sources with different dust obscurations at low and high redshift. Observed UV luminosities are similar for galaxies with dust obscurations $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}`$ spanning four orders of magnitude.
How robust is the claimed existence of the correlation between dust obscuration and luminosity? At low redshift, where objects’ dust luminosities are well constrained by IRAS measurements, its existence is indisputable. At higher redshifts, dust luminosities can only be estimated with the uncertain empirical correlations described in §2, and its existence is somewhat less secure. But even if none of the empirical correlations of §2 hold at high redshift, it is difficult to escape the conclusion that dust obscuration and luminosity are correlated. Without the correlations of §2 we would not know precisely where on the $`x`$-axis to place any object in Figure 17, for example, but we would still know that ISO sources are typically more luminous and more obscured than optical sources, and that SCUBA sources are typically more luminous and more obscured than ISO sources<sup>5</sup><sup>5</sup>5 For example, typical optical sources at $`z1`$ are fainter at $`15\mu `$m and 20cm than ISO sources at $`z1`$; typical $`z1`$ ISO sources must be fainter than typical SCUBA sources, or else they would over-produce the $`850\mu `$m number counts.; and that is enough to establish the existence of the correlation.
A final and more direct argument that UV-selected galaxies contribute significantly to the $`850\mu `$m background comes from the work of Peacock et al. (2000), who showed that positive background fluctuations in a deep $`850\mu `$m image of the HDF (Hughes et al. 1998) tend to be coincident with the locations of UV-selected high-redshift galaxies. Their detailed analysis suggests that UV-selected populations must have produced at least one quarter, and probably at least one half, of the total $`850\mu `$m background.
Taken together the preceding arguments make a reasonably strong case that the majority of high-redshift star formation is detectable in the deepest optical surveys. But this does not imply that IR/sub-mm observations are unnecessary. Although many of the results that have emerged from sub-mm surveys could have been inferred from UV data alone—the brightness of the $`850\mu `$m background, for instance, or the domination of the $`850\mu `$m background by $`1`$mJy sources (e.g. Blain et al. 1999b; cf. Figure 14), or the $``$ thousand-fold increase in number density of ULIRGs from $`z0`$ to $`z\stackrel{>}{}3`$ (e.g. Sanders 1999; cf. Figure 13)—some could not. For example, there was little indication from UV-selected surveys that extremely dusty galaxies with $`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}100`$ even existed (cf. Figure 12), let alone that they hosted significant fraction ($`\stackrel{>}{}5`$%) of high-redshift star formation. Observations in the IR will continue to play a central role in studies of high-redshift star formation, both as tests of the dust corrections that UV-selected surveys require and as the most efficient way to find the dustiest and most rapidly star-forming galaxies.
## 5 SUMMARY
The available evidence strongly suggests that dust in high-redshift galaxies will have absorbed most of the UV radiation emitted by their massive stars. Dust certainly absorbs most of the UV radiation emitted by rapidly star-forming galaxies in the local universe, and the ratio of the far-IR to optical backgrounds shows that the situation must have been similar throughout the history of the universe. Estimating the bolometric dust luminosities of high-redshift galaxies is therefore a crucial step in deriving their star-formation rates. The first part of this paper was primarily concerned with ways bolometric dust luminosities can be estimated even though most of the dust emission is blocked by the earth’s atmosphere before it can be detected on the ground.
We began in §2 by describing correlations between local galaxies’ bolometric dust luminosities and luminosities at more accessible far-UV, mid-IR, sub-mm, and radio wavelengths. These correlations provide the basis for attempts to estimate high-redshift galaxies’ bolometric dust luminosities. Much of this section was simply a review of previously published results; the minor new contribution was a quantification of the uncertainties involved in estimating bolometric dust luminosities from observations at a single rest wavelength $`25\mu \mathrm{m}\stackrel{<}{}\lambda \stackrel{<}{}600\mu \mathrm{m}`$ or $`\lambda 8\mu \mathrm{m}`$. The data suggest that radio, sub-mm, mid-IR, and far-UV estimates of local galaxies’ bolometric dust luminosities are all subject to a comparable uncertainty of 0.2–0.3 dex.
§3 was concerned with establishing whether the bolometric dust luminosities of high-redshift galaxies can be estimated from their fluxes through various atmospheric windows with the correlations of §2. The evidence presented was of uneven quality. The galaxy with the highest quality observations, SMMJ14011+0252 at $`z=2.565`$, obeyed the expected correlations nicely. The observations of more typical $`z3`$ galaxies had extremely low signal to noise ratios ($`\stackrel{<}{}1`$), and were largely inconclusive, though at least consistent with the expected correlations. Galaxies in our $`z1`$ sample were bright enough in the mid-IR and radio for us to show with high significance ($`>`$99%) that their predicted and observed fluxes in each of these bands were positively correlated. The data suggested moreover that this positive correlation was at least broadly consistent with the expected linear correlation $`f_{\mathrm{observed}}=f_{\mathrm{predicted}}`$. Taken together the data in this section make a plausible but still unconvincing case that the bolometric dust luminosities of high-redshift galaxies can be estimated with the expected accuracy using correlations calibrated in the local universe.
Our most important results were presented in the second part of this paper, §4, where we estimated the contributions to the far-IR background from different high-redshift populations. The brightness of the $`850\mu `$m background provides a rough measure of the total amount of star formation at high redshift, and a significant shortfall between the background contribution from known populations and the measured background would show that existing surveys have not detected the majority of star formation at high redshift. We did not find any evidence for a shortfall. The extremely luminous objects currently detectable in the sub-mm can account for $`25`$% of the far-IR background; the somewhat fainter galaxies detected in the mid-IR by ISO can account for perhaps an additional $`15`$%; and the galaxies detected in optical surveys can account for the remainder. Attributing the far-IR background to different galaxy populations in this way obscures the fact that there is substantial overlap between galaxy populations selected at different wavelengths, especially the mid-IR and optical, but the important point is that the brightness of the $`850\mu `$m background is similar to what known high-redshift populations would produce if they obeyed the local correlations of §2. This casts doubt on recent claims that the majority of high-redshift star formation is hidden from existing surveys (e.g. Lanzetta et al. 1999; BCR).
The analysis of §4 suggested that the bulk of the $`850\mu `$m background was produced by moderately obscured galaxies ($`1<L_{\mathrm{bol},\mathrm{dust}}/L_{UV}<100`$) similar to those that host most of the star formation in the local universe and to those that are detected in UV-selected high-redshift surveys. The brightness of the $`850\mu `$m background is sometimes cited as evidence that most of the star formation in the universe occurred in dusty objects missing from UV-selected surveys, but this is a mistaken conclusion: the bright $`850\mu `$m background shows only that a large fraction of star formation occurred in dusty objects, not that it occurred in the extremely dusty objects ($`L_{\mathrm{bol},\mathrm{dust}}\stackrel{>}{}100L_{UV}`$) that are easier to detect in the sub-mm than the far-UV. If we assumed that star formation at high redshift occurred only in known UV-selected populations and asked how dusty these galaxies would have to be to produce the observed $`850\mu `$m background, we would find that the mean required extinction at 1600Å was a factor of $`6`$—precisely the value that is generally inferred for UV-selected galaxies at $`z1`$ and $`z3`$ and similar to the values observed in spirals and UV-selected starbursts at $`z0`$. The brightness of the $`850\mu `$m background does not by itself require any star formation at all to have occurred in galaxies dustier than the moderately obscured ones ($`1<L_{\mathrm{bol},\mathrm{dust}}/L_{UV}<100`$) known to contain most of the star formation in the local universe.
This is not to say that extremely dusty galaxies ($`L_{\mathrm{bol},\mathrm{dust}}/L_{UV}>100`$) do not exist. Some clearly do (e.g. Dey et al. 1999; BCR; Frayer et al. 2000). Part of the far-IR background is produced by them. Our argument is simply that if high-redshift galaxies are similar to local galaxies in either their typical dust obscurations or their correlation of dust obscuration with luminosity, then UV-selected populations at high redshift must have contributed significantly to the $`850\mu `$m background (see §4.5). The fundamental difference between our analysis and the analyses of previous authors who reached opposite conclusions is the assumption made about the correlation between luminosity and dust obscuration. We have assumed that a correlation exists; they have assumed that it does not. The assumed existence of this correlation strongly affects the amount of high-redshift star formation that is detectable in the optical, since it implies first that the $`0.3<f_\nu (850\mu \mathrm{m})<3`$mJy sources which dominate the $`850\mu `$m background are likely to be less obscured than the brighter sources detectable with SCUBA, and second that the dustiest, most rapidly star-forming galaxies at high redshift are not necessarily any harder to detect in the UV than the less obscured sources forming stars at more moderate rates. A correlation similar to the one we have assumed is known to exist in the local universe, and in §4 we presented evidence that it exists at $`z1`$ and $`z3`$ as well. As illustrated by Figure 17, the high-redshift galaxies that have been detected by their dust emission are (with few exceptions) no fainter in the UV than the less obscured UV-selected galaxies at comparable redshifts.
Although we have argued that the majority of high-redshift star formation is detected in UV-selected surveys, and that many of the recent results from far-IR/sub-mm observations could have been predicted from UV observations alone (e.g. the brightness of the $`850\mu `$m background, the shape of the $`850\mu `$m number counts, the $``$ thousand-fold increase in the number density of ULIRGs at high redshift), we do not want to suggest that IR observations are unnecessary. Our arguments imply on the contrary that they are indispensable. Large corrections for dust extinction will be necessary in the interpretation of UV-selected surveys, and only IR observations can show whether the currently adopted corrections are valid or suggest alternatives if they are not. And the fact remains that IR/sub-mm surveys are the most efficient way to find the extremely luminous and dusty galaxies known to produce at least 5% of the $`850\mu `$m background.
It is commonly assumed that star formation at high redshift occurs in two distinct populations of galaxies, one relatively unaffected by dust and the other completely obscured. We favor a different view. High-redshift galaxies exhibit a continuum of dust properties. Some contain little dust, and some contain so much dust that they are nearly invisible in the optical; but the majority of star formation occurs in galaxies between these two extremes, in galaxies with $`1<L_{\mathrm{bol},\mathrm{dust}}/L_{UV}<100`$. The galaxies in this dominant population are undeniably dusty—most emit a larger fraction of their bolometric luminosities in the far-IR than in the far-UV—but with current technology they are easiest to detect in the rest-frame UV. Because in most cases only a small fraction of their luminosities emerge in the far-UV, large corrections for dust obscuration are required when estimating their star formation rates. A major purpose of this paper was to show that these corrections are tractable and produce sensible results. Although the validity of UV-derived dust corrections has not been convincingly demonstrated on an object-by-object basis, and further observations in the infrared are clearly required, our analysis suggests that the majority of high-redshift star formation occurs in objects which can at least be detected in existing UV-selected surveys. This is good news for those attempting to understand galaxy formation at high redshift, since it is likely that UV-selected surveys will continue to provide the most statistically robust information for at least the next decade.
This paper was made possible by several researchers who generously gave us access to their data during the early stages of its publication. We are grateful to D. Calzetti for her ISO measurements of dust emission from local starbursts, to E. Richards for his 20cm image of the Hubble Deep Field, and to T. Haynes, G. Cotter, and A. Bunker for the optical magnitudes of HR10. In addition M. Dickinson, M. Giavalisco, M. Pettini, A. Shapley, and R. Brunner helped obtain some of the data we have presented, and G. Taylor kindly provided us with an estimate of SMMJ14011’s radio flux prior to its publication by other authors. D. Elbaz recommended looking at the $`15\mu `$m fluxes of galaxies in our Balmer-break sample. D. Frayer and A. Barger are thanked for several informative conversations. Useful comments on an earlier draft were provided by A. Blain, M. Dickinson, D. Elbaz, M. Pettini and (especially) S. Lilly, the referee. CCS acknowledges support from the U.S. National Science Foundation through grant AST 95-96229, and from the David and Lucile Packard Foundation. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. |
warning/0001/nucl-th0001027.html | ar5iv | text | # Production of ϕ Mesons in Near-Threshold 𝝅𝑵 and 𝑵𝑵 Reactions
## I Introduction
The present interest in the $`\varphi `$ meson production in different elementary reactions is related to the strangeness degrees of freedom in the nucleon. Since the $`\varphi `$ meson is thought to consist mainly of strange quarks, i.e. $`s\overline{s}`$, with a rather small contribution of the light $`u`$ and $`d`$ quarks, its production should be suppressed if the entrance channel does not possess a considerable admixture of strangeness. Indeed, the recent experiments on the proton annihilation at rest (cf. for references and a compilation of data) point to a large apparent violation of the OZI rule , which is interpreted as a hint to an intrinsic $`s\overline{s}`$ component in the proton. However, the data can be explained as well by modified meson exchange models without introducing any strangeness component in the nucleon or OZI violation mechanisms. On the other hand, the analysis of the $`\pi N`$ sigma term suggests that the proton might contain a strange quark admixture as large as 20%. Thus this issue remains controversial. Therefore it is tempting to look for other observables that are sensitive to the strangeness content of the nucleon. Most of them are related to a possible strong interference of delicate $`s\overline{s}`$ knock-out and shake-off amplitudes and the “non-strange” amplitude which is caused by OZI rule allowed processes, or by processes wherein the standard OZI rule violation comes from the $`\varphi \omega `$ mixing.
As shown in Ref. , through this interference the polarization observables of the $`\varphi `$ photoproduction process are sensitive even to a rather small strangeness admixture in the proton. However, the only $`{}_{}{}^{3,1}S`$ $`s\overline{s}`$ configurations may be seen in a such process. The other configurations, such as $`{}_{}{}^{3,1}P`$, are suppressed by the selection rules and/or form factors. Contrary to this, Ellis et al. argue that the possibly dominant $`{}_{}{}^{3}P_{1}^{}`$ configuration might be seen in hadronic reactions.
Obviously, reliable information about the hidden strangeness manifestation in the $`\varphi `$ production in $`\pi p`$ and $`NN`$ reactions can be obtained only when the conventional, i.e. non-exotic, amplitude has been understood quantitatively. This is the objective of this work. The dominant conventional processes in $`\pi N`$ and $`NN`$ reactions are depicted in Figs. 1 and 2, where (a) is the mesonic exchange process being allowed by the finite $`\varphi \rho \pi `$ coupling strength and (b) is the direct $`\varphi `$ radiation from the nucleon legs, which is proportional to the finite $`\varphi NN`$ interaction strength. It should be emphasized that the process 1a is a subprocess in the diagram 2a, while the process 1b is a subprocess in the diagram 2b, when the exchanged boson is a pion.
While the diagrams in Figs. 1 and 2 look like usual Feynman diagrams it should be stressed they give a guidance of how to obtain from an interaction Lagrangian of hadronic fields a covariant parameterization of observable in strict tree level approximation. Additional ingredients are needed to achieve an accurate description of data within such a framework. In particular, the vertices needs to be dressed by form factors.
The early theoretical studies show, indeed, that predictions for hadronic observables are very sensitive to the parameters of the monopole form factors which can not be fixed unambiguously without adjustments relying on the corresponding experimental data. In our case one can rely on the recent measurement of the ratio of the total cross sections of $`\varphi `$ and $`\omega `$ production in $`pp`$ reactions studied by the DISTO Collaboration at SATURNE at $`\mathrm{T}_{\mathrm{lab}}=2.85`$ GeV as well as on the $`\varphi `$ angular distribution in the $`pppp\varphi `$ reaction and on the total cross section .
In the (sub)threshold region also in heavy-ion reactions the $`\varphi `$ production data is accessible via the $`K^+K^{}`$ decay channel studied with the $`4\pi `$ detector FOPI at SIS in GSI/Darmstadt . However, here an understanding of elementary hadronic reactions serves as a prerequisite for interpreting the data. Rather the upcoming proton and pion beam experiments with the HADES detector system at SIS in GSI offer a chance to enlarge the data base. In particular, HADES can identify the $`\varphi `$ also via the $`e^+e^{}`$ channel.
Finally we mention that for pion-induced reactions at the proton also near-threshold data for the $`\varphi `$ production are available .
An important step towards an understanding of the structure of the “non-strange” $`pppp\varphi `$ reaction mechanisms was made recently . The focus of Ref. is the determination of the parameters of the direct $`\varphi NN`$ interaction, thus reducing the above mentioned uncertainties, by analyzing the combined the $`pppp\omega `$ and $`pppp\varphi `$ reactions and the corresponding DISTO data at a given beam energy; for the reaction $`pppp\omega `$ just the same mechanism is assumed as those for $`pppp\varphi `$, as shown in Fig. 2. Clearly, at one given beam energies the excess energy for both reactions are quite different.
In this paper we therefore attempt a different approach with a similar goal. We reduce the uncertainties of the reaction mechanism by a combined study of the to each other related reactions $`pppp\varphi `$ and $`\pi ^{}pn\varphi `$ using the known data within the same interval of excess energies of $`20100`$ MeV . For the absolute normalization of the angular distribution in $`pppp\varphi `$ we use the recently published total cross sections of the reaction $`pppp\varphi `$ . It turns out that this value is compatible, within given error bars, with an extrapolation of the previously measured ratio of the total cross sections of $`pppp\varphi `$ to $`pppp\varphi `$ reactions, $`\sigma _\varphi /\sigma _\omega =(3.7\pm 0.5)\times 10^3`$ at $`T_{lab}=2.85`$ GeV , by normalizing it to the old bubble chamber data for $`\sigma _\omega `$ at various excess energies . The extrapolated value of the cross section differs from the new value thus influencing to some extent adjusted parameters.
In comparison with previous works we are doing the next step towards an understanding of the dynamics of $`\varphi `$ production in hadronic reactions. We present a systematical analysis of $`\pi ^{}pn\varphi `$, $`pppp\varphi `$ and $`pnpn\varphi `$ reactions in the near-threshold region where the destructive interferences between the two mechanisms (a) and (b) in Figs. 1 and 2 are essential. We are going to show that basically there are two different sets of the model parameters. One of them corresponds to the case when the mesonic exchange channel (a) is dominant for the $`\pi ^{}pn\varphi `$ reaction (Fig. 1), and in the second case the direct emission mechanism (b) is dominant. For both sets of solutions we calculate the total and differential cross sections and spin density matrix, responsible for the $`\varphi e^+e^{}`$ and $`\varphi K^+K^{}`$ decay angular distributions and show in which observables the direct $`\varphi NN`$ interaction might be clearly manifest. We present also a combined analysis of $`pppp\varphi `$ and $`pnpn\varphi `$ reactions at a finite energy excess with taking into account the final state interaction and analyze the deviation of predicted observables from the pure threshold values which is important for the future understanding of the role of an internal strangeness degrees of freedom in the nucleon. For this aim we study the beam-target spin asymmetry and the relative role of the singlet and triplet states in the entrance channel.
Our paper is organized as follows. In Section II, we define the kinematical variables and formulae for calculating the cross sections and polarization observables. The basic amplitudes for the mechanisms illustrated in Figs. 1 and 2 are given explicitly in Section III. In Section IV we discuss results of numerical calculations and predictions. The summary is given in Section V. In the Appendix we describe the formalism for the enhancement factors of final state interaction within the framework of the Jost function and the effective phase-equivalent potentials.
## II Observables
The differential cross section of the reaction $`\pi ^{}p\varphi n`$ (cf. Fig. 1) has the obvious form
$`{\displaystyle \frac{d\sigma }{d\mathrm{\Omega }_\varphi }}={\displaystyle \frac{1}{64\pi ^2s}}{\displaystyle \frac{|𝐪|}{|𝐩_\pi |}}\overline{|T_{(1)}|^2},`$ (1)
where $`p_\pi =(E_\pi ,𝐩_\pi )`$ and $`q=(E_\varphi ,𝐪)`$ are the four-momenta of the pion and the $`\varphi `$ meson in the center of mass system (c.m.s.); $`\overline{|T_{(1)}|^2}`$ means average and sum over the initial and final spin states, respectively.
The differential cross section of $`\varphi `$ production in the reaction $`a+bc+d+\varphi `$, where $`a,b`$ and $`c,d`$ label the incoming and outgoing nucleons (cf. Fig. 2), is related to the invariant amplitude $`T_{(2)}`$ as
$`d\sigma ={\displaystyle \frac{1}{2(2\pi )^5\sqrt{s(s4M_N^2)}}}\overline{|T_{(2)}|^2}{\displaystyle \frac{d𝐩_c}{2E_c}}{\displaystyle \frac{d𝐩_d}{2E_d}}{\displaystyle \frac{d𝐪}{2E_\varphi }}\delta ^{(4)}(P_iP_f).`$ (2)
where $`p_n=(E_n,𝐩_n)`$ with $`n=a,b,c,d`$ are the four-momenta of the nucleons in the c.m.s., $`\sqrt{s}=E_a+E_b`$ is the total c.m.s. energy, $`P_{i,f}`$ are the total four-momenta of the initial or final states. Hereafter $`\theta `$ denotes the polar $`\varphi `$ meson angle and $`\mathrm{\Omega }`$ is its solid angle. We use a coordinate system with $`𝐳𝐩_a`$, $`𝐲𝐩_a\times 𝐪`$. Among the five independent variables for describing the final state we choose $`E_\varphi ,\mathrm{\Omega }`$ and $`\mathrm{\Omega }_c`$. Then the energy $`E_c`$ of particle $`c`$ is defined by $`E_c=\frac{ABC\sqrt{B^2M_N^2(A^2C^2)}}{A^2C^2},`$ with $`A=2(\sqrt{s}E_\varphi )`$, $`B=s2E_\varphi \sqrt{s}+M_\varphi ^2`$, $`C=2|𝐪|\mathrm{cos}\theta _{qp_c}`$, Finally, the fivefold differential cross section reads
$`{\displaystyle \frac{d^5\sigma }{dE_\varphi d\mathrm{\Omega }d\mathrm{\Omega }_c}}={\displaystyle \frac{1}{8(2\pi )^5\sqrt{s(s4M_N^2)}}}\overline{|T_{(2)}|^2}{\displaystyle \frac{|𝐪||𝐩_c|^2}{|A|𝐩_c|+CE_c|}}.`$ (3)
The total and/or partially differential cross sections are found by integration over the available phase space.
In this paper we consider two polarization observables. One of them is the spin density matrix which describes the spin structure of the outgoing $`\varphi `$ meson,
$`\rho _{rr^{}}={\displaystyle \frac{_\beta T_{r,\beta }T_{r^{},\beta }^{}}{_{r,\beta }T_{r,\beta }T_{r,\beta }^{}}},`$ (4)
where $`rm_\varphi =\pm 1,0`$ are the spin projections of the $`\varphi `$ meson, and $`\beta `$ denotes a set of unobserved quantum numbers. The spin density defines the angular distribution in $`\varphi e^+e^{}`$ and $`\varphi K^+K^{}`$ decays, which has a simple form in a system where the $`\varphi `$ meson is at rest (for details see ). The decay angles $`\mathrm{\Theta }`$, $`\mathrm{\Phi }`$ are defined as polar and azimuthal angles of the direction of flight of one of the decay particles in the $`\varphi `$ meson’s rest frame. The decay distributions integrated over the azimuthal angle $`\mathrm{\Phi }`$, $`𝒲(\mathrm{cos}\mathrm{\Theta })`$, depend only on the diagonal matrix elements $`\rho _{00},\rho _{11}=\rho _{11}`$, normalized as $`\rho _{00}+2\rho _{11}=1`$, according to
$`𝒲(\mathrm{cos}\mathrm{\Theta })`$ $`=`$ $`{\displaystyle \frac{3}{2(B+3)}}\left(1+B\mathrm{cos}^2\mathrm{\Theta }\right),`$ (5)
where the $`\varphi `$ decay anisotropies $`B`$ read
$`B^{K^+K^{}}={\displaystyle \frac{13\rho _{00}}{1\rho _{00}}},B^{e^+e^{}}={\displaystyle \frac{13\rho _{00}}{1+\rho _{00}}}.`$ (6)
To exclude the kinematical dependence of $`\rho _{00}`$ or $`B`$ on the $`\varphi `$ meson production angle, we choose the quantization axis along the $`𝐳`$ direction (in the $`\varphi `$ rest system), and using the corresponding Wigner rotation functions $`d_{\mathrm{}}^{\mathrm{}}(\chi )`$ one gets the amplitudes $`T_{r,\beta }`$ in Eq. (4) by
$`T_{m_\varphi ,\beta }^z={\displaystyle \underset{\lambda ,\beta _i^{}}{}}T_{\lambda ,\beta ^{}}^{c.m.s.}d_{\lambda ,m_\varphi }^1(\chi _\varphi ){\displaystyle \underset{i}{}}d_{\beta _i^{},\beta _i}^{\frac{1}{2}}(\chi _i),`$ (7)
where only $`\chi _\varphi =\theta `$ is important, while the other $`\chi _i`$’s disappear in Eq. (4).
Another polarization observable is the beam-target asymmetry in the $`NNNN\varphi `$ reactions which is related to the nucleon spin states via
$$C_{BT}=\frac{d\sigma (S_i=1)d\sigma (S_i=0)}{d\sigma (S_i=1)+d\sigma (S_i=0)},$$
(8)
where $`S_i`$ is the total spin in the entrance channel. It is important to note that spin and parity conservation arguments result in a precise model independent prediction for the beam - target asymmetry: $`C_{BT}=1`$ for the $`pppp\varphi `$ reaction at the threshold. In the $`pnpn\varphi `$ reaction the asymmetry depends on the relative weights of the triplet and singlet states in the entrance channel.
## III Basic amplitudes
Basically, our consideration in this section is similar to the previous study (for the pure mesonic exchange contributions depicted in Figs. 1a and 2a) and to the models for both channels shown in Figs. 1 and 2. The difference between this work and previous ones is in the different form of cut-off form factors for the off-shell nucleons in direct $`\varphi `$ emission (cf. Figs. 1b and 2b) and a different choice of the cut-off parameters in $`\pi N`$ and $`NN`$ interactions which we will discuss below in detail. In spite of the mentioned similarity, for completeness in discussing our predictions for the set of observables which have not been considered before, in this section we display the main formulae which define the basic amplitudes. The meson - nucleon and the $`\varphi \rho \pi `$ interaction Lagrangians read in standard notation
$`_{MNN}`$ $`=`$ $`ig_{\pi NN}\overline{N}\gamma _5𝝉𝝅N`$ (11)
$`g_{\rho NN}\left(\overline{N}\gamma _\mu 𝝉N𝝆^\mu {\displaystyle \frac{\kappa _\rho }{2M_N}}\overline{N}\sigma ^{\mu \nu }𝝉N_\nu 𝝆_\mu \right)`$
$`g_{\varphi NN}\left(\overline{N}\gamma _\mu N\varphi ^\mu {\displaystyle \frac{\kappa _\varphi }{2M_N}}\overline{N}\sigma ^{\mu \nu }N_\nu \rho _\mu \right),`$
$`_{\varphi \rho \pi }`$ $`=`$ $`g_{\varphi \rho \pi }ϵ^{\mu \nu \alpha \beta }_\mu \varphi _\nu \mathrm{Tr}(_\alpha \rho _\beta \pi ),`$ (12)
where $`\mathrm{Tr}(\rho \pi )=\rho ^0\pi ^0+\rho ^+\pi ^{}+\rho ^{}\pi ^+`$, and bold face letters denote isovectors. All coupling constants with off-shell meson are dressed by monopole form factors $`F_i=(\mathrm{\Lambda }_i^2m_i^2)/(\mathrm{\Lambda }_i^2k_i^2)`$, where $`k_i`$ is the four-momentum of the exchanged meson. Following the scheme of the meson photoproduction we assume that $`\varphi NN`$ vertices must be dressed by form factors for off-shell virtual nucleons. But this might result in a violation of the transversality of the amplitude with respect to the $`\varphi `$ meson field. To avoid this problem we use the prescription of Ref. and parameterize the product of the two form factors appearing in the left and the right diagrams in Figs. 1b and 2b in a symmetrical form
$`F_N(p_l,p_r)={\displaystyle \frac{1}{2}}\left({\displaystyle \frac{\mathrm{\Lambda }_N^4}{\mathrm{\Lambda }_N^4+(p_l^2M_N)^2}}+{\displaystyle \frac{\mathrm{\Lambda }_N^4}{\mathrm{\Lambda }_N^4+(p_r^2M_N)^2}}\right);`$ (13)
here $`p_l(p_r)`$ is the four-momentum of the virtual nucleon in the left (right) diagrams in Figs. 1b and 2b.
### A $`𝝅𝑵\mathbf{}𝑵\mathit{\varphi }`$ reaction
The invariant amplitude for the meson exchange channel (a) in Fig. 1 reads
$`T_{(1a)\lambda }=K^{\pi N}ϵ^{ijkl}\left[\overline{u}(p_c)\mathrm{\Gamma }_{\rho l}(k_\rho )u(p_a)\right]q_ik_k\epsilon _j^\lambda I_\pi ,`$ (14)
where
$`\mathrm{\Gamma }_\rho ^i(k)=\gamma ^i+i{\displaystyle \frac{\kappa _\rho }{2M_N}}\sigma ^{ij}k_{\rho j},`$ (15)
$`K^{\pi N}(k_\rho )={\displaystyle \frac{g_{\rho NN}g_{\varphi \rho \pi }}{k_\rho ^2m_\rho ^2}}{\displaystyle \frac{\mathrm{\Lambda }_{\rho NN}^2m_\rho ^2}{\mathrm{\Lambda }_{\rho NN}^2k_\rho ^2}}{\displaystyle \frac{\mathrm{\Lambda }_{\varphi \rho \pi }^{\rho \mathrm{\hspace{0.17em}2}}m_\rho ^2}{\mathrm{\Lambda }_{\varphi \rho \pi }^{\rho 2}k_\rho ^2}}`$ (16)
with $`k_\rho =p_cp_a`$ as the virtual $`\rho `$ meson’s four-momentum; $`\epsilon _j^\lambda `$ is the $`\varphi `$ meson polarization vector, $`I_\pi `$ denotes the isospin factor being equal to $`\sqrt{2}`$ (1) for a $`\pi ^{}`$ ($`\pi ^0`$) meson in the entrance channel, and the nucleon spin indices are not displayed; $`i,j\mathrm{}`$ are Lorentz indices, and $`\gamma _i`$ and $`u`$ denote Dirac matrices and bispinors.
The invariant amplitude for the direct radiation channel (b) in Fig. 1 has the following form
$`T_{(1b)\lambda }=ig_{\varphi NN}g_{\pi NN}\overline{u}(p_c)\left[\mathrm{\Gamma }_\varphi ^i(q){\displaystyle \frac{\overline{)}p_l+M_N}{p_l^2M_N^2}}+{\displaystyle \frac{\overline{)}p_r+M_N}{p_r^2M_N^2}}\mathrm{\Gamma }_\varphi ^i(q)\right]\overline{u}(p_c)\epsilon _i^\lambda I_\pi F_N(p_l,p_r),`$ (17)
where $`\mathrm{\Gamma }_\varphi ^i(q)`$ and $`F_N`$ are defined by Eqs. (15) and (13), respectively, and $`p_l=p_aq`$ and $`p_r=p_cq`$.
### B $`𝑵𝑵\mathbf{}𝑵𝑵\mathit{\varphi }`$ reaction
The total invariant amplitude of meson exchange diagrams (a) in Fig. 2 with internal meson conversion is the sum of 4 amplitudes
$$(T_M)_\alpha =\xi _\alpha ^1T_M[ab;cd]+\xi _\alpha ^2T_M[ab;dc]+\xi _\alpha ^3T_M[ba;dc]+\xi _\alpha ^4T_M[ba;cd]$$
(18)
with $`\xi _{pp}^1=\xi _{pp}^3=\xi _{pp}^2=\xi _{pp}^4=1`$, $`\xi _{pn}^1=\xi _{pn}^3=1`$, $`\xi _{pn}^2=\xi _{pn}^4=2`$. The last two terms stem from the antisymmetrization or from charged meson exchange in $`pp`$ or $`pn`$ reactions, respectively<sup>§</sup><sup>§</sup>§In we used a convention with $`\xi _{pn}^2=\xi _{pn}^4=2`$, which, however, does not change our threshold prediction for the ratio of the total cross sections in $`pn`$ and $`pp`$ interaction made there without the final state interactions.. The first term in Eq. (18) for the $`pp`$ reaction reads
$`T_{(2a)\lambda }[ab;cd]=K^{NN}\left[\overline{u}(p_d)\gamma _5u(p_b)\right]\left[\overline{u}(p_c)\mathrm{\Gamma }_\rho ^j(k)u(p_a)ϵ_{ijkl}k_\rho ^iq_\varphi ^k\epsilon _\lambda ^l\right],`$ (19)
with
$`K^{NN}(k_\pi ^2,k_\rho ^2)={\displaystyle \frac{g_{\pi NN}g_{\rho NN}g_{\varphi \rho \pi }}{(k_\pi ^2m_\pi ^2)(k_\rho ^2m_\rho ^2)}}{\displaystyle \frac{\mathrm{\Lambda }_\pi ^2m_\pi ^2}{\mathrm{\Lambda }_\pi ^2k_\pi ^2}}{\displaystyle \frac{\mathrm{\Lambda }_\rho ^2m_\rho ^2}{\mathrm{\Lambda }_\rho ^2k_\rho ^2}}{\displaystyle \frac{\mathrm{\Lambda }_{\varphi \rho \pi }^{\rho \mathrm{\hspace{0.17em}2}}m_\rho ^2}{\mathrm{\Lambda }_{\varphi \rho \pi }^{\rho 2}k_\rho ^2}}{\displaystyle \frac{\mathrm{\Lambda }_{\varphi \rho \pi }^{\pi \mathrm{\hspace{0.17em}2}}m_\pi ^2}{\mathrm{\Lambda }_{\varphi \rho \pi }^{\pi 2}k_\pi ^2}}.`$ (20)
The amplitude of direct $`\varphi `$ meson emission from the nucleon legs according to Fig. 2b is calculated similarly to the real or virtual photon bremsstrahlung in the few GeV region. The internal zig-zag lines in Fig. 2b correspond to a suitably parameterized $`NN`$ interaction in terms of an effective two-body $`T`$-matrix which is written in the form of the one-boson exchange model (OBE) with effective coupling constants and cut-off parameters and may be interpreted as effective $`\pi ,\omega ,\rho ,\sigma `$ meson exchanges. We would like to stress that this is an effective dynamical model which is appropriate in the few GeV region and which is different from the OBE model in the conventional sense. This model has been applied successfully to different reactions and this encourages us to employ it for the $`\varphi `$ production too.
The total amplitude for the process (b) in Fig. 2 consists of $`28`$ $`(26)`$ contributions for $`pp`$ ($`pn`$) interactions and has a similar structure as Eq. (18) (with $`\xi _{pn}^1=\xi _{pn}^3=1,\xi _{pn}^2=\xi _{pn}^4=0`$, for $`\sigma ,\omega `$ exchanges), where $`T[ab;cd]`$ now reads
$`T_{(2b)\lambda }[ab;cd]=g_{\varphi NN}\epsilon _i^\lambda \left[\overline{u}(p_d)V^mu(p_b)\right]\times `$ (21)
$`{\displaystyle \underset{m=\pi ,\sigma ,\rho ,\omega }{}}[iD^m]\overline{u}(p_c)\left[V^m{\displaystyle \frac{\overline{)}p_l+M_N}{p_l^2M_N^2}}\mathrm{\Gamma }_\varphi ^i(q)+\mathrm{\Gamma }_\varphi ^i(q){\displaystyle \frac{\overline{)}p_r+M_N}{p_r^2M_N^2}}V^m\right]u(p_a).`$ (22)
Here, $`V^m`$ and $`D^m`$ are effective coupling vertices and propagators of the two-body $`T`$ matrix, respectively,
$$D^{\pi ,\sigma }=\frac{i}{k^2m_{\pi ,\sigma }^2},D_{\mu \nu }^{\rho ,\omega }=i\frac{g_{\mu \nu }k_\mu k_\nu m_{\rho ,\omega }^2}{k^2m_{\rho ,\omega }^2},$$
(23)
$$V^\pi =iG_{\pi NN}\gamma _5,V^\sigma =G_{\sigma NN},V_{\rho ,\omega }^i(k)=G_{\rho ,\omega NN}\mathrm{\Gamma }_{\rho ,\omega }^i,$$
(24)
where $`k`$ is the four momentum of the virtual meson $`m`$ and $`G_{mNN}`$ is the vertex function which includes the corresponding cut-off form factor. The numerical values of the $`G_{mNN}`$ are taken from Refs. .
In the near-threshold region the relative velocity of the outgoing nucleons is small which might result in a strong final state interaction (FSI) between them. If the energy excess is a few MeV up to a few tens MeV then one can consider only the $`s`$-wave interaction and account for the final state interaction in terms of the enhancement factors by renormalizing the basic amplitude correspondingly. For instance, for the $`pn`$ reaction we get
$`T_{pn}[ab;cd]T_{pn}[ab;cd]\left(_{0}^{}{}_{pn}{}^{}\delta _{m_cm_d}+_{1}^{}{}_{pn}{}^{}\delta _{m_cm_d}\right),`$ (25)
where $`m_c,m_d`$ are the spin projections of the nucleons in the final state, and $`_0,_1`$ are the singlet and triplet enhancement FSI factors, which are calculated within the Jost function and the phase-equivalent potentials formalism, which we describe in detail in Appendix A. The calculation shows that the singlet enhancement factor is much greater than the triplet one, i.e. $`|_0|^21|_1|^210`$ and greater than the corresponding factors of higher partial waves. Thus, for the $`pp`$ interaction we use
$`T_{pp}[ab;cd]T_{pp}[ab;cd]\left(_{0}^{}{}_{pp}{}^{}\delta _{m_cm_d}+\delta _{m_cm_d}\right),`$ (26)
reminding that at the threshold the $`pp`$ triplet final state is exactly zero. $`_{0}^{}{}_{pn}{}^{}`$ and $`_{0}^{}{}_{pp}{}^{}`$ are different which reflects the difference in the corresponding effective radii and the scattering lengths. Note that the mutual FSI of the $`\varphi `$ and the outgoing nucleons is assumed to be negligible.
In calculating the cross sections and the spin density matrix, squares and bilinear forms of the FSI-corrected amplitudes need to be evaluated numerically.
## IV Results
### A Fixing parameters
The parameters of the two-body $`T`$ matrix for the direct $`\varphi `$ emission depicted in Fig. 2b are taken from Refs. , where a quite reasonable agreement with data of different elastic and inelastic $`NN`$ reactions is found.
The coupling constant $`g_{\varphi \rho \pi }`$ is determined by the $`\varphi \rho \pi `$ decay. The recent value $`\mathrm{\Gamma }(\varphi \rho \pi )=`$ 0.69 MeV results in $`|g_{\varphi \rho \pi }|`$ = 1.10 GeV<sup>-1</sup>. The SU(3) symmetry consideration predicts a negative value for it. Thus, $`g_{\varphi \rho \pi }=1.10`$ GeV<sup>-1</sup>.
The remaining parameters of the meson exchange amplitudes for the processes in Figs. 1a and 1b are taken from the Bonn model as listed in Table B.1 (Model II) of Ref. .
The yet undetermined parameters are: the cut-off parameters for the virtual mesons in the $`\varphi \rho \pi `$ vertex, $`\mathrm{\Lambda }_{\varphi \rho }^\pi `$ and $`\mathrm{\Lambda }_{\varphi \rho }^\rho `$, the cut-off $`\mathrm{\Lambda }_N`$ in Eq. (13), and the parameters of the $`\varphi NN`$ interaction, $`g_{\varphi NN}`$ and $`\kappa _\varphi `$. We can reduce the number of parameters by making the natural assumption $`\mathrm{\Lambda }_{\varphi \rho }^\pi =\mathrm{\Lambda }_{\varphi \rho }^\rho `$ based on the symmetry of the virtual mesons in the $`\varphi \rho \pi `$ vertex . The next consideration is related to the tensor coupling $`\kappa _\varphi `$. Based on the $`\varphi \omega `$ similarity we do not expect a large value for it and in all our subsequent calculations we employ the theoretical estimate $`\kappa _\varphi =0.2`$ as an upper limit.
Even after that we have three free parameters being $`g_{\varphi NN}`$, $`\mathrm{\Lambda }_{\varphi \rho }^\rho `$ and $`\mathrm{\Lambda }_N`$. For $`g_{\varphi NN}`$ the SU(3) symmetry predicts
$`g_{\varphi NN}=\mathrm{tg}\mathrm{\Delta }\theta _Vg_{\omega NN},`$ (27)
where $`\mathrm{\Delta }\theta _V`$ is the deviation from the ideal $`\omega \varphi `$ mixing angle. It is responsible for the “standard” OZI rule violation, and in general, it depends on the method of its determination (Gell-Mann–Okubo linear or quadratic mass formulae, radiative decays, say $`\varphi (\omega )\gamma \pi `$, etc.). Using the quadratic Gell-Mann–Okubo mass formula one gets $`\mathrm{\Delta }\theta _V=3.7^0`$. Sometimes another relation is used, e.g.
$`g_{\varphi NN}=3\mathrm{sin}\mathrm{\Delta }\theta _Vg_{\rho NN},`$ (28)
which is obtained from the SU(3) relation
$`g_{\omega NN}={\displaystyle \frac{3FD}{F+D}}\mathrm{cos}\mathrm{\Delta }\theta _Vg_{\rho NN},`$ (29)
and Eq. (27) with the assumption $`D/F=0`$ in the SU(3) vector meson octet. Using the known values for $`g_{\rho NN}^2/4\pi =0.71.3`$ and $`g_{\omega NN}^2/4\pi =2224`$ , one may obtain $`g_{\varphi NN}=0.570.65`$ and $`g_{\varphi NN}=1.071.09`$ for the expressions (28) and (27), respectively. On the other hand, the theoretical estimates of Ref. give $`g_{\varphi NN}=0.24`$. Thus, we can conclude that even using the standard OZI rule violation (thought non-ideal $`\omega \varphi `$ mixing) one is left with estimated values of $`g_{\varphi NN}`$ within a quite large interval. The possible hidden strangeness in a nucleon may even increase this interval. In this paper we restrict ourselves to the standard OZI rule violation mechanisms and analyze consequences of varying $`g_{\varphi NN}`$ in the region 0.0 – 1.0.
The negative coupling constant $`g_{\varphi \rho \pi }`$ results in a destructive interference between meson exchange amplitudes (a) and direct emission (b) in Figs. 1 and 2. Analyzing the unpolarized $`\pi ^{}pn\varphi `$ reaction, based on the data of Ref. , we find that the yet unconstrained three parameters $`g_{\varphi NN}`$, $`\mathrm{\Lambda }_{\varphi \rho }^\rho `$, $`\mathrm{\Lambda }_N`$ become related to each other as $`\mathrm{\Lambda }_{\varphi \rho }^\rho `$=$`\mathrm{\Lambda }_{\varphi \rho }^\rho (g_{\varphi NN},\mathrm{\Lambda }_{\varphi \rho }^\rho )`$ by the constrains given by the data, and two solutions emerge for this dependence: (i) $`\sigma _{(a)}>\sigma _{(b)}`$ and (ii) $`\sigma _{(a)}<\sigma _{(b)}`$, where $`\sigma _{(a,b)}`$ are the total cross section for the meson exchange process (a) and the direct $`\varphi `$ emission (b) calculated separately. These solutions are displayed in Fig. 3 for several values of $`g_{\varphi NN}`$ as discussed above.
In order to constrain one more free parameter we analyze also the cross section $`d\sigma /d\mathrm{\Omega }`$ for the $`pppp\varphi `$ reaction, using simultaneously the DISTO data , i.e. the angular distribution (we remind that in our notation $`\mathrm{\Omega }`$ is the $`\varphi `$ meson solid angle) and the total cross section , respectively. For this aim we fix the absolute normalization of the angular distribution $`d\sigma /d\mathrm{\Omega }`$ given in Ref. by making use of the recently published the total cross section . As a result we get the fat dots in Fig. 3. The minimum values for $`g_{\varphi NN}`$ are 0.07 (for $`\sigma _{(a)}>\sigma _{(b)}`$) and 0.60 (for $`\sigma _{(a)}<\sigma _{(b)}`$), respectively at $`\mathrm{\Lambda }_N\mathrm{}`$. For both solutions we find that increasing values of $`|g_{\varphi NN}|`$ results in decreasing values of $`\mathrm{\Lambda }_N`$ leaving the total cross section or “integrated” strength of the $`\varphi NN`$ interaction on the same level.
To explore in more detail the relative importance of the direct $`\varphi NN`$ interaction we now employ three parameter sets; two of them correspond to the $`\sigma _{(a)}>\sigma _{(b)}`$ solution and the third one to $`\sigma _{(a)}<\sigma _{(b)}`$:
set A: $`g_{\varphi NN}=0.24`$ , $`\mathrm{\Lambda }_{\varphi \rho }^\rho =1.34`$ GeV, $`\mathrm{\Lambda }_N=1.065`$ GeV ($`\sigma _{(a)}>\sigma _{(b)}`$),
set B: $`g_{\varphi NN}=0.8`$, $`\mathrm{\Lambda }_{\varphi \rho }^\rho =1.34`$ GeV, $`\mathrm{\Lambda }_N=0.715`$ GeV ($`\sigma _{(a)}>\sigma _{(b)}`$),
set C: $`g_{\varphi NN}=0.8`$, $`\mathrm{\Lambda }_{\varphi \rho }^\rho =1.60`$ GeV, $`\mathrm{\Lambda }_N=1.99`$ GeV ($`\sigma _{(a)}<\sigma _{(b)}`$).
In the sets B and C we choose $`g_{\varphi NN}`$ close to its upper limit as predicted by SU(3) symmetry.
Results of our calculation of the total cross section for the $`\pi ^{}pn\varphi `$ reaction for these parameter sets are shown in Fig. 4 as a function of the energy excess $`\mathrm{\Delta }s^{\frac{1}{2}}s^{\frac{1}{2}}s_{0}^{}{}_{\pi N}{}^{\frac{1}{2}}`$ with $`s_{0}^{}{}_{\pi N}{}^{}=(M_N+m_\varphi )^2`$. The contribution of the meson exchange channel (a) is displayed separately by the dot-dashed line, and the direct emission by the nucleon line (b) is depicted by the dashed line. Clearly seen is the strong destructive interference of the channels (a) and (b), in particular for the set C.
In Figs. 5 and 6 we display results of our calculations for $`d\sigma /d\mathrm{\Omega }`$ for the $`pppp\varphi `$ reaction together with the available data for the parameter sets B and C, respectively, at $`\mathrm{\Delta }s^{\frac{1}{2}}=s^{\frac{1}{2}}s_{0}^{}{}_{NN}{}^{\frac{1}{2}}=82`$ MeV with $`s_{0}^{}{}_{NN}{}^{}=(2M_N+m_\varphi )^2`$. The results for the set A are very similar to that of the set B and we do not separately display them here. Interestingly, in all parameter sets considered the channel (a) dominates in the $`NNNN\varphi `$ reaction, but the interferences are different for different parameter sets.
One can see a qualitative difference in $`\pi N`$ and $`NN`$ reactions for the set C. In the $`\pi N`$ reaction the direct radiation channel (b) is dominant, i.e. $`\sigma _{(b)}>\sigma _{(a)}`$. In contrast, in the $`NN`$ reaction the relative contribution of the direct emission channel (b) increases considerably as compared with sets A and B, but it is still smaller that the meson exchange channel (a). The reason for this difference is the following one. The direct emission (b) in the reaction $`\pi N`$ consists of the two competing u-channel and s-channel diagrams shown in Fig. 1b. which add destructively, while the contribution of the of the u-channel amplitude is greater. However, the corresponding contributions of the two competing diagrams in Fig. 2b are numerically nearly the same resulting in a stronger suppression of the direct channel (b).
As we have adjusted our parameters by the data, it is clear that they describe the data with approximately equal quality, and at the present level of the data accuracy it is difficult to give a preference to one of them. Therefore, we now investigate whether other observables can be used to constrain the parameters further and whether the difference between $`pp`$ and $`pn`$ reactions is a sensible measure.
### B $`𝝅^{\mathbf{}}𝒑\mathbf{}𝒏\mathit{\varphi }`$ reaction
The calculated angular differential cross sections of the $`\pi ^{}pn\varphi `$ reaction at $`\mathrm{\Delta }s^{\frac{1}{2}}=50`$ MeV and for the parameter sets A, B, C are shown in Fig. 7. One can see that the shapes of the distributions for the sets A and B are very similar to each other. They are quite smooth and close to the distribution of the meson exchange channel (a). Only in the backward direction the total cross section slightly decreases due to the destructive interference with the direct channel (b), leading to some enhancement of the cross section in forward direction. (We would like to mention here, that the extrapolated value of the $`\varphi `$ production cross section in $`NN`$ reactions from the data would require somewhat different parameter sets, which in turn cause also more pronounced differences between the sets A and B.) Contrary to that, for the model C the largest destructive interference appears at forward direction, where the contributions of the two competing channels (a) and (b) are close to each other. As a result, the cross section is enhanced in the backward direction. So we can conclude that the differential cross section is sensitive to the dynamics of the $`\varphi `$ production and the direct $`\varphi NN`$ coupling (a similar conclusion for $`\omega `$ production has arrived at in Ref. ).
The prediction for spin density matrix elements $`\rho _{00}`$ and $`\rho _{11}`$ for the different models is shown in Fig. 8 as a function of the $`\varphi `$ production angle in c.m.s. at $`\mathrm{\Delta }s^{\frac{1}{2}}=50`$ MeV. The sets A and B deliver standard values, typical for the spin-flip processes, i.e. $`\rho _{00}0`$, $`\rho _{11}0.5`$. But the parameter set C predicts a strong deviation from these values, especially at forward direction. The reason of this effect is the following. In the meson exchange channel (a) the nucleon spin-flip amplitudes result in transitions $`m_im_\varphi ,m_f`$, where $`m_f=m_im_\varphi `$ with $`m_\varphi =\pm 1`$. For instance, the transitions like $`\frac{1}{2}1,\frac{1}{2}`$ are dominant. In the direct radiation channel (b) together with this strong amplitudes we have finite amplitudes for the transition $`m_i0,m_f`$, where $`m_f=m_i`$. In the set C the strongly competing nucleon spin flip amplitudes cancel each other and only the nucleon spin conserving direct emission amplitude (b) survives. This is illustrated in Fig. 9, where we show the nucleon spin flip (left panel) and the nucleon spin conserving (right panel) amplitudes for the set C. Here $`F_\mathrm{z}=\mathrm{Im}T^{\pi pn\varphi }`$ with the quantization axis along the $`𝐳`$ direction ($`T^{\pi pn\varphi }`$ is purely imaginary).
The anisotropies of the decay channels $`\varphi e^+e^{}`$ and $`\varphi K^+K^{}`$ (cf. Eqs. (5, 6) for the different parameter sets are shown in Fig. 10. Again, one can see a strong deviation of our prediction for the set C from the naive expectation $`B^{e^+e^{}}1`$, $`B^{K^+K^{}}1`$ based on a purely mesonic exchange channel or on the sets A and B. Fig. 11 illustrates the the manifestation of this deviation in the real $`e^+e^{}`$ and $`K^+K^{}`$ angular distributions. The distributions $`W_L`$ and $`W_T`$ are the longitudinal (along the quantization axis) and transversal fluxes for the outgoing electrons or kaons. The functions $`W_L`$ and $`W_T`$ are normalized as $`\sqrt{W_L^2(\mathrm{\Theta })+W_T^2(\mathrm{\Theta })}d\mathrm{cos}\mathrm{\Theta }=1`$, where $`\mathrm{\Theta }`$ is defined by Eq. (5). Thus, one can see that the sets A and B predict a practically vanishing kaon flux in the longitudinal direction for all $`\varphi `$ production angles. The set C predicts a finite amount of the longitudinal flux which increases with decreasing $`\varphi `$ production angle in c.m.s. A corresponding modification is predicted for the electron flux too.
### C $`𝒑𝒑\mathbf{}𝒑𝒑\mathit{\varphi }`$ and $`𝒑𝒏\mathbf{}𝒑𝒏\mathit{\varphi }`$ reactions
As shown in Figs. 5 and 6 the meson exchange contribution (a) is the dominating contribution to the $`NNNN\varphi `$ reaction, therefore, it is useful to recall the threshold prediction for this channel in the absence of the final state interaction, which serves as a starting point for further calculations at finite energy. Adopting the notation of Ref. we can express the invariant amplitudes of the reaction $`abab\varphi `$ with $`a=p`$, $`b=p`$ or $`n`$ as following
$`T_{pp}=F_1,T_{pn}={\displaystyle \frac{1}{2}}\left(F_0+F_1\right),`$ (30)
where $`F_0(F_1)`$ is the initial singlet (triplet) amplitude with
$`F_0=f_0(1)^{\frac{1}{2}+m_a}\delta _{m_am_b}(\delta _{\frac{1}{2}m_c}\delta _{\frac{1}{2}m_d}\delta _{\frac{1}{2}m_c}\delta _{\frac{1}{2}m_d}),`$ (31)
$`F_1=f_1(1)^{\frac{1}{2}+m_a}\delta _{m_am_b}(\delta _{\frac{1}{2}m_c}\delta _{\frac{1}{2}m_d}\delta _{\frac{1}{2}m_c}\delta _{\frac{1}{2}m_d}),`$ (32)
where $`m_{a,b}`$ and $`m_{c,d}`$ are again the nucleon spin projections in the initial and final states, respectively, $`f_0=6\sqrt{2}T_0`$, $`f_1=2\sqrt{2}T_0`$, where the threshold amplitude $`T_0`$ is defined by Eq. (25) in Ref. . The above equations lead to the ratio $`f_0/f_1=3`$ and to the ratio of singlet to triplet cross sections
$`{\displaystyle \frac{|F_0|^2}{|F_1|^2}}=9.`$ (33)
The beam target asymmetries (8) read $`C_{BT}^{}{}_{pp}{}^{}=1`$, and $`C_{BT}^{}{}_{pn}{}^{}=0.8`$. The ratio of the total cross sections in $`pn`$ and $`pp`$ reactions is 5. Accordingly, the prediction for the spin density reads $`\rho _{00}=0`$, $`\rho _{11}=0.5`$.
Let us now turn back to the Figs. 5 and 6. These figures show (i) a relatively small contribution of the direct radiation channel (b), which is in agreement with previous works , (ii) the cross sections for the $`pn`$ interaction are qualitatively very similar in shape to these of the $`pp`$ interaction but they are larger, and (iii) the ratio of the corresponding cross sections in $`pn`$ and $`pp`$ reactions is different for the sets A (or B) and C. Later we will discuss this aspect in detail.
The energy dependence of the total cross sections of $`pppp\varphi `$ and $`pnpn\varphi `$ reactions for the sets B and C is shown in Figs. 12 and 13. We do not display the result for the set A because it is practically the same as for the set B. The experimental data is taken from Ref. . One can see that the direct radiation channel (b) is much indeed smaller than the meson exchange contribution (a) in the near-threshold region where our consideration is valid.
Fig. 14 shows the energy dependence of the ratio of the total cross sections of $`pppp\varphi `$ and $`pnpn\varphi `$ reactions for the different parameter sets. One can see that this ratio increases with the energy excess and differs from the threshold value 5 in case of absence of FSI. The FSI is greater in the triplet initial (or singlet final) states and reduces the contribution of the initial singlet state in the $`pn`$ interaction. For the set C the ratio $`\sigma _{pn}/\sigma _{pp}`$ is greater because of the relatively greater contribution of the initial triplet state in the meson exchange channel (a). Fig. 15 shows the energy dependence of the ratio of the singlet to triplet cross sections in $`pn`$ interactions (cf. Eq. (33)). The left panel shows this ratio for the separate channels, while on the right panel one can see our prediction for this ratio for the different parameters sets. One can again see a strong deviation from the threshold prediction (33) without FSI and non-trivial non-monotonic dependence of these ratios with some maximum values around $`\mathrm{\Delta }s^{\frac{1}{2}}20`$ MeV.
Fig. 16 shows beam target asymmetry (8) for the separate channels for $`pp`$ and $`pn`$ interactions. For the $`pp`$ interaction it coincides with its threshold value $`C_{BT}=1`$ up to a relatively large energy excess. For the $`pn`$ interaction the asymmetry is different for the different channels which reflects the different role of the singlet and triplet states in the different amplitudes which are additionally modified by the FSI.
The total asymmetry for the different parameter sets is shown in Fig. 17. It is interesting that even for the sets A and B with small contribution of the direct radiation amplitude (b) the asymmetry for $`pn`$ interaction strongly deviates from the threshold prediction (without FSI: $`C_{BT}^{pn}=0.8`$), displaying a minimum around $`\mathrm{\Delta }s^{\frac{1}{2}}20`$ MeV.
We do not display here our results for the spin density matrix elements because for the sets A, B and C we get almost the threshold values, i.e. $`\rho _{00}=0`$, $`\rho _{11}=0.5`$, which reflects the dominance of the meson exchange channel (a).
## V Summary
We have analyzed the $`\varphi `$ production in $`\pi N`$ and $`NN`$ interactions in the near-threshold region using the conventional “non-strange” hadron dynamics, based on the amplitudes allowed by non-ideal $`\omega \varphi `$ mixing, that is meson conversion in a $`\varphi \pi \rho `$ vertex and direct $`\varphi `$ emission from the nucleon legs by a direct $`\varphi NN`$ coupling. Using the limited body of available experimental data of the total unpolarized reactions we have tried to reduce as much as possible the uncertainty of the model parameters. As a result we get two branches of solutions with either a relatively small or a relatively large contribution of the direct emission channel which is determined by the strength of the $`\varphi NN`$ interaction. By making use of these solutions we have compared various parameter sets with different strengths of the direct $`\varphi NN`$ interaction.
Analyzing the $`\pi pn\varphi `$ reaction we find a strong dependence of the various observables on the strength of the $`\varphi NN`$ interaction. The study of the differential cross section and angular distributions of electrons and kaons in the $`\varphi e^+e^{}`$ and $`\varphi K^+K^{}`$ decays seems to be most promising in investigating the $`\varphi NN`$ dynamics. Experimentally, this study might be performed with the pion beam at the HADES spectrometer in GSI/Darmstadt.
Analyzing the $`NNNN\varphi `$ reaction we find a large difference in $`pp`$ and $`pn`$ reactions due to the different role of the singlet and triplet nucleon spin states in the entrance channel and strong final state interaction. We predict a non-monotonic energy dependence of the ratio $`\sigma _{pn}/\sigma _{pp}`$ and of the beam target asymmetry for the $`pn`$ interaction which deviates strongly from the pure threshold prediction.
Finally, we emphasize once more that the present investigation is completely based on the conventional meson-nucleon dynamics and, therefore, our predictions may be considered as a necessary background for forthcoming studies of the strangeness degrees of freedom in non-strange hadrons. Additionally we would like to mention that fixing the $`\varphi NN`$ coupling is important for an access to the elastic $`\varphi N`$ scattering cross section which determines the degree of thermalization and collective flow properties of the $`\varphi `$ mesons in heavy-ion collisions at SIS energies.
### Acknowledgments
We gratefully acknowledge fruitful discussions with M. Debowski, L.P. Kaptari, N. Kaiser, R. Kotte, J. Ritman, and V.V. Shklyar. One of the authors (A.I.T.) thanks for the warm hospitality of the nuclear theory group in the Research Center Rossendorf. This work is supported by BMBF grant 06DR829/1, Heisenberg-Landau program, and HADES-JINR participation project #03-1-1020-95/2002.
## A Final state interaction
In this appendix we present the formulae for the FSI and corresponding correction factors. We use the general framework for the FSI enhancement factor based on the Jost function formalism. Important aspects of this framework are described in the monograph by Gillespie and some early original papers . With respect to the significance of this problem in studying various near-threshold particle production reactions in the present time with cooled beams, we accumulate here the relevant expressions of this method and give the final result in a form convenient for specific calculations. For the Jost function formalism we use the notation of the textbook by Newton . For simplicity, we limit our consideration to the $`s`$-wave interaction which is dominant in the near-threshold region. A generalization for higher angular momenta may be done straightforwardly.
The FSI enhancement factor for two identical particles with momentum $`k`$ in their c.m.s. reads
$`={\displaystyle \frac{1}{𝒥_+(k)}},`$ (A1)
where $`𝒥_+`$ belongs to a set of functions $`𝒥_\pm (k)`$ which are defined through the Wronskian of two linearly independent solutions of the Schrödinger equation,
$`𝒥_\pm (k)=f_\pm (k,r)\phi ^{}(k,r)f_\pm ^{}(k,r)\phi (k,r),`$ (A2)
where the prime means here the derivative with respect to $`r`$. The function $`𝒥`$= $`𝒥_+`$ is called the Jost function. The integral equations for the regular and irregular functions $`\phi (k,r)`$ and $`f_\pm (k,r)`$ have the standard form
$`\phi (k,r)={\displaystyle \frac{\mathrm{sin}kr}{k}}+{\displaystyle \frac{1}{k}}{\displaystyle _0^{\mathrm{}}}𝑑r^{}\mathrm{sin}k(rr^{})V(r^{})f_\pm (k,r^{}),`$ (A3)
$`f_\pm (k,r)=\mathrm{e}^{\pm kr}{\displaystyle \frac{1}{k}}{\displaystyle _0^{\mathrm{}}}𝑑r^{}\mathrm{sin}k(rr^{})V(r^{})f_\pm (k,r^{}).`$ (A4)
Eq. (A2) and the boundary conditions $`\phi (0)=0,\phi ^{}(0)=1`$ show that $`𝒥_\pm (k)=f_+(k,0)`$, thus allowing the integral representation
$`𝒥(k)=1+{\displaystyle \frac{1}{k}}{\displaystyle _0^{\mathrm{}}}𝑑r\mathrm{sin}krV(r)f_+(k,r).`$ (A5)
The physical wave function $`\psi ^+(k,r)`$ is related to the regular function $`\phi (k,r)`$ as
$`\psi ^+(k,r)={\displaystyle \frac{k\phi (k,r)}{𝒥(k)}},`$ (A6)
which means that the inverse of the square of the modulus of $`𝒥(k)`$ measures the probability of finding the particles near $`r=0`$, relative to a situation without interaction. From Eq. (A5) one can find the important asymptotic condition
$`\underset{|k|\mathrm{}}{lim}𝒥=1,`$ (A7)
which shows that at high energies the enhancement tends to unity, thus leaving the total amplitude unchanged.
The analyticity of $`𝒥`$ together with the asymptotic condition Eq. (A7) leads to the integral representation of the Jost function in terms of the phase shift $`\delta (k)`$
$`𝒥(k)={\displaystyle \underset{n}{}}\left(1+{\displaystyle \frac{\kappa _n^2}{k^2}}\right)\mathrm{exp}\left[{\displaystyle \frac{1}{\pi }}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑k^{}{\displaystyle \frac{\delta (k^{})}{kk^{}+iϵ}}\right],`$ (A8)
where $`\kappa _n`$ is related to the corresponding binding energy $`\kappa _n^2=2\mu ϵ_n>0`$ if bound states appear; $`\mu `$ is the corresponding reduced mass.
For the practical usage of the above formalism it is convenient to work with the effective potentials which give an exact analytical expression for the phase shift. Let us first consider the singlet $`NN`$ scattering (without bound state). The Eckart potential
$`V(r)={\displaystyle \frac{8\alpha ^2}{\alpha ^2\beta ^2}}\left({\displaystyle \frac{\mathrm{e}^{\alpha r}}{\alpha \beta }}+{\displaystyle \frac{\mathrm{e}^{\alpha r}}{\alpha +\beta }}\right)^2,\alpha >0,\beta >0,`$ (A9)
gives the $`s`$-wave phase shift
$`k\mathrm{cot}\delta _0={\displaystyle \frac{\alpha \beta }{\alpha \beta }}+{\displaystyle \frac{k^2}{\alpha \beta }}`$ (A10)
reproducing the effective-range phase shift exactly,
$`k\mathrm{cot}\delta _0={\displaystyle \frac{1}{a_0}}+{\displaystyle \frac{1}{2}}r_0k^2,`$ (A11)
with
$`\alpha ={\displaystyle \frac{1}{r_0}}\left(\sqrt{12r_0a_0^1}+1\right),\beta ={\displaystyle \frac{1}{r_0}}\left(\sqrt{12r_0a_0^1}1\right).`$ (A12)
The insertion of $`\delta _0`$ from (A10),
$`\delta _0={\displaystyle \frac{i}{2}}\mathrm{ln}\left[{\displaystyle \frac{(ki\alpha )(k+i\beta )}{(k+i\alpha )(ki\beta )}}\right],`$ (A13)
in Eq. (A8) gives the analytical expression for the Jost function
$`𝒥(k)={\displaystyle \frac{k+i\beta }{k+i\alpha }},`$ (A14)
and the resulting enhancement factor for singlet interaction reads therefore as
$`_0(k)=C_0(k^2){\displaystyle \frac{\mathrm{sin}\delta _0\mathrm{e}^{i\delta _0}}{k}},`$ (A15)
$`C_0(k^2)={\displaystyle \frac{k^2+\alpha ^2}{\alpha \beta }}={\displaystyle \frac{(kr_0)^2+2\left(1r_0a_0^1+\sqrt{12r_0a_0^1}\right)}{2r_0}},`$ (A16)
which coincides with the classical Watson enhancement factor
$`_W(k)=C_W{\displaystyle \frac{\mathrm{sin}\delta _0\mathrm{e}^{i\delta _0}}{k}}`$ (A17)
in the limit of $`kr_00`$. Note that expression (A17) is commonly used in calculations of the FSI in the near-threshold region (cf. ), where the constant $`C_W`$ is fixed by a comparison of calculation and experimental data. Eq. (A16) is superior to Eq. (A17) because it yields a definite value of $`C_W=C_0(k^2=0)`$ by making use of the independent phase shift data and satisfies at the same time the required asymptotic behavior according to Eq. (A7),
$`\underset{|k|\mathrm{}}{lim}_0(k)=1,`$ (A18)
contrary to Eq. (A17), where $`_W(k)0`$ at $`|k|\mathrm{}`$.
Fig. 18 illustrates the effect of FSI when using different enhancement factors for our parameter set B for the reaction $`pppp\varphi `$, where we keep here only the dominant mesonic exchange diagram in Fig. 2a and use the threshold value. In this case the energy dependence comes only from the enhancement factor and the phase space volume. The cross section calculated with enhancement factors from Eqs. (A16) and (A17) with $`C_W=C_0(0)`$ are shown by the dashed and the solid lines, respectively. For comparison we show also the cross section calculated without FSI (dot-dashed line). One can see, that the difference between the two factors (A16) and (A17) is indeed negligible at sufficiently small energy excess, say at $`\mathrm{\Delta }s^{\frac{1}{2}}<10`$ MeV with $`kr_01`$, where the Watson theory is valid, thus supporting the approach of . At $`\mathrm{\Delta }s^{\frac{1}{2}}>50`$ MeV the difference between the two variants results in a factor 2 and greater.
Following Ref. , for the triplet $`pn`$ interaction one can use the effective potential of the Bargmann type
$`V(r)=4\kappa {\displaystyle \frac{d}{dr}}\left[\mathrm{sh}\alpha _1r{\displaystyle \frac{g(\kappa ,r)}{g(\kappa +\alpha _1,r)g(\kappa \alpha _1,r)}}\right],`$ (A19)
$`g(q,r)=\left(\mathrm{e}^{qr}+2\mathrm{s}\mathrm{h}qr\right)k^1`$ (A20)
which reproduces the known phase shift and the deuteron binding energy. The Jost function in this case reads
$`𝒥(k)={\displaystyle \frac{ki\kappa }{k+i\alpha _1}},`$ (A21)
with
$`\alpha _1=(2\kappa r_1)r_1^1,`$ (A22)
where $`\kappa ^2=2\mu ϵ_d`$, $`ϵ_d`$ is the deuteron binding energy, and $`r_1`$ and $`a_1=(\kappa (1\kappa r/2))^1`$ are the triplet effective radius and scattering length, respectively. The triplet enhancement factor reads explicitly
$`_1(k)=C_1(k^2){\displaystyle \frac{\mathrm{sin}\delta _1\mathrm{e}^{i\delta _1}}{k}},`$ (A23)
$`C_1(k^2)={\displaystyle \frac{k^2+\alpha _1^2}{\alpha _1+\kappa }}={\displaystyle \frac{(kr_1)^2+2\left(1r_1a_1^1+\sqrt{12r_1a_1^1}\right)}{2r_1}},`$ (A24)
In our calculation we use $`a_{0,1}`$ and $`r_{0,1}`$ from Ref.
$`pn\mathrm{singlet}:a_{0pn}=23.768\mathrm{fm},r_{0pn}=2.75\mathrm{fm},`$ (A25)
$`pp\mathrm{singlet}:a_{0pp}=7.8098\mathrm{fm},r_{0pp}=2.767\mathrm{fm},`$ (A26)
$`pn\mathrm{triplet}:a_{1pn}=\mathrm{\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}5.424}\mathrm{fm},r_{1pn}=1.759\mathrm{fm},\kappa ^1=4.318\mathrm{fm}.`$ (A27)
Finally, we would like to mention that the approach presented here is equivalent to the approach in Ref. if the off-shell correction factor $`𝒫`$ (cf. Eq. (9) in ) takes the form $`𝒫(E,k)=a_xr_x^1\left(1+\sqrt{12r_xa_x^1}\right)`$ with $`x=0,1`$. |
warning/0001/math0001071.html | ar5iv | text | # Free Field Construction for the ABF Models in Regime II
## 1. Introduction
In conformal field theory (CFT), free field construction (or the ‘Coulomb gas’ representation) is the most effective calculational tool for physical quantities. The same can be said about the vertex operator approach to solvable lattice models (see e.g. ). The latter can be viewed as a $`q`$-deformation of some conformal field models, the most typical example being the Andrews-Baxter-Forrester (ABF) models in regime III which corresponds to the minimal unitary series. Despite its technical importance, however, no uniform recipe is known at present for finding a free field representation of a given CFT. Even when it is known, an equally non-trivial task is to do the same for the corresponding off-critical solvable lattice models. In this article we address this issue in the case of the ABF models in regime II .
By the level-rank duality for the Boltzmann weights , the ABF models in regime II at level $`k`$ are equivalent to the $`A_{k1}^{(1)}`$ face models in regime III at level $`2`$. For the latter, a free field realization is already available . However this picture is rather complicated, since one has to deal with $`k1`$ kinds of oscillators. The Felder complex in higher rank is also quite cumbersome . On the other hand, the ABF models in regime II are described in the conformal limit by the $`_k`$ parafermionic CFT , and in the massive scaling limit by its perturbation by the first energy operator . Therefore one naturally expects that an alternative construction on the lattice is to invoke the $`q`$-deformation of the parafermion theory , where it is sufficient to deal with only two kinds of oscillators and the resolution of Fock modules has a simpler structure. In this paper we work out this point in detail. Our purpose is to show that the known results in representation theory fits nicely the interpretation as a bosonization of lattice models.
The text is organized as follows. In section 2 we recall the ABF models and set up the notation. In section 3 we review the free field realization of the deformed parafermion theory. We then present various commutation relations of operators. Comparing them with those in the lattice theory, we identify the deformed counterpart of the primary fields and the parafermionic currents with the vertex operators (VO’s) of type I and type II, respectively. A simplifying feature is that these deformed parafermionic currents mutually commute by scalars. To our knowledge, this is the only known example in the vertex operator approach<sup>1)</sup><sup>1)</sup>1)as well as the lattice counterpart of the non-unitary minimal models $`M_{2,2n+1}`$ which are expected to have the same property. where the type II VO’s do not involve integrals of screening operators. We note also the existence of a deformation of the $`W_k`$ currents, which we believe to be equivalent to the one in with specialization $`r=k+2`$. This conforms with the known equivalence between the $`_k`$ parafermionic CFT and the first member in the $`W_k`$ minimal series. In section 4 we calculate form factors of height variables on one and neighboring two lattice sites. In section 5 we take their continuous limit. Section 6 is devoted to the summary and discussions about open problems. Some technical formulas are gathered in the Appendices.
## 2. ABF model in regime II
### 2.1. Boltzmann weights
In this section we recall the ABF model in regime II and set up the notation. Throughout this article, we fix a positive integer $`k2`$.
The Boltzmann weights of the ABF model have the form
(2.1) $`W({\displaystyle \genfrac{}{}{0pt}{}{a}{c}}{\displaystyle \genfrac{}{}{0pt}{}{b}{d}}\left|u\right)=\rho (u)\overline{W}({\displaystyle \genfrac{}{}{0pt}{}{a}{c}}{\displaystyle \genfrac{}{}{0pt}{}{b}{d}}\left|u\right).`$
Here $`a,b,c,d`$ denote positive integers which we refer to as height variables. Besides, the weights depend on two real parameters $`u`$ and $`x`$. If $`1a,b,c,dk+1`$ and $`|ab|=|bd|=|dc|=|ca|=1`$, then we set
(2.2) $`\overline{W}({\displaystyle \genfrac{}{}{0pt}{}{a}{a\pm 1}}{\displaystyle \genfrac{}{}{0pt}{}{a\pm 1}{a\pm 2}}\left|u\right)=1,`$
(2.3) $`\overline{W}({\displaystyle \genfrac{}{}{0pt}{}{a}{a\pm 1}}{\displaystyle \genfrac{}{}{0pt}{}{a\pm 1}{a}}\left|u\right)={\displaystyle \frac{[a\pm u][1]}{[a][1u]}},`$
(2.4) $`\overline{W}({\displaystyle \genfrac{}{}{0pt}{}{a}{a1}}{\displaystyle \genfrac{}{}{0pt}{}{a\pm 1}{a}}\left|u\right)={\displaystyle \frac{[a1][u]}{[a][1u]}}.`$
Here $`[u]`$ stands for the function
$`[u]=x^{\frac{u^2}{k+2}u}\mathrm{\Theta }_{x^{2(k+2)}}(x^{2u}),`$
and we use the standard symbols
$`\mathrm{\Theta }_p(z)=(z;p)_{\mathrm{}}(pz^1;p)_{\mathrm{}}(p;p)_{\mathrm{}},`$
$`(z_1,\mathrm{},z_n;p_1,\mathrm{},p_m)_{\mathrm{}}={\displaystyle \underset{j=1}{\overset{n}{}}}{\displaystyle \underset{l_1,\mathrm{},l_m0}{}}(1p_1^{l_1}\mathrm{}p_m^{l_m}z_j).`$
For all other values of $`a,b,c,d`$, we set the weight (2.1) to $`0`$.
In this note we restrict to regime II defined by
$`0<x<1,{\displaystyle \frac{k}{2}}<u<0.`$
We choose the scalar factor $`\rho (u)`$ in (2.1) to ensure that the partition function per site equals to $`1`$. Explicitly
$`\rho (u)=x^{\frac{2u}{k(k+2)}}{\displaystyle \frac{\rho _+(u)}{\rho _+(u)}},`$
$`\rho _+(u)={\displaystyle \frac{(x^{2k+2+2u},x^{2k+2+2u};x^{2k},x^{2(k+2)})_{\mathrm{}}}{(x^{2k+2u},x^{2k+4+2u};x^{2k},x^{2(k+2)})_{\mathrm{}}}}.`$
We have
$`\rho (u)\rho (u)=1,\rho (u)\rho (ku)={\displaystyle \frac{[1u]^2}{[u][2u]}}.`$
In , a free field construction of the ABF model was found in regime III defined by $`0<x<1,0<u<1`$. The Boltzmann weights in regime II differs from that case only in the choice of $`\rho (u)`$ made above.
We represent the Boltzmann weights graphically as follows.
### 2.2. Vertex operators in the naïve picture
In this subsection we summarize the vertex operator approach in the naïve picture .
In regime II, there are $`2k`$ different ground states labeled by $`m/2k`$. Each of them is invarinat under the shift in the NE-SW direction, and hence is specified by a sequence of heights on a column. Fix a reference column and a site (say, site $`1`$) on it. Then the $`m`$-th ground state is given by the following sequence $`\{\overline{l}_j^{(m)}\}_j`$ on that column
$`\overline{l}_{jm}^{(m)}=\{\begin{array}{cc}j\hfill & (1jk)\hfill \\ 2k+2j\hfill & (k+1j2k)\hfill \end{array},`$
$`\overline{l}_{j+2k}^{(m)}=\overline{l}_j^{(m)}(j).`$
Here the sites on the reference column are numbered by $`j`$ from south to north in the increasing order.
Let $`_{m,a}`$ denote the space of states of the half-infinite lattice, in the sector where the central height (i.e. the one on the reference site $`1`$) is fixed to $`a`$ and the boundary heights are in the ground state $`m`$. Namely we set formally
$`_{m,a}=\text{ span }\{(l_j)_{j=1}^{\mathrm{}}1l_jk+1,l_1=a,l_j=\overline{l}_j^{(m)}(j1)\}.`$
Notice that $`_{m,a}=0`$ if $`ammod2`$. The corner transfer matrices (CTM’s) associated with the north-west, south-west, south-east and north-east quadrants operate respectively as
$`A_{NW}^{(a)}(u)`$ $`:`$ $`_{m,a}_{m,a},`$
$`A_{SW}^{(a)}(u)`$ $`:`$ $`_{m,a}_{m,a},`$
$`A_{SE}^{(a)}(u)`$ $`:`$ $`_{m,a}_{m,a},`$
$`A_{NE}^{(a)}(u)`$ $`:`$ $`_{m,a}_{m,a}.`$
In the limit of an infinitely large lattice, they have the simple form
$`A_{NW}^{(a)}(u)=x^{2uH_C^{(a)}},`$
$`A_{SW}^{(a)}(u)=M^{(a)}x^{2uH_C^{(a)}},`$
$`A_{SE}^{(a)}(u)=M^{(a)}x^{2uH_C^{(a)}}M_{}^{(a)}{}_{}{}^{1},`$
$`A_{NE}^{(a)}(u)=[a]x^{2(u+k)H_C^{(a)}}M_{}^{(a)}{}_{}{}^{1},`$
where the ‘corner Hamiltonian’ $`H_C^{(a)}:_{m,a}_{m,a}`$ has a discrete and equidistant spectrum, and $`M^{(a)}:_{m,a}_{m,a}`$ is an operator which does not play a role in the following. Both of them are independent of $`u`$.
Likewise let
$`\mathrm{\Phi }_N^{(a,b)}(u)`$ $`:`$ $`_{m,b}_{m+1,a},`$
$`\mathrm{\Phi }_W^{(b,a)}(u)`$ $`:`$ $`_{m+1,a}_{m,b},`$
$`\mathrm{\Phi }_S^{(a,b)}(u)`$ $`:`$ $`_{m,b}_{m+1,a},`$
$`\mathrm{\Phi }_E^{(b,a)}(u)`$ $`:`$ $`_{m+1,a}_{m,b}`$
be the half-infinite transfer matrices extending to north, west, south and east directions, respectively. (Note that the $`(m+1)`$-th ground state is obtained from the $`m`$-th one by shifting it one step to the left.) Normally we suppress the dependence on the boundary condition $`m`$ in the notation. Unlike the case of regime III, our Boltzmann weights do not have the crossing symmetry. Accordingly there is no simple relation between the operators $`\mathrm{\Phi }_N^{(a,b)}(u)`$ and $`\mathrm{\Phi }_W^{(a,b)}(u)`$.
Figure : Corner transfer matrices and VO’s of type I.
Formal manipulations show that these operators should satisfy various commutation relations. We have
$`\mathrm{\Phi }_N^{(a,b)}(u)=x^{2uH_C^{(a)}}\mathrm{\Phi }_N^{(a,b)}(0)x^{2uH_C^{(b)}},`$
$`\mathrm{\Phi }_W^{(a,b)}(u)=x^{2uH_C^{(a)}}\mathrm{\Phi }_W^{(a,b)}(0)x^{2uH_C^{(b)}}.`$
The VO’s satisfy the commutation relations
(2.5) $`\mathrm{\Phi }_N^{(a,b)}(u_2)\mathrm{\Phi }_N^{(b,c)}(u_1)={\displaystyle \underset{g}{}}W({\displaystyle \genfrac{}{}{0pt}{}{a}{b}}{\displaystyle \genfrac{}{}{0pt}{}{g}{c}}|u_1u_2)\mathrm{\Phi }_N^{(a,g)}(u_1)\mathrm{\Phi }_N^{(g,c)}(u_2),`$
(2.6) $`\mathrm{\Phi }_W^{(a,b)}(u_2)\mathrm{\Phi }_W^{(b,c)}(u_1)={\displaystyle \underset{g}{}}W({\displaystyle \genfrac{}{}{0pt}{}{c}{g}}{\displaystyle \genfrac{}{}{0pt}{}{b}{a}}|u_1u_2)\mathrm{\Phi }_W^{(a,g)}(u_1)\mathrm{\Phi }_W^{(g,c)}(u_2),`$
(2.7) $`\mathrm{\Phi }_N^{(a,b)}(u_1)\mathrm{\Phi }_W^{(b,c)}(u_2)={\displaystyle \underset{g}{}}W({\displaystyle \genfrac{}{}{0pt}{}{g}{a}}{\displaystyle \genfrac{}{}{0pt}{}{c}{b}}|u_1u_2)\mathrm{\Phi }_W^{(a,g)}(u_2)\mathrm{\Phi }_N^{(g,c)}(u_1).`$
We have in addition
(2.8) $`{\displaystyle \underset{g}{}}\mathrm{\Phi }_W^{(a,g)}(u)\mathrm{\Phi }_N^{(g,a)}(u)=\mathrm{id},`$
(2.9) $`{\displaystyle \underset{g}{}}{\displaystyle \frac{[g]}{[a]}}\mathrm{\Phi }_N^{(a,g)}(u)\mathrm{\Phi }_W^{(g,a)}(k+u)=\mathrm{id}.`$
The last two equations are consequences of the unitarity of the Boltzmann weights
$$\underset{g}{}W(\genfrac{}{}{0pt}{}{a}{g}\genfrac{}{}{0pt}{}{b}{c}\left|u\right)W(\genfrac{}{}{0pt}{}{a}{d}\genfrac{}{}{0pt}{}{g}{c}|u)=\delta _{bd},$$
and the second inversion relation
$$\underset{g}{}\frac{[g]}{[c]}W(\genfrac{}{}{0pt}{}{d}{b}\genfrac{}{}{0pt}{}{c}{g}|ku)W(\genfrac{}{}{0pt}{}{a}{c}\genfrac{}{}{0pt}{}{b}{g}\left|u\right)=\delta _{ad}\frac{[b]}{[d]}.$$
In Section 3 we shall present a bosonic realization of the relations (2.5)–(2.9).
### 2.3. Local height probabilities
Consider successive $`n`$ sites $`i_1,\mathrm{},i_n`$ on a row of the lattice, numbered from right to left. We regard $`i_1`$ as the reference site used to label the ground states. Fixing the boundary heights to the ground state $`m`$, we let $`P_{a_n,\mathrm{},a_1}(m)`$ denote the joint probability that the height variable $`l_{i_j}`$ takes the value $`a_j`$, $`j=1,\mathrm{},n`$. These are the $`n`$ point local height probabilities (LHP’s). In terms of the CTM and VO’s, they can be expressed as
$`P_{a_n,\mathrm{},a_1}(m)={\displaystyle \frac{1}{Z_m}}[a_1]`$
$`\times \mathrm{tr}_{_{m,a_1}}\left(x^{2kH_C^{(a_1)}}\mathrm{\Phi }_W^{(a_1,a_2)}(0)\mathrm{}\mathrm{\Phi }_W^{(a_{n1},a_n)}(0)\mathrm{\Phi }_N^{(a_n,a_{n1})}(0)\mathrm{}\mathrm{\Phi }_N^{(a_2,a_1)}(0)\right),`$
$`Z_m={\displaystyle \underset{a=1}{\overset{k+1}{}}}[a]\mathrm{tr}_{_{m,a}}\left(x^{2kH_C^{(a)}}\right).`$
It turns out that (with an appropriate normalization of $`H_C^{(a)}`$) $`Z=Z_m`$ is independent of $`m`$. Obviously $`P_{a_n,\mathrm{},a_1}(m)=0`$ for $`a_1mmod2`$.
The following result for $`n=1`$ is due to Andrews, Baxter and Forrester .
(2.10) $`P_a(m)={\displaystyle \frac{[a]\mathrm{tr}_{_{m,a}}\left(x^{2kH_C^{(a)}}\right)}{Z}}=x^{\frac{k+2}{4}}[a]c_{\mathrm{\Lambda }_k(m)}^{\mathrm{\Lambda }_k(a1)}(\tau ),`$
where $`c_{\mathrm{\Lambda }_k(m)}^{\mathrm{\Lambda }_k(l)}(\tau )`$ stands for the string function for integrable $`\widehat{\text{sl}}_2`$-modules. Explicitly it is
$`c_{\mathrm{\Lambda }_k(m)}^{\mathrm{\Lambda }_k(l)}(\tau )=\eta (\tau )^3`$
$`\times \left({\displaystyle \underset{n_1|n_2|}{}}{\displaystyle \underset{n_1>|n_2|}{}}\right)(1)^{2n_1}\left(x^{2k}\right)^{\frac{(l+1+2n_1(k+2))^2}{4(k+2)}\frac{(m+2n_2k)^2}{4k}}`$
for $`lmmod2`$ and $`c_{\mathrm{\Lambda }_k(m)}^{\mathrm{\Lambda }_k(l)}(\tau )=0`$ for $`lmmod2`$. In the above, we set $`x^{2k}=e^{2\pi i\tau }`$, $`\eta (\tau )=(x^{2k})^{1/24}(x^{2k};x^{2k})_{\mathrm{}}`$, $`\mathrm{\Lambda }_k(j)=(kj)\mathrm{\Lambda }_0+j\mathrm{\Lambda }_1`$, and the sum is taken over $`n_1,n_2\frac{1}{2}`$ with $`n_1n_2`$. We have
$`P_a(m)=0(ammod2),`$
$`{\displaystyle \underset{a=1}{\overset{k+1}{}}}P_a(m)=1,`$
$`P_a(m)=P_a(m).`$
As observed in for regime III, the nearest neighbor LHP’s ($`n=2`$) can be written in terms of the one point LHP. To see this it suffices to note the following relations which are obvious consequences of the definition.
$`P_{1,2}(m)=P_1(m+1),P_{2,1}(m)=P_1(m),`$
$`P_{a1,a}(m)+P_{a+1,a}(m)=P_a(m),`$
$`P_{a,a1}(m)+P_{a,a+1}(m)=P_a(m+1).`$
These properties fix $`P_{a\pm 1,a}(m)`$ uniquely in terms of the one point LHP (2.10) as follows.
(2.11) $`P_{a+1,a}(m)={\displaystyle \underset{\genfrac{}{}{0pt}{}{1sa}{samod2}}{}}P_s(m){\displaystyle \underset{\genfrac{}{}{0pt}{}{1sa}{samod2}}{}}P_s(m+1),`$
(2.12) $`P_{a1,a}(m)={\displaystyle \underset{\genfrac{}{}{0pt}{}{1s<a}{samod2}}{}}P_s(m)+{\displaystyle \underset{\genfrac{}{}{0pt}{}{1s<a}{samod2}}{}}P_s(m+1).`$
The right hand sides of (2.11)–(2.12) automatically satisfy $`P_{k+2,k+1}(m)=P_{k+1,k+2}(m)=0`$, as it should be.
## 3. Free Field Realization
The $`q`$-deformation of the Wakimoto module over the affine Lie algebra $`\widehat{\text{sl}}_2`$ was found by several authors (See refs. in as to the other variants of free field realizations.), using three kinds of bosonic fields. The deformed parafermion theory is obtained by dropping one of these fields. In this section we give some details of the latter theory, and interpret it as a free field realization of the ABF model in regime II. A part of the results are given in . The various relations given in Propositions 3.1, 3.33.6 are new.
### 3.1. Basic operators
The main objects in the deformed parafermion theory are the vertex operators (VO’s) of type I
(3.1) $`\mathrm{\Phi }_\pm ^{}(u),\mathrm{\Phi }_\pm (u),`$
and of type II
(3.2) $`\mathrm{\Psi }_+(u)=\mathrm{\Psi }^{}(u),\mathrm{\Psi }_{}(u)=\mathrm{\Psi }(u).`$
The VO’s of type I (3.1) are a lattice counterpart of the simplest primary fields in conformal theory, while those of type II (3.2) corresponds to the parafermions. In addition, there are also some auxiliary operators: the so-called screening current and the $`\xi `$-$`\eta `$ system which we denote by
$`S(u),\xi (u),\eta (u),`$
respectively. All these operators act on the direct sum of Fock modules $`=_{mlmod2}_{m,l}`$ labeled by $`m,l`$. Their explicit formulas are given in Appendix A. Here we only mention their basic features.
The operators
$`\mathrm{\Psi }_\pm (u),\mathrm{\Phi }_{}^{}(u),\mathrm{\Phi }_{}(u),S(u),\xi (u),\eta (u)`$
are either an exponential of a bosonic field or a sum of two such terms. In contrast, the formulas for $`\mathrm{\Phi }_+^{}(u),\mathrm{\Phi }_+(u)`$ comprise integrals. To elaborate on the last point, let us introduce the screening charge
$`X(u)={\displaystyle _{C_X(z)}}\underset{¯}{dz}^{}S(u^{}){\displaystyle \frac{[uu^{}\frac{1}{2}P_1]}{[uu^{}\frac{1}{2}]}}.`$
Here we set $`z=x^{2u}`$, $`z^{}=x^{2u^{}}`$, $`\underset{¯}{dz}^{}=dz^{}/(2\pi iz^{})`$, and $`P_1`$ denotes a ‘zero-mode’ operator. For more details we refer to Appendix A. The contour $`C_X(z)`$ is a simple closed curve encircling the origin counterclockwise in the $`z^{}`$-plane, such that $`z^{}=zx^{1+2(k+2)n}`$ ($`n0`$) are inside $`C_X(z)`$ and $`z^{}=zx^{1+2(k+2)n}`$ ($`n<0`$) are outside $`C_X(z)`$. Using the screening charge, the ‘plus’ components of the type I VO’s are given by
$`\mathrm{\Phi }_+(u)=\mathrm{\Phi }_{}(u)X(u+k+1),`$
$`\mathrm{\Phi }_+^{}(u)=\mathrm{\Phi }_{}^{}(u)X(u),`$
where the product $`A(u)B(v)`$ is defined to be the analytic continuation from the domain $`|x^{2u}||x^{2v}|`$.
Each of the operators $`\mathrm{\Phi }_\pm (u)`$, $`\mathrm{\Phi }_\pm ^{}(u)`$, $`\mathrm{\Psi }_\pm (u)`$, $`S(u)`$, $`\xi (u)`$, $`\eta (u)`$ sends one Fock space $`_{m,l}`$ to another $`_{m^{},l^{}}`$ with a different label $`(m^{},l^{})`$. The change of label is listed in the following table.
| | $`\mathrm{\Phi }_\pm (u)`$ | $`\mathrm{\Phi }_\pm ^{}(u)`$ | $`\mathrm{\Psi }_\pm (u)`$ | $`S(u)`$ | $`\xi (u)`$ | $`\eta (u)`$ |
| --- | --- | --- | --- | --- | --- | --- |
| $`_{m^{},l^{}}`$ | $`_{m+1,l1}`$ | $`_{m1,l1}`$ | $`_{m2,l}`$ | $`_{m,l2}`$ | $`_{mk,l(k+2)}`$ | $`_{m+k,l+k+2}`$ |
### 3.2. Commutation relations
In the next three propositions we state the commutation relations among the operators $`\mathrm{\Phi }_\pm (u)`$, $`\mathrm{\Phi }_\pm ^{}(u)`$ and $`\mathrm{\Psi }_\pm (u)`$. Set
$`P=P_1+1.`$
###### Proposition 3.1.
The operators $`\mathrm{\Phi }_\pm (u)`$, $`\mathrm{\Phi }_\pm ^{}(u)`$ satisfy the commutation relations
$`\mathrm{\Phi }_{\epsilon _2}(u_2)\mathrm{\Phi }_{\epsilon _1}(u_1)={\displaystyle }W({\displaystyle \genfrac{}{}{0pt}{}{P}{P+\epsilon _2}}{\displaystyle \genfrac{}{}{0pt}{}{P+\epsilon _1^{}}{P+\epsilon _1+\epsilon _2}}|u_1u_2)\mathrm{\Phi }_{\epsilon _1^{}}(u_1)\mathrm{\Phi }_{\epsilon _2^{}}(u_2),`$
$`\mathrm{\Phi }_{\epsilon _2}^{}(u_2)\mathrm{\Phi }_{\epsilon _1}^{}(u_1)={\displaystyle }W({\displaystyle \genfrac{}{}{0pt}{}{P+\epsilon _1+\epsilon _2}{P+\epsilon _1^{}}}{\displaystyle \genfrac{}{}{0pt}{}{P+\epsilon _2}{P}}|u_1u_2)\mathrm{\Phi }_{\epsilon _1^{}}^{}(u_1)\mathrm{\Phi }_{\epsilon _2^{}}^{}(u_2),`$
$`\mathrm{\Phi }_{\epsilon _1}(u_1)\mathrm{\Phi }_{\epsilon _2}^{}(u_2)={\displaystyle }W({\displaystyle \genfrac{}{}{0pt}{}{P+\epsilon _2^{}}{P}}{\displaystyle \genfrac{}{}{0pt}{}{P+\epsilon _1+\epsilon _2}{P+\epsilon _1}}|u_1u_2)\mathrm{\Phi }_{\epsilon _2^{}}^{}(u_2)\mathrm{\Phi }_{\epsilon _1^{}}(u_1),`$
where the Boltzmann weights are given in (2.2)–(2.4). The sums are taken over $`\epsilon _1^{},\epsilon _2^{}=\pm 1`$ such that $`\epsilon _1^{}+\epsilon _2^{}=\epsilon _1+\epsilon _2`$. Moreover we have the inversion relations
(3.3) $`{\displaystyle \underset{\epsilon }{}}\mathrm{\Phi }_\epsilon ^{}(u)\mathrm{\Phi }_\epsilon (u)=\mathrm{id},`$
(3.4) $`{\displaystyle \underset{\epsilon }{}}\mathrm{\Phi }_\epsilon (u)[P]\mathrm{\Phi }_\epsilon ^{}(k+u)=[P]\mathrm{id}.`$
In the next propositions we set
$`[u]^{}=x^{\frac{u^2}{k}u}\mathrm{\Theta }_{x^{2k}}(x^{2u}).`$
###### Proposition 3.2.
The following commutation relations hold.
$`\mathrm{\Psi }_\pm (u_2)\mathrm{\Psi }_\pm (u_1)={\displaystyle \frac{[u_1u_2+1]^{}}{[u_1u_21]^{}}}\mathrm{\Psi }_\pm (u_1)\mathrm{\Psi }_\pm (u_2),`$
$`\mathrm{\Psi }_\pm (u_2)\mathrm{\Psi }_{}(u_1)={\displaystyle \frac{[u_1u_21+\frac{k}{2}]^{}}{[u_1u_2+1+\frac{k}{2}]^{}}}\mathrm{\Psi }_{}(u_1)\mathrm{\Psi }_\pm (u_2).`$
Moreover, as $`u_2u_1\frac{k}{2}`$ we have
$`\mathrm{\Psi }_\pm (u_2)\mathrm{\Psi }_{}(u_1)={\displaystyle \frac{k}{2\pi }}{\displaystyle \frac{g^{}}{u_2u_1+\frac{k}{2}}}\times \mathrm{id}+O(1),`$
where $`g^{}={\displaystyle \frac{\pi }{k\mathrm{log}x}}{\displaystyle \frac{(x^{2k2};x^{2k})_{\mathrm{}}}{(x^2;x^{2k})_{\mathrm{}}}}`$.
###### Proposition 3.3.
$`\mathrm{\Psi }_+(u_2)\mathrm{\Phi }_\pm (u_1)={\displaystyle \frac{[u_1u_2\frac{1}{2}+\frac{k}{2}]^{}}{[u_1u_2+\frac{1}{2}+\frac{k}{2}]^{}}}\mathrm{\Phi }_\pm (u_1)\mathrm{\Psi }_+(u_2),`$
$`\mathrm{\Psi }_+(u_2)\mathrm{\Phi }_\pm ^{}(u_1)={\displaystyle \frac{[u_1u_2+\frac{1}{2}+\frac{k}{2}]^{}}{[u_1u_2\frac{1}{2}+\frac{k}{2}]^{}}}\mathrm{\Phi }_\pm ^{}(u_1)\mathrm{\Psi }_+(u_2),`$
$`\mathrm{\Psi }_{}(u_2)\mathrm{\Phi }_\pm (u_1)={\displaystyle \frac{[u_1u_2+\frac{1}{2}]^{}}{[u_1u_2\frac{1}{2}]^{}}}\mathrm{\Phi }_\pm (u_1)\mathrm{\Psi }_{}(u_2),`$
$`\mathrm{\Psi }_{}(u_2)\mathrm{\Phi }_\pm ^{}(u_1)={\displaystyle \frac{[u_1u_2\frac{1}{2}]^{}}{[u_1u_2+\frac{1}{2}]^{}}}\mathrm{\Phi }_\pm ^{}(u_1)\mathrm{\Psi }_{}(u_2).`$
### 3.3. Free fields resolution
The space of states for the deformed parafermion theory is constructed following a procedure well known in conformal field theory . This is done in two steps. The first step is to introduce a certain subspace $`\stackrel{~}{}_{m,l}`$ of the Fock space $`_{m,l}`$ using $`\eta (u)`$. The second step is to consider a complex consisting of these $`\stackrel{~}{}_{m,l}`$, in which the coboundary maps are given by powers of the screening charge $`X(u)`$. The space of states is then defined as the $`0`$-th cohomology of this complex.
Let
$`\eta _0:_{m,l}_{m+k,l+k+2}`$
denote the zeroth Fourier coefficient of $`\eta (u)`$. It is well defined provided $`mlmod2`$. We set
$`\stackrel{~}{}_{m,l}=\mathrm{Ker}\eta _0|_{_{m,l}}.`$
Then we have the following resolution of $`\stackrel{~}{}_{m,l}`$ by Fock spaces (see Appendix B)
(3.5)
$`0\stackrel{~}{}_{m,l}_{m,l}\stackrel{\eta _0}{}_{m+k,l+k+2}\stackrel{\eta _0}{}_{m+2k,l+2(k+2)}\stackrel{\eta _0}{}\mathrm{},`$
(3.6)
$`\mathrm{}\stackrel{\eta _0}{}_{m2k,l2(k+2)}\stackrel{\eta _0}{}_{mk,l(k+2)}\stackrel{\eta _0}{}\stackrel{~}{}_{m,l}0.`$
The following proposition shows that the operators $`S(u)`$, $`\mathrm{\Psi }_\pm (u)`$, $`\mathrm{\Phi }_\pm (u)`$ and $`\mathrm{\Phi }_\pm ^{}(u)`$ have a well-defined action on the subspace $`\stackrel{~}{}_{m,l}`$.
###### Proposition 3.4.
We have
$`\eta _0S(u)=S(u)\eta _0,`$
$`\eta _0\mathrm{\Psi }_\pm (u)=\mathrm{\Psi }_\pm (u)\eta _0,`$
$`\eta _0\mathrm{\Phi }_\pm (u)=\mathrm{\Phi }_\pm (u)\eta _0,`$
$`\eta _0\mathrm{\Phi }_\pm ^{}(u)=\mathrm{\Phi }_\pm ^{}(u)\eta _0.`$
Next we fix $`l,m`$ with $`0lk`$, $`lmmod2`$. Consider a sequence $`𝒞_{m,l}`$
(3.7)
$`\mathrm{}\stackrel{X_2}{}\stackrel{~}{}_{m,l2+2(k+2)}\stackrel{X_1}{}\stackrel{~}{}_{m,l}\stackrel{X_0}{}\stackrel{~}{}_{m,l2}\stackrel{X_1}{}\stackrel{~}{}_{m,l2(k+2)}\stackrel{X_2}{}\mathrm{},`$
defined by appropriate powers of the screening charge $`X(u)`$, i.e.,
$`X_{2j}=X(u)^{l+1}:\stackrel{~}{}_{m,l2j(k+2)}\stackrel{~}{}_{m,l22j(k+2)},`$
$`X_{2j+1}=X(u)^{kl+1}:\stackrel{~}{}_{m,l22j(k+2)}\stackrel{~}{}_{m,l2(j+1)(k+2)}.`$
The following can be shown in exactly the same way as in .
###### Proposition 3.5.
The maps $`X_j`$ are independent of $`u`$, and $`𝒞_{m,l}`$ is a cochain complex:
$`X_jX_{j1}=0(j).`$
The following statement concerning the cohomology group of this complex seems quite plausible (cf. ).
(3.8) $`H^j(𝒞_{m,l})=0(j0).`$
As we do not have a rigorous mathematical proof, we assume henceforth the validity of (3.8). By the Euler-Poincaré principle, the character of the remaining cohomology $`H^0(𝒞_{m,l})`$ then becomes
$`\mathrm{tr}_{H^0(𝒞_{m,l})}\left(x^{2kD}\right)=x^{\frac{k^2}{4(k+2)}}(x^{2k};x^{2k})_{\mathrm{}}c_{\mathrm{\Lambda }_k(m)}^{\mathrm{\Lambda }_k(l)}(\tau ),`$
$`Z={\displaystyle \underset{\genfrac{}{}{0pt}{}{0lk}{lmmod2}}{}}[l+1]\mathrm{tr}_{H^0(𝒞_{m,l})}\left(x^{2kD}\right)=x^{\frac{k+1}{k+2}}(x^{2k};x^{2k})_{\mathrm{}},`$
giving the same formula as that of the one point LHP (2.10).
###### Proposition 3.6.
$`X_j\mathrm{\Psi }_\pm (u)=\mathrm{\Psi }_\pm (u)X_j,`$
$`X_j\mathrm{\Phi }_\pm (u)=\mathrm{\Phi }_{}(u)X_j,`$
$`X_j\mathrm{\Phi }_\pm ^{}(u)=\mathrm{\Phi }_{}^{}(u)X_j.`$
This proposition ensures that $`\mathrm{\Phi }_\pm (u)`$, $`\mathrm{\Phi }_\pm ^{}(u)`$ and $`\mathrm{\Psi }_\pm (u)`$ give rise to well-defined operators on the cohomology $`H^0(𝒞_{m,l})`$. We abuse the notation and denote them by the same letters.
### 3.4. Identification with lattice theory
The construction of this Section is related to the lattice theory in Section 2 as follows. We make the following identification:
1. The space of states of the lattice model (with central height $`a`$ and boundary condition $`m`$) with the $`0`$-th cohomology of $`𝒞_{m,a1}`$,
$`_{m,a}=H^0(𝒞_{m,a1}).`$
2. The corner Hamiltonian $`H_C^{(a)}`$ with the grading operator $`D`$.
3. The half-infinite transfer matrices with the type I VO’s
$`\mathrm{\Phi }_N^{(a\epsilon ,a)}(u)=\mathrm{\Phi }_\epsilon (u)|_{H^0(𝒞_{m,a1})},`$
$`\mathrm{\Phi }_W^{(a\epsilon ,a)}(u)=\mathrm{\Phi }_\epsilon ^{}(u)|_{H^0(𝒞_{m,a1})}.`$
4. The creation/annihilation operators of particles and anti-particles with the type II VO’s $`\mathrm{\Psi }_\pm (u)`$.
As we already mentioned, with this identification the characters of the two spaces match. The commutation relations for VO’s of type I expected from the lattice theory (2.5)–(2.9) are recovered in Proposition 3.1. In Section 5, we will comment about the agreement with the known results on the excitation spectrum and the $`S`$-matrix in the scaling limit.
Correlation functions and form factors of local operators are given as traces of operators acting on the physical space $`H^0(𝒞_{m,l})`$. Let $`𝒪`$ stand for $`x^{2kD}`$ times a product of operators of the form $`\mathrm{\Psi }_\epsilon (u)`$, $`\mathrm{\Phi }_\epsilon (u)`$ and $`\mathrm{\Phi }_\epsilon ^{}(u)`$. For simplicity we consider the case where there are an equal number of $`\mathrm{\Psi }_+(u)`$’s and $`\mathrm{\Psi }_{}(u)`$’s. Proposition 3.6 shows that
$`𝒪_i\eta _0=\eta _0𝒪_i,`$
$`X_j𝒪_i=𝒪_{1i}X_j(jimod2),`$
where $`𝒪_0=𝒪`$, and $`𝒪_1`$ signifies the operator obtained from $`𝒪`$ by negating the indices $`\epsilon `$ of $`\mathrm{\Phi }_\epsilon (u)`$ and $`\mathrm{\Phi }_\epsilon ^{}(u)`$. The resolutions (3.5),(3.6),(3.7) afford a procedure for computing the trace as follows.
(3.9) $`\mathrm{tr}_{H^0(𝒞_{m,l})}(𝒪)={\displaystyle \underset{s}{}}\mathrm{tr}_{\stackrel{~}{}_{m,l2s(k+2)}}(𝒪_0){\displaystyle \underset{s}{}}\mathrm{tr}_{\stackrel{~}{}_{m,l22s(k+2)}}(𝒪_1),`$
$`\mathrm{tr}_{\stackrel{~}{}_{m,l}}(𝒪_i)`$ $`=`$ $`{\displaystyle \underset{n0}{}}(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)n}}(𝒪_i)`$
$`=`$ $`{\displaystyle \underset{n<0}{}}(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)n}}(𝒪_i).`$
Taking $`𝒪=x^{2kD}\mathrm{\Phi }_\epsilon ^{}(u)\mathrm{\Phi }_\epsilon (u)`$, we have verified directly that the formula thus obtained reproduces the LHP (2.11)–(2.12) for neighboring lattice sites.
### 3.5. Fusion of parafermions and $`W`$ currents
The type II VO’s $`\mathrm{\Psi }_{}(u)`$, $`\mathrm{\Psi }_+(u)`$ play the role of creation operators of excitations over the ground state. Besides them, there are altogether $`k1`$ kinds of ‘elementary’ particles in regime II, as expected from the level-rank duality . The standard fusion procedure provides us with a free field realization $`\mathrm{\Psi }_a(u)`$ defined recursively as
$`\mathrm{\Psi }_1(u^{}{\displaystyle \frac{a}{2}})\mathrm{\Psi }_a(u+{\displaystyle \frac{1}{2}})`$
$`=`$ $`{\displaystyle \frac{1}{1\frac{z}{z^{}}}}x^{\frac{a(a+1)}{k}}{\displaystyle \frac{(x^{2a},x^{2a+2k+2};x^{2k})_{\mathrm{}}}{(x^{2k},x^{2k+2};x^{2k})_{\mathrm{}}}}\mathrm{\Psi }_{a+1}(u)+O(1)(u^{}u),`$
where $`\mathrm{\Psi }_1(u)=(xx^1)\mathrm{\Psi }_{}(u)`$. A more explicit expression is given in (A.1). We have the following commutation relations
$`\mathrm{\Psi }_a(u_2)\mathrm{\Phi }_\pm (u_1)={\displaystyle \frac{[u_1u_2+\frac{a}{2}]^{}}{[u_1u_2\frac{a}{2}]^{}}}\mathrm{\Phi }_\pm (u_1)\mathrm{\Psi }_a(u_2),`$
$`\mathrm{\Psi }_a(u_2)\mathrm{\Psi }_b(u_1)`$
$`=`$ $`{\displaystyle \underset{s=1}{\overset{a}{}}}{\displaystyle \underset{s^{}=1}{\overset{b}{}}}{\displaystyle \frac{[u_1u_2+1+\frac{ab}{2}(ss^{})]^{}}{[u_1u_21+\frac{ab}{2}(ss^{})]^{}}}\mathrm{\Psi }_b(u_1)\mathrm{\Psi }_a(u_2)`$
$`=`$ $`{\displaystyle \frac{[u+\frac{a+b}{2}]^{}}{[u\frac{a+b}{2}]^{}}}{\displaystyle \frac{[u+\frac{|ab|}{2}]^{}}{[u\frac{|ab|}{2}]^{}}}{\displaystyle \underset{s=1}{\overset{\mathrm{min}(a,b)1}{}}}{\displaystyle \frac{[u+\frac{|ab|}{2}+s]_{}^{}{}_{}{}^{2}}{[u\frac{|ab|}{2}s]_{}^{}{}_{}{}^{2}}}\mathrm{\Psi }_b(u_1)\mathrm{\Psi }_a(u_2),`$
where $`u=u_1u_2`$. These relations agree with the known results about the excitation spectra (5.1) and the scattering matrices (5.6) in the scaling field theory discussed in Section 5.
The operator $`\mathrm{\Psi }_k(u)`$ commutes with $`\mathrm{\Psi }_\pm (v)`$ and $`X_j`$, commutes or anticommutes with $`\eta _0`$, and anticommutes with $`\mathrm{\Phi }_\pm (v)`$. It can be shown that it is also invertible. We expect that on the cohomology it is independent of $`u`$ and defines an isomorphism $`\iota :H^0(𝒞_{m,l})H^0(𝒞_{m+2k,l})`$. We also expect that $`\mathrm{\Psi }_{k1}(u)`$ defines the same operator as $`\iota \mathrm{\Psi }_+(u)`$.
The level-rank duality also suggests the existence of the deformed $`W_k`$ currents in the parafermionic description of the ABF models. Indeed, the first deformed $`W`$ current $`W^1(u)`$ can be obtained either as fusion of type I VO’s ($`z_2x^{2(k+2)}z_1`$)
$`\mathrm{\Phi }_{}(u_2)\mathrm{\Phi }_+^{}(u_1)`$
$`=`$ $`\left(1x^{2(k+2)}{\displaystyle \frac{z_1}{z_2}}\right){\displaystyle \frac{x^{\frac{1}{k}}}{[k+1]_x}}{\displaystyle \frac{(x^{2k},x^4;x^{2k})_{\mathrm{}}}{(x^2,x^2;x^{2k})_{\mathrm{}}}}W^1(u_1)+\mathrm{},`$
or as that of type II VO’s ($`z_2x^{k+2}z_1`$)
$`\mathrm{\Psi }_{}(u_2)\mathrm{\Psi }_+(u_1)`$
$`=`$ $`{\displaystyle \frac{1}{1x^{k+2}\frac{z_1}{z_2}}}{\displaystyle \frac{x^{\frac{2}{k}+1}}{(xx^1)^2[k+1]_x}}{\displaystyle \frac{(x^4;x^{2k})_{\mathrm{}}}{(x^{2k};x^{2k})_{\mathrm{}}}}W^1(u_1\frac{k+1}{2})+\mathrm{}.`$
Explicit formulas of $`W^1(u)`$, as well as those of the higher currents $`W^j(u)`$ ($`j=2,3,\mathrm{}`$), are given in Appendix A. We expect that these $`W^j(u)`$ ($`j1`$) generate the same deformed $`W`$ algebra for $`\text{sl}_N`$ in , under the following identification
$`N=k,q=x^{2(k+1)},t=x^{2(k+2)}(p=qt^1=x^2).`$
We have checked a part of the relations for $`W^j(u)`$ (eq.(8) in ), but have not verified such relations as $`W^k(u)=1`$ which are expected to hold only at the level of the cohomology.
## 4. Form Factors
### 4.1. Traces of type II operators
As a simplest example of form factors of local operators, let us consider the quantity
(4.1) $`Q_a^{(n,n)}(m)=\left(Zg^n\right)^1[a]`$
$`\times \mathrm{tr}_{_{m,a}}\left(x^{2kD}\mathrm{\Psi }_+(v_1)\mathrm{}\mathrm{\Psi }_+(v_n)\mathrm{\Psi }_{}(v_1^{})\mathrm{}\mathrm{\Psi }_{}(v_n^{})\right).`$
To simplify the notation, we have not exhibited explicitly the dependence of $`Q_a^{(n,n)}(m)`$ on the parameters $`v_j,v_j^{}`$. For the same reason we will often suppress the superscript $`(n,n)`$. Note that $`Q_a(m)=0`$ for $`ammod2`$.
Hereafter we set
$`\tau ={\displaystyle \frac{ik}{\pi }}\mathrm{log}x.`$
It is a standard task to compute traces of bosonic operators over the Fock space. After the working outlined in Appendix C, we find the following expression.
(4.2) $`Q_a(m)={\displaystyle \underset{\genfrac{}{}{0pt}{}{\mu _1,\mathrm{},\mu _n=\pm 1}{\nu _1,\mathrm{},\nu _n=\pm 1}}{}}\kappa (v,\mu ,\nu ){\displaystyle \underset{i=1}{\overset{n}{}}}\mu _i\nu _i`$
$`\times {\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{[v_jv_i+\frac{\mu _i\mu _j}{2}]^{}}{[v_jv_i1]^{}}}{\displaystyle \frac{F(v_jv_i)}{F(0)}}`$
$`\times {\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{[v_j^{}v_i^{}\frac{\nu _i\nu _j}{2}]^{}}{[v_j^{}v_i^{}1]^{}}}{\displaystyle \frac{F(v_j^{}v_i^{})}{F(0)}}`$
$`\times {\displaystyle \underset{i,j=1}{\overset{n}{}}}{\displaystyle \frac{[v_j^{}v_i\frac{k}{2}\frac{\mu _i+\nu _j}{2}]^{}}{[v_j^{}v_i\frac{k}{2}]^{}}}{\displaystyle \frac{F(0)}{F(v_j^{}v_i\frac{k}{2})}}`$
$`\times \mathrm{\Gamma }_{m,a1}({\displaystyle \frac{\tau }{2k}}(\mu \nu ),{\displaystyle \frac{\tau }{k}}({\displaystyle \frac{2v}{k}}{\displaystyle \frac{\mu +\nu }{2}})\left|\tau \right).`$
In the above, we have set $`v=_{i=1}^n(v_i^{}v_i)`$, $`\mu =_{i=1}^n\mu _i`$, $`\nu =_{i=1}^n\nu _i`$. The function $`\mathrm{\Gamma }_{m,l}(y_1,y_2|\tau )=\mathrm{\Gamma }_{m,l}^{(0,0)}(y_1,y_2|\tau )`$ is defined in (C). The functions $`F(v)`$ and $`\kappa (v,\mu ,\nu )`$ are defined as follows.
(4.3) $`F(v)={\displaystyle \frac{(x^{2(k+1+v)},x^{2(k+1v)};x^{2k},x^{2k})_{\mathrm{}}}{(x^{2(k1+v)},x^{2(k1v)};x^{2k},x^{2k})_{\mathrm{}}}},`$
$`\kappa (v,\mu ,\nu )={\displaystyle \frac{[a]}{Z}}\left(i\tau {\displaystyle \frac{\eta (\tau )^3}{[1]^{}}}\right)^nx^{(\frac{\mu +\nu }{k}\frac{2}{k}n)v+\frac{n^2}{2}+\frac{\mu ^2+\nu ^2}{4k}\frac{kn}{4}\frac{k+1}{2k}\mu \nu +\frac{kc}{12}}`$
with $`c=2(k1)/(k+2)`$.
Using (C.7),(C.8) we see that $`Q_a(m+2k)=Q_{k+2a}(m+k)=Q_a(m)`$.
Remark. We see from (C.3), (C.4) that the contribution of the oscillator part to the trace (4.1) is convergent if $`x^{\epsilon _2\epsilon _1}<|z_1/z_2|<x^{\epsilon _2\epsilon _12k}`$. For $`k>2`$ there is a non-empty domain of convergence common to all $`\epsilon _1,\epsilon _2`$. The case $`k=2`$ of the Ising model is exceptional and needs a separate treatment by analytic continuation. For the rest of the paper we assume $`k3`$.
### 4.2. Neighboring heights
As the next example, let us consider the trace involving two type I VO’s,
$`Q_{b,a}(m)`$ $`=`$ $`\left(Zg^n\right)^1[a]\mathrm{tr}_{_{m,a}}(x^{2kD}\mathrm{\Phi }^{(a,b)}(u)\mathrm{\Phi }^{(b,a)}(u)`$
$`\times \mathrm{\Psi }_+(v_1)\mathrm{}\mathrm{\Psi }_+(v_n)\mathrm{\Psi }_{}(v_n^{})\mathrm{}\mathrm{\Psi }_{}(v_n^{})).`$
Here $`\mathrm{\Phi }^{(a\epsilon ,a)}(u)`$ means $`\mathrm{\Phi }_\epsilon (u)|_{_{m,a}}`$, and similarly for $`\mathrm{\Phi }^{(a,b)}(u)`$.
We have already mentioned that the two point LHP for the neighboring height variables (2.11),(2.12) can be expressed simply in terms of one point LHP’s. Let us apply the same argument to compute $`Q_{b,a}(m)`$. The first inversion identity (3.3) entails that
(4.4) $`Q_{a+1,a}(m)+Q_{a1,a}(m)=Q_a(m),`$
with $`Q_a(m)`$ given by (4.1). On the other hand, the second inversion relation (3.4) together with the cyclicity of the trace implies
(4.5) $`Q_{b,b1}(m)+Q_{b,b+1}(m)=Q_b(m+1)G.`$
The factor
$`G={\displaystyle \underset{j=1}{\overset{n}{}}}{\displaystyle \frac{[uv_j+\frac{1}{2}+\frac{k}{2}]^{}}{[uv_j\frac{1}{2}+\frac{k}{2}]^{}}}{\displaystyle \frac{[uv_j^{}\frac{1}{2}]^{}}{[uv_j^{}+\frac{1}{2}]^{}}}`$
arises from the commutation relations between type I and type II VO’s. Solving the relations (4.4),(4.5) under the condition $`Q_{2,1}(m)=Q_1(m)`$, $`Q_{1,2}(m)=Q_1(m+1)`$ (which can be verified using the integral representations), we obtain
$`Q_{a+1,a}(m)={\displaystyle \underset{\genfrac{}{}{0pt}{}{1sa}{samod2}}{}}Q_s(m){\displaystyle \underset{\genfrac{}{}{0pt}{}{1sa}{samod2}}{}}Q_s(m+1)G,`$
$`Q_{a1,a}(m)={\displaystyle \underset{\genfrac{}{}{0pt}{}{1s<a}{samod2}}{}}Q_s(m)+{\displaystyle \underset{\genfrac{}{}{0pt}{}{1s<a}{samod2}}{}}Q_s(m+1)G.`$
The above reasoning does not seem to generalize easily to the case where more than two VO’s of type I are present.
## 5. Scaling limit
Basic facts about the general RSOS models and their scaling limit have been worked out by Bazhanov and Reshetikhin , by diagonalizing the row-to-row transfer matrix. First we briefly review their results specializing to the present case of the ABF models in regime II. Then we work out new expressions for form factors by taking the scaling limit of the formulas on the lattice.
### 5.1. Review of known results
Denote by $`H`$ the Hamiltonian of the lattice model related with the row-to-row transfer matrix $`T`$ as
$`H={\displaystyle \frac{1}{4\pi \delta }}{\displaystyle \frac{d}{du}}\mathrm{log}T(u)|_{u=0},`$
where $`\delta `$ is the lattice spacing. The excitation spectrum over the ground state found from the Bethe ansatz has the form
(5.1) $`ϵ_a(v)={\displaystyle \frac{1}{4\pi \delta }}{\displaystyle \frac{d}{dv}}\mathrm{log}{\displaystyle \frac{[v+\frac{a}{2}]^{}}{[v\frac{a}{2}]^{}}},(1ak1).`$
The energies $`|ϵ_a(v)|`$ are periodic functions with the period $`\frac{i\pi }{\mathrm{log}x}`$, and have a minimum at $`v=\frac{i\pi }{2\mathrm{log}x}`$. The analysis of this function shows that there is a non-zero gap in the spectrum, and the corresponding continuous model is a massive field theory.
Introduce the rapidity variable $`\beta `$ by
(5.2) $`v={\displaystyle \frac{ik}{2\pi }}\beta {\displaystyle \frac{i\pi }{2\mathrm{log}x}},`$
and let
(5.3) $`p=e^{\frac{\pi ^2}{(k+2)\mathrm{log}x}}`$
be the temperature parameter of the lattice model. In the scaling limit
(5.4) $`\mathrm{log}x0,\delta 0,p^{\frac{k+2}{k}}\delta ^1\text{const}={\displaystyle \frac{kM}{2\mathrm{sin}\frac{\pi }{k}}},`$
while keeping $`\beta `$ fixed, (5.1) gives a massive relativistic spectrum
$`\underset{\mathrm{log}x0}{lim}ϵ_a\left({\displaystyle \frac{ik}{2\pi }}\beta {\displaystyle \frac{i\pi }{\mathrm{log}x}}\right)=M_a\mathrm{cosh}\beta .`$
Here the mass $`M_a`$ of the particle $`a`$ is defined by
$`M_a={\displaystyle \frac{\mathrm{sin}\frac{\pi a}{k}}{\mathrm{sin}\frac{\pi }{k}}}M.`$
The scattering matrices $`S_{ab}(\beta )`$ of these particles are diagonal and have a very simple form. For the fundamental particle/antiparticle, it is given by ‘minimal’ $`S`$ matrices of the $`_k`$ symmetric model proposed for the first time in ,
(5.5) $`S_{11}(\beta )=S_{\overline{1}\overline{1}}(\beta )={\displaystyle \frac{\mathrm{sinh}(\frac{\beta }{2}+\frac{i\pi }{k})}{\mathrm{sinh}(\frac{\beta }{2}\frac{i\pi }{k})}},`$
$`S_{1\overline{1}}(\beta )=S_{\overline{1}1}(\beta )=S_{11}(i\pi \beta ).`$
In general, we have
(5.6) $`S_{ab}(\beta )=f_{a+b}(\beta )f_{|ab|}(\beta ){\displaystyle \underset{s=1}{\overset{\mathrm{min}(a,b)1}{}}}f_{|ab|+2s}(\beta )^2,`$
where
$`f_A(\beta )={\displaystyle \frac{\mathrm{sinh}(\frac{\beta }{2}+\frac{i\pi A}{2k})}{\mathrm{sinh}(\frac{\beta }{2}\frac{i\pi A}{2k})}}.`$
The particles in the scaling theory are not self-conjugate except for the case $`k=2`$ of the Ising model. The charge conjugation identifies the antiparticle $`\overline{a}`$ of $`a`$ with the particle $`ka`$, reflecting the additional $`_2`$ symmetry of the model.
It has been found in that the ultraviolet properties of the scaling theory are described by the parafermionic CFT with the central charge
$`c=2{\displaystyle \frac{k1}{k+2}}.`$
According to these works, one can treat the model in the scaling region as the CFT perturbed by the first energy operator with the left and right conformal dimensions
$`\mathrm{\Delta }_{0,2}={\displaystyle \frac{2}{k+2}}.`$
All these results agree well with the algebraic picture we have followed. The vertex operators of type II are identified with operators that create eigenstates of the row-to-row transfer matrix . Writing the formal limit (5.4) of $`\mathrm{\Psi }_a(v)`$ as $`𝒵_a(\beta )`$, where $`v`$ and $`\beta `$ are related as in (5.2), we find that the commutation relations of Proposition 3.2 and (3.5) become
(5.7) $`𝒵_a(\beta _1)𝒵_b(\beta _2)=S_{ab}(\beta _1\beta _2)𝒵_b(\beta _2)𝒵_a(\beta _1).`$
Thus the operators $`𝒵_a`$ can be interpreted as generators of the Zamolodchikov-Faddeev algebra in the angular quantization approach . The commutation relations between $`\mathrm{\Phi }_\pm (0)`$ and $`\mathrm{\Psi }_a(v)`$ (Proposition 3.3) ensure that the states created by acting with the type II operators on the vacuum have the eigenvalues (5.1) of the corresponding Hamiltonian.
Our present aim is to take the continuous limit of the formulas in the previous section, and to interpret them as $`_k`$-neutral form factors of some operators in the $`_k`$-symmetric model. We will however discuss neither the Lagrangian description of the present model nor the problems of identification and normalization of local operators.
General aspects of the form factors in diagonal scattering theories with $`_k`$ symmetry were discussed in , where a recursive system of functional equations have been written. Another well-known example of theories with such a symmetry is the affine $`A_{k1}`$ Toda models. Its $`S`$ matrix for fundamental particles differs from $`S_{11}`$ (5.5) by a coupling-dependent factor which has no poles and zeros in the physical region. For the affine Toda models, the two- and four-particle form factors have been determined in . Our results below are very similar to those for the affine Toda case.
### 5.2. Projection operators
Let us denote by $`|0_m`$ the $`2k`$ degenerate ground states in regime II. They are related to each other by a spatial translation. For the discussion of the continuous limit it is more convenient to deal with eigenstates with respect to the translation operator on the lattice,
(5.8) $`{\displaystyle \frac{1}{\sqrt{2k}}}{\displaystyle \underset{m^{}=k}{\overset{k1}{}}}e^{\frac{i\pi mm^{}}{k}}|0_m^{}.`$
Let $`\mathrm{Pr}_a`$ stand for the projection operator onto the sector where the central height takes the value $`a`$. We shall focus attention to form factors of the linear transform
$`\widehat{\mathrm{Pr}}_a={\displaystyle \frac{1}{\sqrt{2k}}}{\displaystyle \underset{a^{}=1}{\overset{k+1}{}}}{\displaystyle \frac{\mathrm{sin}\frac{\pi aa^{}}{k+2}}{\mathrm{sin}\frac{\pi a^{}}{k+2}}}\mathrm{Pr}_a^{}`$
with respect to the translationally invariant vacuum $`m=0`$ in (5.8). Thus the quantities we study in the continuous limit are
(5.9) $`\widehat{𝒬}_a^{(n,n)}={\displaystyle \frac{1}{2k}}{\displaystyle \underset{a^{}=1}{\overset{k+1}{}}}{\displaystyle \underset{m=k}{\overset{k1}{}}}{\displaystyle \frac{\mathrm{sin}\frac{\pi aa^{}}{k+2}}{\mathrm{sin}\frac{\pi a^{}}{k+2}}}Q_a^{}^{(n,n)}(m),`$
where $`Q_a^{}^{(n,n)}(m)`$ is given by (4.1). In the angular quantization approach to affine Toda field theory, similar operators are attributed to the exponential operators . Note that $`\widehat{𝒬}_a^{(n,n)}=0`$ if $`a`$ is even.
### 5.3. Two-particle form factors
To illustrate the procedure of taking the scaling limit, let us consider in some details the simplest case $`n=1`$, corresponding to the particle-antiparticle form factors.
To find the limit (5.4) of $`\widehat{𝒬}_a^{(1,1)}`$ it is convenient to introduce the conjugate modulus transformation. The standard formulas for the theta functions give
$`[u]=\sqrt{{\displaystyle \frac{ik}{\tau (k+2)}}}e^{i\frac{k+2}{4k}\pi \tau }\theta _1({\displaystyle \frac{\pi u}{k+2}};{\displaystyle \frac{k}{\tau (k+2)}}),`$
$`[u]^{}=\sqrt{{\displaystyle \frac{i}{\tau }}}e^{\frac{i\pi }{4}\tau }\theta _1({\displaystyle \frac{\pi u}{k}};{\displaystyle \frac{1}{\tau }}),`$
$`\theta _1(u;\tau )=2{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}(1)^ne^{i\pi \tau (n\frac{1}{2})^2}\mathrm{sin}(2n1)u.`$
To deal with the function $`\mathrm{\Gamma }_{m,l}`$ we borrow the technique of , as mentioned in Appendix C. With the notation $`\beta =\beta _1\beta _1^{}`$, the expression (4.2) can be rewritten in the following form.
(5.10) $`Q_a^{(1,1)}(m)={\displaystyle \frac{e^{\frac{i\pi }{2k}\tau (\frac{i\beta }{\pi }+1)^2}}{k+2}}\eta (\frac{1}{\tau })^2{\displaystyle \frac{\theta _1(\frac{\pi a}{k+2};\frac{k}{\tau (k+2)})}{\theta _1(\frac{\pi }{k};\frac{1}{\tau })}}{\displaystyle \frac{F(0)}{F\left(\frac{k}{2\pi i}(\beta i\pi )\right)}}`$
(5.11) $`\times {\displaystyle \underset{\mu ,\nu =\pm 1}{}}\mu \nu {\displaystyle \frac{\theta _1(\frac{i\beta }{2}+\frac{\pi }{2k}(\mu +\nu )+\frac{\pi }{2};\frac{1}{\tau })}{\theta _1(\frac{i\beta }{2}+\frac{\pi }{2};\frac{1}{\tau })}}\mathrm{\hspace{0.17em}2}{\displaystyle \underset{l^{}=0}{\overset{k}{}}}{\displaystyle \underset{m^{}=0}{\overset{k1}{}}}\mathrm{sin}{\displaystyle \frac{\pi a(l^{}+1)}{k+2}}`$
$`\times e^{\frac{i\pi mm^{}}{k}}\mathrm{\Gamma }_{m^{},l^{}}({\displaystyle \frac{\mu \nu }{2k}},{\displaystyle \frac{i\beta }{\pi k}}{\displaystyle \frac{\mu +\nu }{2k}}|{\displaystyle \frac{1}{\tau }}).`$
From the sum over $`m`$ and $`a^{}`$, the linear transform (5.9) selects only one term ($`m^{}=0,l^{}=a1`$). Thus we have to analyze the limiting behavior of $`\mathrm{\Gamma }_{0,l}`$,
$`\mathrm{\Gamma }_{0,l}({\displaystyle \frac{\mu \nu }{2k}},{\displaystyle \frac{i\beta }{\pi k}}{\displaystyle \frac{\mu +\nu }{2k}}|{\displaystyle \frac{1}{\tau }})`$
$`=`$ $`{\displaystyle \frac{1}{\eta (\frac{1}{\tau })^2}}\left({\displaystyle \underset{n_1|n_2|}{}}{\displaystyle \underset{n_1>|n_2|}{}}\right)(1)^{2n_1}p^{2\frac{k+2}{k}\left(\frac{(l+1+2n_1(k+2))^2}{4(k+2)}n_2^2k\right)}`$
$`\times e^{\frac{\pi i}{2k}(\mu \nu )(l+1+2(k+2)n_1)+2\pi i(\frac{i\beta }{\pi }+\frac{\mu +\nu }{2})n_2},`$
where $`p`$ is defined in (5.3) and the sum is taken over $`n_1,n_2\frac{1}{2}`$ with $`n_1n_2`$. The leading contribution to the sums comes from the term with $`n_1=n_2=0`$. In the limit (5.4), we thus find
$`\mathrm{\Gamma }_{0,l}({\displaystyle \frac{\mu \nu }{2k}},{\displaystyle \frac{i\beta }{\pi k}}{\displaystyle \frac{\mu +\nu }{2k}}|{\displaystyle \frac{1}{\tau }})=p^{2\frac{k+2}{k}(\mathrm{\Delta }_{0,l}\frac{c}{24})}(e^{\frac{\pi i}{2k}(\mu \nu )(l+1)}+O(p^{2\frac{k+2}{k}})).`$
Here $`\mathrm{\Delta }_{0,l}`$ stands for the conformal dimension of $`_k`$-neutral primary fields in the parafermionic CFT
$`\mathrm{\Delta }_{0,l}={\displaystyle \frac{(l+1)^21}{4(k+2)}}(l0mod2).`$
The limit of the remaining terms can be computed directly. For example,
$`\underset{\mathrm{log}x0}{lim}{\displaystyle \frac{F(0)}{F\left(\frac{k}{2\pi i}(\beta \pi i)\right)}}=F_{1\overline{1}}^{min}(\beta ),`$
where we introduced the minimal form factor for the particle-antiparticle scattering
$`F_{1\overline{1}}^{min}(\beta )={\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }(\frac{i\beta }{2\pi }+\frac{1}{k}+n+\frac{1}{2})\mathrm{\Gamma }(\frac{i\beta }{2\pi }+\frac{1}{k}+n\frac{1}{2})\mathrm{\Gamma }(n\frac{1}{k})^2}{\mathrm{\Gamma }(\frac{i\beta }{2\pi }\frac{1}{k}+n+\frac{1}{2})\mathrm{\Gamma }(\frac{i\beta }{2\pi }\frac{1}{k}+n\frac{1}{2})\mathrm{\Gamma }(n+\frac{1}{k})^2}},`$
normalized as $`F_{1\overline{1}}^{min}(i\pi )=1`$. As for the sum over different components of vertex operators we use
$`{\displaystyle \underset{\mu ,\nu =\pm 1}{}}\mu \nu e^{\frac{\pi i}{2k}(\mu \nu )a}\mathrm{cosh}\left({\displaystyle \frac{\beta }{2}}+{\displaystyle \frac{i\pi }{2k}}(\mu +\nu )\right)`$
$`=4\mathrm{cosh}{\displaystyle \frac{\beta }{2}}\mathrm{sin}{\displaystyle \frac{\pi (a1)}{2k}}\mathrm{sin}{\displaystyle \frac{\pi (a+1)}{2k}}.`$
This leads us to the simple expression for the two-point form factor of the projection operators
$`\widehat{𝒬}_a^{(1,1)}=2{\displaystyle \frac{\mathrm{sin}\frac{\pi (a1)}{2k}\mathrm{sin}\frac{\pi (a+1)}{2k}}{\mathrm{sin}\frac{\pi }{k}}}p^{2\frac{k+2}{k}\mathrm{\Delta }_{0,a1}}F_{1\overline{1}}^{min}(\beta )\times \left(1+o(1)\right).`$
As we expect from the results of , there is no singularity at $`\beta =i\pi `$.
In the next subsection, we compute the scaling limit for the case of multi-particles in a similar manner. We note in passing that
$`\underset{\mathrm{log}x0}{lim}{\displaystyle \frac{F\left(\frac{k}{2\pi i}\beta \right)}{F(0)}}={\displaystyle \frac{F_{11}^{min}(\beta )}{\mathrm{sinh}\frac{\beta }{2}\mathrm{sinh}(\frac{\beta }{2}+\frac{i\pi }{k})}},`$
where the minimal two particle form factor $`F_{11}^{min}(\beta )`$ reads
$`F_{11}^{min}(\beta )=\mathrm{sinh}{\displaystyle \frac{\beta }{2}}\mathrm{sinh}\left({\displaystyle \frac{\beta }{2}}+{\displaystyle \frac{i\pi }{k}}\right)`$
$`\times {\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }(\frac{i\beta }{2\pi }\frac{1}{k}+n)\mathrm{\Gamma }(\frac{i\beta }{2\pi }\frac{1}{k}+n)\mathrm{\Gamma }(n+\frac{1}{k})^2}{\mathrm{\Gamma }(\frac{i\beta }{2\pi }+\frac{1}{k}+n)\mathrm{\Gamma }(\frac{i\beta }{2\pi }+\frac{1}{k}+n)\mathrm{\Gamma }(n\frac{1}{k})^2}}.`$
### 5.4. Many particles
To present the result in the general case, it is convenient to use the formal bosonization rule as explained in . Let us introduce
(5.12) $`_1^a(\beta )={\displaystyle \underset{\mu =\pm 1}{}}\mu e^{\frac{\mu \pi ia}{2k}}𝒵_{1,\mu }(\beta ),`$
(5.13) $`_{\overline{1}}^a(\beta )={\displaystyle \underset{\mu =\pm 1}{}}\mu e^{\frac{\mu \pi ia}{2k}}𝒵_{\overline{1},\mu }(\beta ).`$
Here $`𝒵_{1,\mu }(\beta )`$ and $`𝒵_{\overline{1},\mu }(\beta )`$ are some operators, for which we assume the contraction rules
$`𝒵_{1,\mu }(\beta _1)𝒵_{1,\nu }(\beta _2)=𝒵_{\overline{1},\mu }(\beta _1)𝒵_{\overline{1},\nu }(\beta _2)`$
$`={\displaystyle \frac{F_{11}^{min}(\beta )}{\mathrm{sinh}(\frac{\beta }{2}+\frac{i\pi }{k})\mathrm{sinh}(\frac{\beta }{2}\frac{i\pi }{k})}}{\displaystyle \frac{\mathrm{sinh}\left(\frac{\beta }{2}\frac{i\pi }{2k}(\mu \nu )\right)}{2\mathrm{sin}\frac{\pi }{k}\mathrm{sinh}\frac{\beta }{2}}},`$
$`𝒵_{1,\mu }(\beta _1)𝒵_{\overline{1},\nu }(\beta _2)=𝒵_{\overline{1},\mu }(\beta _1)𝒵_{1,\nu }(\beta _2)`$
$`=F_{1\overline{1}}^{min}(\beta ){\displaystyle \frac{\mathrm{cosh}\left(\frac{\beta }{2}+\frac{i\pi }{2k}(\mu +\nu )\right)}{2\mathrm{sin}\frac{\pi }{k}\mathrm{cosh}\frac{\beta }{2}}},`$
where $`\beta =\beta _1\beta _2`$, and the Wick theorem as in (C.2) where $`C`$ is replaced by
$`C_1={\displaystyle \frac{1}{\sqrt{2\mathrm{sin}\frac{\pi }{k}}}}.`$
The prescription (5.12), (5.13) is analogous to the one for the free field representation of form factors of the ‘exponential fields’ in the sine-Gordon and affine Toda theories. The exponential factors $`e^{\pm \frac{\pi ia}{2k}}`$ are the only remnant of the ‘zero mode’ part.
The limit of the formula (5.9) can be written compactly as
(5.14) $`\underset{\mathrm{log}x0}{lim}\left(p^{2\frac{k+2}{k}\mathrm{\Delta }_{0,a1}}\widehat{𝒬}_a^{(n,n)}\right)`$
$`=_{\overline{1}}^a(\beta _1)\mathrm{}_{\overline{1}}^a(\beta _n)_1^a(\beta _1^{})\mathrm{}_1^a(\beta _n^{})`$
$`={\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{F_{11}^{min}(\beta _i\beta _j)}{(x_i\omega ^2x_j)(x_i\omega ^2x_j)}}{\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{F_{11}^{min}(\beta _i^{}\beta _j^{})}{(y_i\omega ^2y_j)(y_i\omega ^2y_j)}}`$
$`\times {\displaystyle \underset{1i,jn}{}}{\displaystyle \frac{F_{1\overline{1}}^{min}(\beta _i\beta _j^{})}{x_i+y_j}}\times C_1^{2n}\mathrm{\hspace{0.17em}2}^{2n(n1)}\sigma _n^{n1}\tau _n^{n1}R_{\frac{a+1}{2}}^{(n,n)}(x;y).`$
The notation is as follows. We set $`x_i=e^{\beta _i}`$, $`y_i=e^{\beta _i^{}}`$, $`\sigma _r=\sigma _r(x)`$, $`\tau _r=\sigma _r(y)`$ with
$`{\displaystyle \underset{j=1}{\overset{n}{}}}(t+x_j)={\displaystyle \underset{r=0}{\overset{n}{}}}t^{nr}\sigma _r(x),`$
$`\omega =e^{\pi i/k}`$, and $`\{l\}=\omega ^l\omega ^l`$. The polynomials $`R_\alpha ^{(m,n)}(x;y)`$ are defined by
$`R_\alpha ^{(m,n)}(x;y)={\displaystyle \underset{\genfrac{}{}{0pt}{}{\mu _1,\mathrm{},\mu _m=\pm 1}{\nu _1,\mathrm{},\nu _n=\pm 1}}{}}{\displaystyle \underset{i=1}{\overset{m}{}}}\mu _i\omega ^{(\alpha m+n)\mu _i}{\displaystyle \underset{j=1}{\overset{n}{}}}\nu _j\omega ^{(\alpha +mn)\nu _j}`$
$`\times {\displaystyle \underset{1i<jm}{}}{\displaystyle \frac{x_i\omega ^{\mu _i}x_j\omega ^{\mu _j}}{x_ix_j}}{\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{y_i\omega ^{\nu _i}y_j\omega ^{\nu _j}}{y_iy_j}}{\displaystyle \underset{\genfrac{}{}{0pt}{}{1im}{1jn}}{}}(x_i\omega ^{\mu _i}+y_j\omega ^{\nu _j}).`$
For example,
$`R_\alpha ^{(1,1)}=\{\alpha \}\{\alpha 1\}(\sigma _1+\tau _1),`$
$`R_\alpha ^{(2,2)}=\{\alpha \}\{\alpha 1\}(\{\alpha \}\{\alpha 1\}(\sigma _1+\tau _1)(\sigma _2\tau _1+\sigma _1\tau _2)`$
$`+\{\alpha +1\}\{\alpha 2\}(\sigma _2\tau _2)^2),`$
$`R_\alpha ^{(3,3)}=\{\alpha \}^3\{\alpha 1\}^3(\sigma _1+\tau _1)(\sigma _3+\sigma _2\tau _1+\sigma _1\tau _2+\tau _3)(\sigma _3\tau _2+\sigma _2\tau _3)`$
$`\{\alpha +2\}\{\alpha \}^2\{\alpha 1\}^2\{\alpha 3\}(\sigma _1+\tau _1)(\sigma _3+\tau _3)(\sigma _3\tau _2+\sigma _2\tau _3)`$
$`+\{\alpha +1\}^2\{\alpha \}\{\alpha 1\}\{\alpha 2\}^2(\sigma _2\tau _1+\sigma _1\tau _2)(\sigma _3+\tau _3)^2`$
$`+\{\alpha +1\}\{\alpha \}^2\{\alpha 1\}^2\{\alpha 2\}((\sigma _1+\tau _1)(\sigma _3\tau _1\sigma _1\tau _3)^2`$
$`+(\sigma _2\tau _2)^2(\sigma _3\tau _2+\sigma _2\tau _3)3(\sigma _1+\tau _1)(\sigma _3+\tau _3)(\sigma _3\tau _2+\sigma _2\tau _3))`$
$`+\{\alpha +2\}\{\alpha +1\}\{\alpha \}\{\alpha 1\}\{\alpha 2\}\{\alpha 3\}(\sigma _3+\tau _3)^3.`$
We note that $`R_1^{(n,n)}=0`$ holds for all $`n`$.
It can be shown that $`R_\alpha ^{(m,n)}(x;y)`$ is a sum of products of determinants. Namely let $`\mathrm{\Lambda }(m,n)`$ denote the set of partitions $`\lambda =(\lambda _1,\mathrm{},\lambda _m)`$ satisfying $`n\lambda _1\mathrm{}\lambda _m0`$. For $`\lambda \mathrm{\Lambda }(m,n)`$ we write $`\stackrel{~}{\lambda ^{}}=(m\lambda _n^{},\mathrm{},m\lambda _1^{})`$, where $`\lambda ^{}=(\lambda _1^{},\mathrm{},\lambda _n^{})`$ denotes the conjugate partition. Then
$`R_\alpha ^{(m,n)}(x;y)`$ $`=`$ $`{\displaystyle \underset{i=1}{\overset{m}{}}}\{\alpha i+n\}{\displaystyle \underset{j=1}{\overset{n}{}}}\{\alpha j+m\}`$
$`\times {\displaystyle \underset{\lambda \mathrm{\Lambda }(m,n)}{}}S_\lambda (x;\alpha +n,\omega )S_{\stackrel{~}{\lambda ^{}}}(y;\alpha +m,\omega ),`$
where
$`S_\lambda (x;\alpha ,\omega )=det\left({\displaystyle \frac{\{\alpha \lambda _i^{}+i2j\}}{\{\alpha i\}}}\sigma _{\lambda _i^{}i+j}(x)\right)_{1i,jN},`$
with $`Nn`$.
## 6. Discussions
In this paper we have developed an algebraic approach to the ABF models in regime II, using free fields. We have found the following.
First, our results immediately give an integral representation for correlation functions of these non-critical models. We computed exactly the simplest integrals for the nearest neighbor correlation functions. The result supports the validity of our construction. Beyond this case, it remains a technical open problem to perform multiple contour integrals explicitly.
Second, following the prescription of we found states which diagonalize the row-to-row transfer matrix in the thermodynamic limit. These are obtained by acting with type II vertex operators on the vacuum. Compared with other lattice models studied so far, a nice feature is that these operators do not contain contour integrals. This gives us a hope for handling them more effectively in further analysis.
As an application we studied the continuous limit of the traces of vertex operators of type II. We obtained a family of functions satisfying Watson’s equations as well as kinematical pole conditions. The former is a consequence of the commutation relation (5.7) and the property of the corner Hamiltonian as the grading operator, while the latter follows from
$`i_{\overline{1}}^a(\beta _1)_1^a(\beta _2)={\displaystyle \frac{1}{\beta _1\beta _2i\pi }}\times \mathrm{id}+O(1)(\beta _1\beta _2+i\pi ).`$
On these grounds, we interpreted the resulting formulas as form-factors of some local operators in the $`_k`$ invariant massive scattering theory with minimal $`S`$-matrices. Our analysis is not complete, since we have considered only the $`_k`$ neutral sector, which corresponds to having the same number of $`\mathrm{\Psi }_+(v)`$’s and $`\mathrm{\Psi }_{}(v)`$’s in the trace. The bound state conditions also remain to be worked out.
There are other unsolved problems which deserve attention. Of particular interest is the problem of identifying the local operators corresponding to the projection operators. The form-factors $`\widehat{𝒬}_a^{(n,n)}`$ are proportional to $`M^{2\mathrm{\Delta }_{0,a1}}`$. This mass dependence indicates that the projection operators correspond to perturbations of the neutral primary fields in the parafermionic CFT with the respective conformal dimensions.<sup>2)</sup><sup>2)</sup>2)A similar study has been made in for the scaling limit of the XXZ model. An argument in favor of this proposal is that the cluster property characteristic to the ‘exponential operators’ seems to hold.
It is natural to identify the case $`a=1`$, $`\mathrm{\Delta }_{0,0}=0`$, with the form-factor of the identity operator. The fact that $`\widehat{𝒬}_1^{(n,n)}=0`$ for all $`n`$ agrees with this identification. In the case $`a=3`$, $`\mathrm{\Delta }_{0,2}=\frac{2}{k+2}`$, we infer that the $`\widehat{𝒬}_3^{(n,n)}`$ give the form factors of the first energy operator. Since this operator perturbs the parafermionic CFT into the massive region, the corresponding form factors are proportional to those of the trace of stress-energy tensor. One can show that, for $`a=3`$, the polynomial $`R_2^{(n,n)}(x;y)`$ in the formula (5.14) contains a factor
$`(\sigma _1+\tau _1)\left({\displaystyle \frac{\sigma _{n1}}{\sigma _n}}+{\displaystyle \frac{\tau _{n1}}{\tau _n}}\right),`$
as we expect for form factors of the stress-energy tensors .
However, our understanding of this problem of identification of local operators, as well as that of deriving the vacuum expectation values of local fields, is still incomplete.
In the scaling limit we considered only the translationally invariant sector $`m=0`$. From the lattice point of view, the other sectors have equal rights. It would be interesting to understand the field theoretical meaning of the traces corresponding to the $`m0`$ sectors.
Although practically all the technical elements of the construction have been known, we think that the interpretation as a free field realization of ABF models in regime II is rather interesting. It opens up a way for applying algebraic methods to the study of physical phenomena in these models.
Acknowledgments.
We thank Michael Lashkevich and Takeshi Oota for helpful discussions. M. J. is grateful to Professor Shi-Shyr Roan for kind invitation and warm hospitality during his visit to Academia Sinica, Taiwan. H. K. thanks V. A. Fateev and Al. B. Zamolodchikov for discussions and colleagues in LAPTH for kind hospitality during his stay in Annecy. This work is partially supported by the Grant-in-Aid for Scientific Research no.10440042 and no.11640030, the Ministry of Education.
## Appendix A Formulas for operators
We give here explicit formulas for the operators in the deformed parafermion theory.
Oscillators We consider two kinds of oscillators $`a_{j,n}`$ ($`j=1,2`$, $`n\backslash \{0\}`$) and ‘zero-mode’ operators $`P_j,Q_j`$ ($`j=1,2`$) satisfying
$`[a_{1,n},a_{1,n^{}}]={\displaystyle \frac{[2n]_x[(k+2)n]_x}{n}}\delta _{n+n^{},0},`$
$`[a_{2,n},a_{2,n^{}}]={\displaystyle \frac{[2n]_x[kn]_x}{n}}\delta _{n+n^{},0},`$
$`[P_1,Q_1]=2(k+2),[P_2,Q_2]=2k,`$
where
$`[n]_x={\displaystyle \frac{x^nx^n}{xx^1}}.`$
Fock space For $`l,m`$ we set
$`_{m,l}=[a_{1,n},a_{2,n}(n>0)]|m,l,`$
$`P_1|m,l=l|m,l,P_2|m,l=m|m,l,`$
$`e^{\frac{Q_1}{2(k+2)}}|m,l=|m,l+1,e^{\frac{Q_2}{2k}}|m,l=|m+1,l.`$
Grading operator We set $`D=D^{osc}+D^{zero}`$, with
$`D^{osc}={\displaystyle \underset{n>0}{}}{\displaystyle \frac{n^2}{[2n]_x[(k+2)n]_x}}a_{1,n}a_{1,n}{\displaystyle \underset{n>0}{}}{\displaystyle \frac{n^2}{[2n]_x[kn]_x}}a_{2,n}a_{2,n},`$
$`D^{zero}={\displaystyle \frac{(P_1+1)^21}{4(k+2)}}{\displaystyle \frac{P_2^2}{4k}}.`$
We have
$`[D,a_{j,n}]=na_{j,n},`$
$`[D,Q_1]=P_1+1,[D,Q_2]=P_2,`$
$`D|m,l=\left({\displaystyle \frac{(l+1)^21}{4(k+2)}}{\displaystyle \frac{m^2}{4k}}\right)|m,l.`$
Notational convention We use the following notation.
$`\varphi _j(A;B,C|z;C^{})={\displaystyle \frac{A}{BC}}\left(Q_j+P_j\mathrm{log}z\right)+{\displaystyle \underset{n0}{}}{\displaystyle \frac{[An]_x}{[Bn]_x[Cn]_x}}a_{j,n}z^nx^{C^{}|n|},`$
$`\varphi _j(B|z;C)=\varphi _j(A;A,B|z;C),`$
$`\varphi _j^{(\pm )}(A;B|z)={\displaystyle \frac{A}{B}}P_j\mathrm{log}x+(xx^1){\displaystyle \underset{n>0}{}}{\displaystyle \frac{[An]_x}{[Bn]_x}}a_{j,\pm n}z^n.`$
We use the ‘additive’ parameters $`u,u^{},\mathrm{}`$ and the ‘multiplicative’ ones $`z,z^{},\mathrm{}`$ on an equal footing. Unless otherwise stated explicitly, they are related by
$$z=x^{2u},z^{}=x^{2u^{}},\mathrm{}.$$
The following are the list of operators used in the text.
Type II VO’s
$`\mathrm{\Psi }_\pm (u)={\displaystyle \frac{1}{xx^1}}\left(\mathrm{\Psi }_{\pm ,+}(u)\mathrm{\Psi }_{\pm ,}(u)\right),`$
$`\mathrm{\Psi }_{\pm ,\epsilon }(u)=z^{\frac{1}{k}}\times `$
$`:\mathrm{exp}(\pm \varphi _2(k|z;\pm {\displaystyle \frac{k}{2}})\pm \epsilon \varphi _1^{(\epsilon )}(1;2|zx^{\pm \epsilon \frac{k}{2}})\epsilon \varphi _2^{(\epsilon )}(1;2|zx^{\pm \epsilon \frac{k+2}{2}})):.`$
More generally we set
(A.1) $`\mathrm{\Psi }_a(u)=z^{\frac{a(a1)}{k}}{\displaystyle \underset{j=0}{\overset{a}{}}}(1)^jx^{(a1)(j\frac{a}{2})}{\displaystyle \underset{i=1}{\overset{j}{}}}{\displaystyle \frac{[ai+1]_x}{[i]_x}}`$
$`\times :{\displaystyle \underset{i=1}{\overset{aj}{}}}\mathrm{\Psi }_,(\frac{a+1}{2}+i+u){\displaystyle \underset{i=aj+1}{\overset{a}{}}}\mathrm{\Psi }_{,+}(\frac{a+1}{2}+i+u):.`$
Screening current
$`S(u)={\displaystyle \frac{1}{xx^1}}\left(S_+(u)S_{}(u)\right),`$
$`S_\pm (u)=`$
$`:\mathrm{exp}(\varphi _1(k+2|z;{\displaystyle \frac{k+2}{2}})\pm \varphi _2^{(\pm )}(1;2|zx^{\frac{k+2}{2}})\pm \varphi _1^{(\pm )}(1;2|zx^{\frac{k}{2}})):.`$
Type I VO’s
$`\mathrm{\Phi }_{}(u)=z^{\frac{k1}{2k(k+2)}}\times `$
$`:\mathrm{exp}(\varphi _1(1;2,k+2|zx^k;{\displaystyle \frac{k+2}{2}})\varphi _2(1;2,k|zx^k;{\displaystyle \frac{k}{2}})):,`$
$`\mathrm{\Phi }_+(u)={\displaystyle _{C_\mathrm{\Phi }(z)}}\underset{¯}{dz}^{}\mathrm{\Phi }_{}(u)S(u^{}){\displaystyle \frac{[uu^{}\frac{3}{2}P_1]}{[uu^{}\frac{3}{2}]}},`$
$`\mathrm{\Phi }_{}^{}(u)={\displaystyle \frac{g}{[P_1+1]}}z^{\frac{k1}{2k(k+2)}}\times `$
$`:\mathrm{exp}(\varphi _1(1;2,k+2|z;{\displaystyle \frac{k+2}{2}})\varphi _2(1;2,k|z;{\displaystyle \frac{k}{2}})):,`$
$`\mathrm{\Phi }_+^{}(u)={\displaystyle _{C_\mathrm{\Phi }^{}(z)}}\underset{¯}{dz}^{}\mathrm{\Phi }_{}^{}(u)S(u^{}){\displaystyle \frac{[uu^{}\frac{1}{2}P_1]}{[uu^{}\frac{1}{2}]}}.`$
The contours $`C_\mathrm{\Phi }(z),C_\mathrm{\Phi }^{}(z)`$ are specified by the rules
$`C_\mathrm{\Phi }(z):z^{}`$ $`=`$ $`zx^{3+2(k+2)n}(n>0)\mathrm{inside},`$
$`=`$ $`zx^{1+2(k+2)n}(n0)\mathrm{outside},`$
$`C_\mathrm{\Phi }^{}(z):z^{}`$ $`=`$ $`zx^{1+2(k+2)n}(n0)\mathrm{inside},`$
$`=`$ $`zx^{1+2(k+2)n}(n0)\mathrm{outside},`$
and
$`g`$ $`=`$ $`(xx^1)x^{\frac{1}{k+2}}`$
$`\times {\displaystyle \frac{(x^{2k+2},x^{4k+2};x^{2k},x^{2k+4})_{\mathrm{}}}{(x^{2k},x^{4k+4};x^{2k},x^{2k+4})_{\mathrm{}}}}(x^{2k+2};x^{2k+4})_{\mathrm{}}^2(x^{2k+4};x^{2k+4})_{\mathrm{}}.`$
$`\xi `$-$`\eta `$ system
$`\xi (u)=:\mathrm{exp}(\varphi _1(2|z;{\displaystyle \frac{k}{2}})+\varphi _2(2|z;{\displaystyle \frac{k+2}{2}})):,`$
$`\eta (u)=:\mathrm{exp}(\varphi _1(2|z;{\displaystyle \frac{k}{2}})\varphi _2(2|z;{\displaystyle \frac{k+2}{2}})):.`$
Deformed $`W`$ currents The first DWA current $`W^1(u)`$ is
$`W^1(u)=[k+1]_x\mathrm{\Lambda }_{}(u)+[k+2]_x\mathrm{\Lambda }_0(u)+[k+1]_x\mathrm{\Lambda }_+(u),`$
where $`\mathrm{\Lambda }_\pm (u)`$ and $`\mathrm{\Lambda }_0(u)`$ are
$`\mathrm{\Lambda }_\pm (u)`$ $`=`$ $`x^{\pm 1}g^1x^{\frac{1k}{k}}z^{\frac{1k}{k(k+2)}}:\mathrm{\Phi }_{}(k+2+u)[P]\mathrm{\Phi }_{}^{}(u)S_{}(\frac{k+1}{2}\frac{k+2}{2}+u):,`$
$`\mathrm{\Lambda }_0(u)`$ $`=`$ $`g^1x^{\frac{1k}{k}}z^{\frac{1k}{k(k+2)}}:\mathrm{\Phi }_{}(k+2+u)[P]\mathrm{\Phi }_{}^{}(u)S_+(k+\frac{3}{2}+u):.`$
In general the DWA currents $`W^j(u)`$ ($`j=1,2,\mathrm{}`$) are given by
$`W^j(u)`$ $`=`$ $`{\displaystyle \underset{\genfrac{}{}{0pt}{}{a,b,c0}{a+b+c=j}}{}}C_{a,b,c}^j:{\displaystyle \underset{i=1}{\overset{a}{}}}\mathrm{\Lambda }_{}(\frac{j+1}{2}i+u)`$
$`\times {\displaystyle \underset{i=a+1}{\overset{a+b}{}}}\mathrm{\Lambda }_0(\frac{j+1}{2}i+u){\displaystyle \underset{i=a+b+1}{\overset{j}{}}}\mathrm{\Lambda }_+(\frac{j+1}{2}i+u):,`$
where $`C_{a,b,c}^j`$ is
$`C_{a,b,c}^j={\displaystyle \underset{i=1}{\overset{a}{}}}{\displaystyle \frac{[k+2i]_x}{[i]_x}}{\displaystyle \underset{i=1}{\overset{b}{}}}{\displaystyle \frac{[k+1+i]_x}{[i]_x}}{\displaystyle \underset{i=1}{\overset{c}{}}}{\displaystyle \frac{[k+2i]_x}{[i]_x}}.`$
Comparison of notation We have followed the notation in with minor changes. Referring to the symbols in by the subscript ‘$`K`$’, we have
$`x=q_K,`$
$`_{m,l}=_{l,m,K}^{PF},`$
$`S(u)=S_K(z),`$
$`\eta (u)=\eta _K(z),`$
$`\mathrm{\Phi }_{}(u)=\varphi _{1,1,K}(q^kz)z^{\frac{k1}{2k(k+2)}},`$
$`\mathrm{\Phi }_{}^{}(u)={\displaystyle \frac{g}{[P_1+1]}}\varphi _{1,1,K}(z)z^{\frac{k1}{2k(k+2)}},`$
$`\mathrm{\Psi }_+(u)=\mathrm{\Psi }_K^{}(z)z^{\frac{1}{k}},`$
$`\mathrm{\Psi }_{}(u)=\mathrm{\Psi }_K(z)z^{\frac{1}{k}}.`$
We have introduced fractional powers of $`z`$ in order that the homogeneity property
$`x^{2vD}Y(u)x^{2vD}=Y(u+v)`$
holds for $`Y=S,\mathrm{\Phi }_\pm ,\mathrm{\Phi }_\pm ^{},\mathrm{\Psi }_\pm ,\xi ,\mathrm{\Psi }_a,W^j`$. As for $`\eta `$ we have $`x^{2vD}\eta (u)x^{2vD}=\eta (u+v)x^{2v}`$.
## Appendix B Resolution by $`\xi `$-$`\eta `$ system
We summarize the main points concerning the resolution by the $`\xi `$-$`\eta `$ system.
On the space $`_{m,l}`$ with $`mlmod2`$, we have expansions of the form
$`\eta (u)={\displaystyle \underset{n}{}}\eta _nz^{n1},\xi (u)={\displaystyle \underset{n}{}}\xi _nz^n.`$
The Fourier components satisfy
$`[\eta _n,\eta _n^{}]_+=0,[\xi _n,\xi _n^{}]_+=0,[\eta _n,\xi _n^{}]_+=\delta _{n+n^{},0}.`$
The components $`\eta _0,\xi _0`$ commute with $`D`$. Since $`\xi _0^2=\eta _0^2=0`$ and $`\xi _0\eta _0+\eta _0\xi _0=\mathrm{id}`$, the complex
$`\mathrm{}\stackrel{\eta _0}{}_{mk,l(k+2)}\stackrel{\eta _0}{}_{m,l}\stackrel{\eta _0}{}_{m+k,l+k+2}\stackrel{\eta _0}{}_{m+2k,l+2(k+2)}\stackrel{\eta _0}{}\mathrm{}`$
is exact. Let
$`\stackrel{~}{}_{m,l}`$ $`=`$ $`\mathrm{Ker}\eta _0|_{_{m,l}}`$
$`=`$ $`\mathrm{Coker}\eta _0|_{_{mk,l(k+2)}}.`$
Then, for an operator $`𝒪`$ on $`_{mlmod2}_{m,l}`$ commuting with $`\eta _0`$, we have
$`\mathrm{tr}_{\stackrel{~}{}_{m,l}}(𝒪)`$ $`=`$ $`{\displaystyle \underset{n0}{}}(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)n}}(𝒪)`$
$`=`$ $`{\displaystyle \underset{n<0}{}}(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)n}}(𝒪).`$
## Appendix C Calculation of the trace
We outline the computation of the trace of a product of type II operators focusing attention to the neutral case
(C.1) $`𝒪=x^{2kD}\mathrm{\Psi }_+(v_1)\mathrm{}\mathrm{\Psi }_+(v_n)\mathrm{\Psi }_{}(v_1^{})\mathrm{}\mathrm{\Psi }_{}(v_n^{}).`$
It is convenient in what follows to consider the oscillator part and the zero mode part separately. Let us write
$`\mathrm{\Psi }_{\pm ,\epsilon }(v)=\mathrm{\Psi }_{\pm ,\epsilon }^{osc}(v)\mathrm{\Psi }_{\pm ,\epsilon }^{zero}(v),`$
$`\mathrm{\Psi }_{\pm ,\epsilon }^{zero}(v)=e^{Q_2/k}z^{P_2/k1/k}x^{\pm \epsilon P_1/2\epsilon P_2/2}.`$
Oscillators The contributions from the oscillator part
$`𝒪={\displaystyle \frac{\mathrm{tr}_{_{m,l}}\left(x^{2kD^{osc}}𝒪^{osc}\right)}{\mathrm{tr}_{_{m,l}}(x^{2kD^{osc}})}}`$
are given by the following rules:
(C.2) $`C^N\mathrm{\Psi }_{\epsilon _1,\epsilon _1^{}}(v_1)\mathrm{}\mathrm{\Psi }_{\epsilon _N,\epsilon _N^{}}(v_N)`$
$`={\displaystyle \underset{1i<jN}{}}C^2\mathrm{\Psi }_{\epsilon _i,\epsilon _i^{}}(v_i)\mathrm{\Psi }_{\epsilon _j,\epsilon _j^{}}(v_j)`$
for $`_{i=1}^N\epsilon _i=0`$, and
(C.3) $`\mathrm{\Psi }_{+,\epsilon _1}(v_1)\mathrm{\Psi }_{+,\epsilon _2}(v_2)=\mathrm{\Psi }_{,\epsilon _1}(v_1)\mathrm{\Psi }_{,\epsilon _2}(v_2)`$
$`=C^2F(v){\displaystyle \frac{[v+\frac{\epsilon _1\epsilon _2}{2}]^{}}{[v1]^{}}}x^{\frac{2}{k}(1+\frac{\epsilon _1\epsilon _2}{2})v+\frac{1+\epsilon _1\epsilon _2}{2k}+1+\frac{\epsilon _1\epsilon _2}{2}},`$
(C.4) $`\mathrm{\Psi }_{+,\epsilon _1}(v_1)\mathrm{\Psi }_{,\epsilon _2}(v_2)=\mathrm{\Psi }_{,\epsilon _1}(v_1)\mathrm{\Psi }_{+,\epsilon _2}(v_2)`$
$`=C^2F(v\frac{k}{2})^1{\displaystyle \frac{[v\frac{k}{2}\frac{\epsilon _1+\epsilon _2}{2}]^{}}{[v\frac{k}{2}]^{}}}x^{\frac{\epsilon _1+\epsilon _2}{k}v\frac{1+k}{2k}(1+\epsilon _1\epsilon _2)\frac{\epsilon _1+\epsilon _2}{2}},`$
where $`v=v_2v_1`$, $`F(v)`$ is given in (4.3), and
$`C=(x^{2k};x^{2k})_{\mathrm{}}{\displaystyle \frac{(x^{2+4k};x^{2k},x^{2k})_{\mathrm{}}}{(x^{2+2k};x^{2k},x^{2k})_{\mathrm{}}}}.`$
Zero mode The trace over the zero mode in the Fock space is taken in a standard fashion, following (3.9),(3.4). As we show at the end of this appendix, our operator (C.1) satisfies in addition
(C.5) $`\mathrm{tr}_{_{m,l}}(𝒪_0)=\mathrm{tr}_{_{m,l2}}(𝒪_1).`$
Then, for any fixed real numbers $`h,h^{}`$, the trace can be rewritten as follows.
$`\mathrm{tr}_{H^0(𝒞_{m,l})}(𝒪)`$
$`=`$ $`\left({\displaystyle \underset{sh}{}}{\displaystyle \underset{n0}{}}{\displaystyle \underset{s>h}{}}{\displaystyle \underset{n<0}{}}\right)(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)(n2s)}}(𝒪_0)`$
$`\left({\displaystyle \underset{s<h^{}}{}}{\displaystyle \underset{n0}{}}{\displaystyle \underset{sh^{}}{}}{\displaystyle \underset{n<0}{}}\right)(1)^n\mathrm{tr}_{_{m+kn,l2+(k+2)(n2s)}}(𝒪_1)`$
$`=`$ $`\left({\displaystyle \underset{\genfrac{}{}{0pt}{}{n0}{sh}}{}}{\displaystyle \underset{\genfrac{}{}{0pt}{}{n<0}{s>h}}{}}{\displaystyle \underset{\genfrac{}{}{0pt}{}{n0}{s>n+h^{}}}{}}+{\displaystyle \underset{\genfrac{}{}{0pt}{}{n<0}{sn+h^{}}}{}}\right)(1)^n\mathrm{tr}_{_{m+kn,l+(k+2)(n2s)}}(𝒪_0)`$
$`=`$ $`\left({\displaystyle \underset{n_1h^{}n_2n_1+h}{}}{\displaystyle \underset{n_1+h<n_2<n_1h^{}}{}}\right)(1)^{2n_1}\mathrm{tr}_{_{m+2n_2k,l+2n_1(k+2)}}(𝒪_0),`$
where the last sum is taken over $`n_1,n_2\frac{1}{2}`$ such that $`n_1n_2`$. Note that the result is independent of the choice of $`h,h^{}`$.
Accordingly, the sum of the zero-mode contributions coming from different Fock spaces can be expressed by the function
$`\mathrm{\Gamma }_{m,l}^{(h,h^{})}(y_1,y_2|\tau )(lmmod2)`$
$`=`$ $`{\displaystyle \frac{1}{\eta (\tau )^2}}\left({\displaystyle \underset{n_1h^{}n_2n_1+h}{}}{\displaystyle \underset{n_1+h<n_2<n_1h^{}}{}}\right)(1)^{2n_1}`$
$`\times e^{2\pi i\tau \left(\frac{(l+1+2n_1(k+2))^2}{4(k+2)}\frac{(m+2n_2k)^2}{4k}\right)}e^{\pi i(l+1+2n_1(k+2))y_1\pi i(m+2n_2k)y_2},`$
where the sum is taken over $`n_1,n_2\frac{1}{2}`$ with $`n_1n_2`$, and we set $`\mathrm{\Gamma }_{m,l}^{(h,h^{})}(y_1,y_2|\tau )=0`$ for $`lmmod2`$. Choosing $`h=h^{}=0`$, for example, we obtain the formula (4.2) for the $`Q_a(m)`$. We note the properties
(C.7) $`\mathrm{\Gamma }_{m+2k,l}^{(h,h^{})}(y_1,y_2|\tau )=\mathrm{\Gamma }_{m,l}^{(h+1,h^{}1)}(y_1,y_2|\tau ),`$
(C.8) $`\mathrm{\Gamma }_{m+k,kl}^{(h,h^{})}(y_1,y_2|\tau )=\mathrm{\Gamma }_{m,l}^{(h^{}+\epsilon ^{},h+\epsilon )}(y_1,y_2|\tau ),`$
where $`\epsilon =\{\genfrac{}{}{0pt}{}{0(h)}{1(h)}`$ and $`\epsilon ^{}=\{\genfrac{}{}{0pt}{}{1(h^{})}{0(h^{})}`$.
Modular transform To study the continuous limit, it is useful to consider the modular transformation $`\tau 1/\tau `$. Unfortunately the functions $`\mathrm{\Gamma }_{m,l}^{(h,h^{})}(y_1,y_2|\tau )`$ do not enjoy simple transformation properties individually. A way around this difficulty is proposed in . Recall that up to an overall scalar factor the $`Q_{l+1}^{(n,n)}(m)`$ has the form
(C.9) $`{\displaystyle \underset{\mu ,\nu }{}}f_{\mu ,\nu }^{(n)}\mathrm{\Gamma }_{m,l}^{(h,h^{})}(y_1,y_2|\tau ),`$
where $`f_{\mu ,\nu }^{(n)}`$ is a function of $`\mu ,\nu `$ given below (C), and $`y_1=\frac{\tau }{2k}(\mu \nu )`$, $`y_2=\frac{\tau }{k}(\frac{2v}{k}\frac{\mu +\nu }{2})`$ with $`v=_{i=1}^n(\frac{k}{2}v_iu_i)`$. Making use of the independence of (C.9) on $`h,h^{}`$, it is shown in that a certain generating function of (C.9) with respect to $`m,h,h^{}`$ can be reexpressed in terms of theta functions and (the zero mode contribution of) the characters of the $`N=2`$ superconformal algebra. From the knowledge of the modular property for the latter, the following relation can be deduced for the sum (C.9).
$`{\displaystyle \underset{\mu ,\nu }{}}f_{\mu ,\nu }^{(n)}\mathrm{\Gamma }_{m,l}^{(h,h^{})}(y_1,y_2|\tau )`$
$`={\displaystyle \frac{1}{\sqrt{k(k+2)}}}{\displaystyle \underset{\mu ,\nu }{}}f_{\mu ,\nu }^{(n)}e^{\frac{i\pi }{2\tau }\left((k+2)y_1^2ky_2^2\right)}`$
$`\times {\displaystyle \underset{l^{}=0}{\overset{k}{}}}{\displaystyle \underset{m^{}=k}{\overset{k1}{}}}\mathrm{sin}{\displaystyle \frac{\pi (l+1)(l^{}+1)}{k+2}}e^{\frac{i\pi mm^{}}{k}}\mathrm{\Gamma }_{m^{},l^{}}^{(h,h^{})}({\displaystyle \frac{y_1}{\tau }},{\displaystyle \frac{y_2}{\tau }}|{\displaystyle \frac{1}{\tau }}).`$
By noting (C.8) and the property $`f_{\nu ,\mu }^{(n)}=f_{\mu ,\nu }^{(n)}`$ (see below), the sums in the last line can also be written as
$`2{\displaystyle \underset{l^{}=0}{\overset{k}{}}}{\displaystyle \underset{m^{}=0}{\overset{k1}{}}}\mathrm{sin}{\displaystyle \frac{\pi (l+1)(l^{}+1)}{k+2}}e^{\frac{i\pi mm^{}}{k}}\mathrm{\Gamma }_{m^{},l^{}}^{(h,h^{})}({\displaystyle \frac{y_1}{\tau }},{\displaystyle \frac{y_2}{\tau }}|{\displaystyle \frac{1}{\tau }}).`$
An identity Let us verify the property (C.5). For $`\mu ,\nu `$, set
$`f_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n)`$
$`=`$ $`{\displaystyle \underset{i=1}{\overset{n}{}}\mu _i\nu _i\underset{1i,jn}{}[u_i+v_j+\frac{\mu _i+\nu _j}{2}]^{}}`$
$`\times {\displaystyle \underset{1i<jn}{}}{\displaystyle \frac{[u_iu_j\frac{\mu _i\mu _j}{2}]^{}}{[u_iu_j]^{}}}{\displaystyle \frac{[v_iv_j\frac{\nu _i\nu _j}{2}]^{}}{[v_iv_j]^{}}}.`$
Here the sum is taken over $`\mu _i,\nu _i=\pm 1`$ ($`i=1,\mathrm{},n`$) satisfying $`_{1=1}^n\mu _i=\mu `$, $`_{i=1}^n\nu _i=\nu `$. It is easy to see that this function is holomorphic and symmetric in $`(u_1,\mathrm{},u_n)`$ (resp. $`(v_1,\mathrm{},v_n)`$). The relation (C.5) reduces to the identity
(C.11) $`f_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n)=f_{\nu ,\mu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n).`$
Clearly (C.11) is true for $`n=1`$. To show it in general, let $`g_{\mu ,\nu }^{(n)}`$ stand for the difference of the left hand side and the right hand side of (C.11). Then
$`g_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n)=g_{\mu ,\nu }^{(n)}(v_1,\mathrm{},v_n;u_1,\mathrm{},u_n),`$
$`g_{\mu ,\nu }^{(n)}(u_1+k,\mathrm{},u_n;v_1,\mathrm{},v_n)=(1)^ng_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n),`$
$`g_{\mu ,\nu }^{(n)}(u_1+\frac{k}{\tau },\mathrm{},u_n;v_1,\mathrm{},v_n)`$
$`=\left(e^{\frac{i\pi }{\tau }}\right)^ne^{\frac{2\pi i}{k}(nu_1+v_1+\mathrm{}+v_n+\frac{\mu +\nu }{2})}g_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_n;v_1,\mathrm{},v_n),`$
$`g_{\mu ,\nu }^{(n)}(u_1,\mathrm{},u_{n1},u;v_1,\mathrm{},v_{n1},u)`$
$`={\displaystyle \underset{\mu _n=\pm 1}{}}[\mu _n]^{}{\displaystyle \underset{i=1}{\overset{n1}{}}}[u_iu+\mu _n]^{}[v_i+u+\mu _n]^{}`$
$`\times g_{\mu \mu _n,\nu \mu _n}^{(n1)}(u_1,\mathrm{},u_{n1};v_1,\mathrm{},v_{n1}).`$
From these properties, we conclude that $`g_{\mu ,\nu }^{(n)}=0`$ by induction on $`n`$. |
warning/0001/cond-mat0001084.html | ar5iv | text | # THERMODYNAMIC AND SPECTROSCOPIC PROPERTIES AND LOW TEMPERATURE THERMOCHROMISM OF CHROMIUM TRIS–ACETYLACETONATE
## 1 Introduction
The chromium tris–acetylacetonate $`Cr(C_5H_7O_2)_3`$ (or $`Cr(AA)_3`$) belongs to the $`\beta `$–diketonate complexes of transition metals crystallizing in the molecular type lattices. The $`\beta `$–diketonates of metals $`Me(AA)_3`$ are widely used in practice (to separate metals, to apply metal coatings, as catalysts etc.). The field of their application is extending with time, which encourages their further comprehensive investigations. The compounds $`Me(AA)_3`$ were actively studied in thermodynamical \[1–3\] and crystallochemical aspects. In the recent years they have became the subject of theoretical inquiry .
When investigating the properties of $`\beta `$–diketonates we have found some interesting effects in these compounds, before unknown. In this work two of such effects have been detected in the properties of $`Cr(AA)_3`$: 1) anomalies in low temperature heat capacity; 2) a reversible change in color of crystals when cooling them to liquid nitrogen temperature (thermochromism).
## 2 Experimental
The samples $`Cr(C_5H_7O_2)_3`$ synthesized for this investigation are crystalline powder with the average size of crystallites 0.3–0.5 mm. At the room temperature they are dark violet. The samples were defined by the methods of chemical analysis, IR–spectroscopy, derivatography and X–ray phase analysis.
The heat capacity of $`Cr(AA)_3`$ has been measured by the adiabatic method within the range $`5320K`$ using the installation described in Refs . The anomalous component was extracted by subtracting the regular heat capacity obtained by means of technique described in . Three anomaly were found in the heat capacity: the small peak at $`30K`$, the anomaly with maximum at $`60K`$, and the broad anomaly within the range $`110240K`$ (see Fig. 1).
The anomalies in maxima are correspondingly 0.9 %, 3.2 % and 0.3 % of a regular heat capacity. The corresponding contributions to the entropy are: $`\mathrm{\Delta }S0.07J`$ mol$`{}_{}{}^{1}K{}_{}{}^{}1`$ ($`30K`$), $`\mathrm{\Delta }S=1.20\pm 0.05J`$ mol$`{}_{}{}^{1}K_{}^{1}`$ ($`60K`$) and $`\mathrm{\Delta }S=1.5\pm 0.25J`$ mol$`{}_{}{}^{1}K_{}^{1}`$ (the broad anomaly).
To understand the nature of found anomalies we have measured the static magnetic susceptibility, the Raman spectra, the ESR–spectra on the ion $`Cr^{3+}`$, the transmission spectra in visible region and have followed the change in color with temperature.
The static magnetic susceptibility of $`Cr(AA)_3`$ was measured by MPMS–5s SQUID–magnitometer (of Quantum Design) within the temperature range $`2300K`$ (Fig. 2). The obtained data showed the experimental points to fit in well with the Curie–Weiss low. No pronounced magnetic anomalies were observed. But at the attentive consideration one could see a weak deviation of experimental points from the Curie–Weiss low within the temperature range $`140190K`$.
The ESR–spectra on the ion $`Cr^{3+}`$ in $`Cr(AA)_3`$ single crystals were recorded using $`E109`$ Varian spectrometer at the frequency $`9.5`$ GHz and at the temperatures $`300K`$ and $`78K`$ (Fig. 3). The ground orbital state of $`Cr^{3+}`$ ion in the octahedral surrounding of $`Cr(AA)_3`$ is singlet; $`g`$—factor of $`3d^3`$ configuration of ion $`Cr^{3+}`$ is close to $`g=2`$ . The large anisotropy of ESR–spectra has been observed which results from large shifts of spectra due to the large second order corrections to the parameters of fine structure when the initial split is comparable with Zeeman interaction. The decrease of temperature from $`300K`$ to $`78K`$ results in the broadening of lines without any change of their positions, for some lines such a broadening results in their total disappearance.
The transmission spectra of $`Cr(AA)_3`$ within the frequency range $`1100030000`$ cm<sup>-1</sup> have been obtained at room and nitrogen temperatures using the two–ray spectrophotometer Specord of $`UV200`$ type ( Fig. 4). The sample was prepared of powder placed between two quartz plates and then heated up to $`Cr(AA)_3`$ melting temperature ($`489K`$). After cooling the solid continuous layer of $`0.01`$ mm thick was formed.
Comparing the spectra obtained in the green region ($`18100`$ cm<sup>-1</sup>) at room and nitrogen temperatures one can see the shift of the transmission band maximum to the violet side by $`600`$ cm<sup>-1</sup>. This shift exceeds the expected shift resulting from the temperature expansion of crystal lattice.
The Raman spectra of $`Cr(AA)_3`$ (Fig. 5) were recorded with a Triplemate, SPEX spectrometer equipped with a O–SMA, $`Si`$–diode array. The $`633`$ nm, $`50`$ mW line of an $`HeNe`$–laser was used for the spectral excitation. For the low temperature measurements the sample was fixed on a cold finger of the helium cryostat (APD Cryogenic Inc). The measurements were carried out within the temperature range $`5220K`$. The temperature was established within the accuracy of $`0.1K`$.
The frequency interval $`40100`$ cm<sup>-1</sup> is the range of crystal lattice vibrations. The interval $`100150`$ cm<sup>-1</sup> is the borderline range of the lattice and molecular vibrations. The higher frequencies are rated as molecular ones. The band within the range $`40100`$ cm<sup>-1</sup> being wide at $`220K`$ splits up into the separate components when the temperature decreases. The detail analysis showed that the frequencies of these components depend on the temperature, but their relative intensity does’nt depend on it. When the temperature reaches $`60K`$ the new band at $`109`$ cm<sup>-1</sup> appears which was absent at higher temperature.
It was discovered in this work that the sample dark–violet at the room temperature becomes red when immersed into liquid nitrogen (thermochromism). This change in color is reversible. In order to follow the change in color with temperature, a special experiment was carried out. The sample placed on a copper substrate was immersed in the bath with liquid nitrogen. Temperature of the substrate and sample was measured by copper–constantan thermocouple, the color of the sample was estimated by eye. When the temperature slowly increased, the red color had remained within the range $`78120K`$. The color of the sample varied smoothly within the range $`120210K`$, at $`210K`$ it had turned dark violet and then it stayed as such at the further increase of temperature up to room one.
An additional experiment with 36 compounds showed that the reversible effects of thermochromism was observed for most of $`\beta `$–diketonates of transition metals, when cooling them with the liquid nitrogen. Most of tried 36 compounds grew light when the temperature decreased. The most pronounced changes in color were observed for cobalt (III), copper (II) and chromium (III) compounds. The compound $`Co(AA)_3`$, being dark green at room temperature, became violet. The compound $`Cu(AA)(hfa)`$–copper(II) acetylacetonate – hexafluoroacetylacetonate, being green at room temperature, became blue. The change in color of $`Cr(AA)_3`$ has been described above.
## 3 Discussion
The nature of discovered anomalies in heat capacity is still unknown.
For the moment we have no additional information about the anomaly at $`30K`$.
The $`\lambda `$–like anomaly at $`60K`$ might display some long range ordering in the substance. Its temperature is close to temperature $`T_s`$, where the total entropy of $`Cr(AA)_3`$ crystal has the maximal temperature derivative. The collective long range ordering in solids is often having been detected at that temperature $`T_s`$, where the derivative of the total entropy in temperature is maximal (see, for example, superconducting effect in HTSC or the ordering of another nature in Refs. ).
The magnetic susceptibility data show that the anomalies in the heat capacity do not result from any magnetic phase transitions. Then, one can suppose the anomaly to result from the change in the structure of the crystal or of the molecule. But in Raman spectra no significant change indicative of structural transformation in the crystal (of phase transition) was observed. That’s why the crystal is’nt thought to be arranged as a whole. In this case one may expect a rearrangement or an ordering of the loosely bound atoms or separate structural fragments: molecules $`Cr(AA)_3`$, ligands $`C_5H_7O_2`$ or groups $`CH`$ and $`CH_3`$.
The analysis of the internal vibrations of the molecule in Raman spectra has shown no change in structure of molecule with temperature. The absence of magnetic anomalies testifies that the symmetry of ion $`Cr^{3+}`$ crystal surrounding does not change substantially within the temperature range under investigation. This means that the change in geometry of molecule is hardly probable in the vicinity of chromium ion. One should rather expect some change connected with the dynamics of the methyl groups.
The torsional vibrations of methyl groups were identified in Raman spectra of $`Al`$–and $`Ga`$$`\beta `$–diketonates . They were detected below $`77K`$ in the frequency region $`100140`$ cm<sup>-1</sup>. One may suppose that just as in Ref. the new band $`109`$ cm<sup>-1</sup> appeared in Raman spectra of $`Cr(AA)_3`$ below $`60K`$ (Fig.5) results from the scattering on the torsional vibrations of the methyl groups. Only these vibrations can show an observed temperature behavior of the line intensity in Raman spectra.
According to Bose–Einstein distribution the function $`\varphi (z)`$ entering in the general expression for the heat capacity and characterizing the statistical weight of vibrational modes is
$$\varphi (z)=\frac{z^2e^z}{\left(1+e^z\right)^2},z=\frac{\mathrm{}\omega }{kT}.$$
The maximal change of statistical weight of vibrational mode $`\mathrm{}\omega `$ occurs at the temperature $`T=\mathrm{}\omega /3k`$. Then the maximal change of the statistical weight of the torsion vibrations $`109`$ cm<sup>-1</sup> ($`157K`$) falls at $`52K`$. It is interesting to note that this value is close to the temperature of the anomaly in heat capacity ($`60K`$).
This $`\lambda `$–like anomaly (Fig.1) might display some long range ordering all over the crystal below $`60K`$. The torsional vibrations of the methyl groups observed in Raman spectra at low temperatures arise at the same temperature $`60K`$. The assumed long range ordering might be attributed just to the dynamics of the methyl groups.
Within the temperature range 110–240 K the diffuse anomaly in heat capacity takes place. Within the same temperature range (120–210 K) the smooth change in the sample color has been detected. When temperature decreases from $`300K`$ to $`78K`$ the ESR spectra exhibit the broadening of lines without any change of their energetic position (Fig.3). Such a broadening might be explained by the appearance of additional indirect anisotropic interaction which results from the decrease of the mobility of separate fragments of $`Cr(AA)_3`$ molecule with temperature. This broadening is observed within the same temperature range where the diffuse anomaly in heat capacity and the smooth change in the sample color take place. The magnetic susceptibility shows the weak deviation of experimental points from the Curie–Weiss low within the same temperature range $`140190K`$.
One can suppose that ions $`Cr^{3+}`$ responsible for the paramagnetic behavior of $`Cr(AA)_3`$ crystals might be slightly affected by the change of some fragments of $`Cr(AA)_3`$ molecule with temperature. The change of energetic state of these fragments might result in the small perturbation of the crystal field. The last results in redistribution of electron density, shift of the transmission spectra and change in color of crystals when temperature varies from liquid nitrogen to room one.
## 4 Conclusion
In the present study some effects have been discovered in the low temperature properties of chromium tris–acetylacetonate $`Cr(AA)_3`$, which have not been observed before: the effect of thermochromism and three anomalies in heat capacity.
Anomalies in heat capacity generally display some transformations in a substance.
There are no considerations as regards the anomaly at $`30K`$.
As for the $`60K`$–anomaly, the assumption has been put forward that it might display some long range ordering taken place in the $`CH_3`$ – subsystem at this temperature. The rise of torsional vibrations of $`CH_3`$ – methyl groups at the same temperature was the basis for this assumption.
Besides, it was detected that the wide diffuse anomaly in the temperature range $`110240K`$ was accompanied with some features in ESR– and transmission spectra, with weak deviation of the magnetic susceptibility from the Curie–Weiss low, and change in color of the sample in the same temperature range. All these features were assumed to be attributed to a change with temperature of some fragments of the $`Cr(AA)_3`$ molecule which are still to be recognized.
To verify our tentative insight into the nature of discovered effects the complex investigation of $`\beta `$–diketonates of other transition metals with different ligands should be carried out.
## Acknowledgements
The authors thank A.A.Rastorguev for helpful discussions of essential aspects of this work.
This work is supported by RFBR (grant No 99–03–33370). |
warning/0001/hep-ph0001216.html | ar5iv | text | # Sphaleron rate in the symmetric electroweak phase
## I Introduction
Baryon number is violated in the standard model , and while the rate of the violation is negligibly small in vacuum, it can be significant at higher temperatures .
Recently there has been substantial interest in how fast baryon number is violated in the standard model at high temperatures. This is both because the rate is important for baryogenesis, and because it is an interesting problem in the more general framework of understanding dynamics of hot gauge theories.
The last 3 years have seen major progress in this problem. First, analytic work has clarified what physics is relevant. While it has long been known that infrared or “soft” gauge fields with momenta $`pg^2T`$ are responsible for baryon number violation , the recent work of Arnold, Son, and Yaffe has shown that “hard” modes with momenta $`pT`$ play an essential role in modifying the dynamics of the $`pg^2T`$ degrees of freedom . Bödeker has gone further, showing the role of scatterings between such hard modes by exchange of $`g^2T\text{ }\stackrel{<}{}\text{ }p\text{ }\stackrel{<}{}\text{ }gT`$ modes . There has also been progress in numerically studying the effective infrared (IR) theories to determine the size of baryon number violation .
The rate of baryon number violation is characterized by the sphaleron rate<sup>*</sup><sup>*</sup>*Our normalization is $`D_i=_i+iT^aA_i^a`$, and $`F_{ij}=[D_i,D_j]`$. $`\alpha `$ means $`\alpha _w`$, and $`g`$ means $`g_w`$.
$`\mathrm{\Gamma }`$ $``$ $`\underset{V,t\mathrm{}}{lim}{\displaystyle \frac{(N_{\mathrm{CS}}(t)N_{\mathrm{CS}}(0))^2}{Vt}},`$ (1)
$`N_{\mathrm{CS}}(t)N_{\mathrm{CS}}(0)`$ $`=`$ $`{\displaystyle \frac{1}{16\pi ^2}}{\displaystyle _0^t}{\displaystyle d^3xϵ_{ijk}E_i^aF_{jk}^a}.`$ (2)
Bödeker has shown that in pure Yang-Mills theory, $`\mathrm{\Gamma }`$ has the parametric form
$$\mathrm{\Gamma }=\kappa ^{}\left(\mathrm{log}\frac{m_\mathrm{D}}{g^2T}+C+O(1/\mathrm{log})\right)\left(\frac{g^2T^2}{m_\mathrm{D}^2}\right)\alpha ^5T^4.$$
(3)
The constantActually it is not a constant, but contains a $`\mathrm{log}(\mathrm{log}(m_D/g^2T))`$, see Eq. (44). $`C`$ can and recently has been determined analytically . The leading coefficient $`\kappa ^{}`$ can be obtained numerically by studying a local, UV finite theory, which is precisely Langevin dynamics for classical 3 dimensional nonabelian gauge fields:
$`D_\tau A_i^a(x,\tau )`$ $`=`$ $`g^2{\displaystyle \frac{H_A(A(\tau ))}{A_i^a(x,\tau )}}+\xi _i^a(x,\tau ),`$ (4)
$`H_A`$ $`=`$ $`{\displaystyle d^3x\frac{1}{4g^2}F_{ij}^aF_{ij}^a(x)},g^2{\displaystyle \frac{H_A}{A_i^a}}=(D_jF_{ji})^a=(D\times B)_i^a,`$ (5)
$`\xi _i^a(x_1,\tau _1)\xi _j^b(x_2,\tau _2)`$ $`=`$ $`2g^2T\delta (x_1x_2)\delta (\tau _1\tau _2)\delta _{ij}\delta ^{ab},`$ (6)
Here $`D_\tau `$ is the covariant derivative in Langevin time $`\tau `$, $`F_{ij}^a`$ the (nonabelian) magnetic field, and $`\xi `$ Gaussian white noise normalized as shown.Note that the Langevin time $`\tau `$ has dimensions of length squared, and recall our scaling convention for $`A`$..
The form of this theory is identical to the nonabelian Ampere’s law, but with the current replaced with $`\stackrel{}{j}=\sigma \stackrel{}{E}`$, as we would expect in a conducting medium;
$$D\times B=\sigma E(+\mathrm{noise}),$$
(7)
with the Langevin time identified as $`\tau =\sigma t`$. Here $`\sigma `$ is a “color conductivity” which describes the current response to very infrared external fields. It can be treated as a constant, to (next to) leading order in $`\mathrm{log}(1/g)`$, because the mean free path for color changing collisions is shorter than the $`1/g^2T`$ scale where $`\mathrm{\Gamma }`$ is set, by a factor of $`\mathrm{log}(1/g)`$. An explicit expression for $`\sigma `$ at next to leading log order, found by Arnold and Yaffe, , will be presented in Eq. (44). $`\kappa ^{}`$ is determined in this theory as
$$\kappa ^{}=\frac{3\sigma }{2\pi }\underset{V,\tau \mathrm{}}{lim}\frac{((N_{\mathrm{CS}}(\tau )N_{\mathrm{CS}}(0))^2}{V\tau }.$$
(8)
This effective theory was studied numerically in , with the result (see below) that $`\kappa ^{}10`$. This effective theory is not sufficient to determine the $`O(1/\mathrm{log})`$ coefficient in Eq. (3), and neither will anything we discuss.
Most of what we have described, and in particular everything in references , is for the case of pure Yang-Mills theory. This is appropriate if we are interested in temperatures very much higher than the electroweak phase transition (or crossover) temperature, because in that case the Higgs boson has a large thermal mass and can be removed perturbatively. However, the main application we are interested in, baryogenesis, requires knowing $`\mathrm{\Gamma }`$ in a theory with at least one Higgs boson, in the symmetric phase but at or slightly below the equilibrium temperature $`T_{\mathrm{eq}}`$ for the electroweak phase transition. It is not clear whether discarding the Higgs physics is justified in this case, and we should rethink both the appropriate effective theory, and the numerical determination of $`\mathrm{\Gamma }`$, in this light.
In this paper we will attempt to fill this gap. First we discuss how important we expect Higgs physics to be by considering the thermodynamics of Yang-Mills Higgs fields in Section II. Then we construct an appropriate infrared (IR) effective theory which generalizes Bödeker’s effective theory in Section III. We discuss the numerical implementation in Section IV and present numerical results in V. We find that, where the phase transition is strong enough to preserve baryon number after its completion, the change in $`\mathrm{\Gamma }`$ due to Higgs physics is quite a small effect, so the error in using pure Yang-Mills theory is small, of order 20%. While we work in the minimal standard model (at experimentally excluded values of the Higgs mass, to get a strong enough phase transition), we expect the results to hold as well in extensions such as the MSSM with a light scalar top quark. This conjecture could be tested by simulations in that theory, along the lines of what we do here.
## II Thermodynamic influence of the Higgs
To get an idea of how important Higgs physics will be for the sphaleron rate, we will try to get an idea of how important it is thermodynamically for the infrared transverse gauge boson excitations which we expect to be responsible for baryon number violation. As we will see, this in fact gives a reasonable estimate for what difference the Higgs will make in the sphaleron rate.
To a very good approximation the thermodynamics of infrared bosonic fields in the hot electroweak theory can be described by a three dimensional path integral . In fact this can be understood as a special case of the statement that the IR physics is essentially classical, since the three dimensional path integral we arrive at coincides with the partition function of the classical bosonic theory. Up to parametrically suppressed corrections<sup>§</sup><sup>§</sup>§The 3-D theory should be viewed as an IR effective theory for the thermodynamics below the scale $`T`$. At some level of precision it becomes necessary to include high dimension operators. If we are interested in thermodynamics at the length scale $`1/g^2T`$, then neglecting the high dimension operators causes errors of $`O((g^2T/T)^2)`$ times an additional explicit factor of $`\alpha `$ because the high dimension operators are radiatively induced, leading to an $`\alpha ^3`$ error. Of course, on less infrared scales the effective theory is less accurate., the partition function governing the thermodynamics is
$`Z`$ $`=`$ $`{\displaystyle 𝒟A_i𝒟A_0𝒟\mathrm{\Phi }\mathrm{exp}}H/T,`$ (9)
$`H`$ $`=`$ $`{\displaystyle d^3x\frac{1}{4g^2}F_{ij}^aF_{ij}^a}+{\displaystyle \frac{1}{2g^2}}(D_iA_0)^a(D_iA_0)^a+{\displaystyle \frac{m_\mathrm{D}^2(T)}{2g^2}}A_0^aA_0^a+{\displaystyle \frac{\lambda _A}{4g^4}}(A_0^aA_0^a)^2`$ (11)
$`+(D_i\mathrm{\Phi })^{}(D_i\mathrm{\Phi })+m_H^2(T)\mathrm{\Phi }^{}\mathrm{\Phi }+\lambda (\mathrm{\Phi }^{}\mathrm{\Phi })^2+{\displaystyle \frac{\lambda _{A\mathrm{\Phi }}}{2g^2}}A_0^aA_0^a\mathrm{\Phi }^{}\mathrm{\Phi }.`$
The couplings $`\lambda `$, $`g^2`$, $`\lambda _A\alpha ^2`$, and $`\lambda _{A\mathrm{\Phi }}=g^2/2+O(\alpha ^2)`$ are determined by a matching calculation; $`g^2`$, $`\lambda `$, and the field wave function normalizations correspond to those of the 4-D theory at a renormalization point given roughly by $`T`$. We will always treat $`\lambda g^2`$ for power counting purposes, as is appropriate given the renormalization structure of the theory. It is sometimes useful to consider $`\lambda g^2`$ and to expand in $`\lambda /g^2`$, but if we do so it is implied that $`\lambda `$ is still $`g^4`$, for instance. $`A_0`$ is the remnant of the temporal connection and can be thought of as an adjoint scalar field. Its mass term is the Debye mass responsible for charge screening, and is given by
$$m_\mathrm{D}^2(T)=\frac{g^2T^2}{12}(4N_\mathrm{c}+N_f+2N_s)\mathrm{\hspace{0.17em}3}\mathrm{D}\mathrm{counterterm}+O(g^4T^2),$$
(12)
with $`N_\mathrm{c}=2`$ the second Casimir of the group, $`N_f`$ the number of chiral, fundamental representation fermions, and $`N_s`$ the number of fundamental representation complex scalars. In the minimal standard model, the $`O(g^2)`$ piece is $`m_\mathrm{D}^2=(11/6)g^2T^2`$.
Because of the Debye mass, the $`A_0`$ field is heavy and has little influence on the very infrared thermodynamics, leading only to a rescaling of the effective gauge coupling by an $`O(g)`$ correction . However the same is not generically true for the Higgs field. This is because, unlike the $`A_0`$ field, it has a negative vacuum mass squared which may approximately cancel the positive induced thermal mass;
$`m_H^2(T)=m_{\mathrm{th}}^2+m_{\mathrm{vac}}^2,m_{\mathrm{vac}}^2<0,m_{\mathrm{th}}^2g^2T^2,`$ (13)
$`\left(\mathrm{specifically},m_{\mathrm{th}}^2={\displaystyle \frac{(3g^2+g_{}^{}{}_{}{}^{2}+4y_t^2+8\lambda )T^2}{16}}\mathrm{in}\mathrm{the}\mathrm{MSM}\right).`$ (14)
At very large temperatures, $`m_H^2(T)g^2T^2`$ and its thermodynamic effects are parametrically suppressed. However, the electroweak phase transition occurs where the positive thermal mass squared and the negative tree one cancel up to $`O(g^4T^2)`$ corrections, and it is in this regime that we need to know the sphaleron rate. It makes sense, then, to treat $`m_H^2(T)`$ parametrically as $`O(g^4T^2)`$, though depending on the strength of the phase transition, it may either be “large” or “small” within this parametric range.
Let us now consider the thermodynamic influence of the Higgs boson on the Yang-Mills fields. Since we are only making estimates here, we will be content with one loop calculations even where the loop expansion is not very reliable. We can estimate the Higgs field’s importance by considering Higgs contributions to the gauge field self-energy at soft external momentum $`pg^2T`$. The sum of diagram (a) in Fig. 1 and any $`p`$ independent contributions from (b) vanish in any gauge invariance respecting regularization; the scalar contribution to the gauge field self-energy is then
$$\mathrm{\Pi }_{ij}^{ab}(p)=\frac{g^2T\delta ^{ab}}{2}\frac{d^3k}{(2\pi )^3}\left(\frac{2k_i2k_j}{((k\frac{p}{2})^2+m_H^2)((k+\frac{p}{2})^2+m_H^2)}\frac{2k_i2k_j}{(k^2+m_H^2)^2}\right).$$
(15)
If we take $`pm_H`$, the result is
$$\mathrm{\Pi }_{ij}^{ab}(p)=\frac{g^2T\delta ^{ab}}{48\pi m_H}\left(\delta _{ij}p^2p_ip_j\right),$$
(16)
whereas in the regime where $`m_Hp`$, we get
$$\mathrm{\Pi }_{ij}^{ab}(p)=\frac{g^2T\delta ^{ab}}{32|p|}\left(\delta _{ij}p^2p_ip_j\right).$$
(17)
In either case the sign of the effect is such that it reduces the size of infrared gauge field excitations; so the presence of a Higgs field should reduce $`\mathrm{\Gamma }`$.
The calculation is for the symmetric phase, where there is no Higgs condensate. When there is a condensate $`\varphi _0`$ (normalized as $`\varphi _0^2=2\mathrm{\Phi }^{}\mathrm{\Phi }`$) then there is an induced mass for the gauge bosons of $`m_W^2=g^2\varphi _0^2/4`$, which will of course further reduce $`\mathrm{\Gamma }`$. However, we are primarily interested right now in the symmetric phase case.
To decide how important the scalar contribution is, we need to know what momentum $`p`$ is characteristic for the baryon number violating processes we intend to study. Since the physics we are after is nonperturbative physics, we expect the answer to be, $`p`$ such that perturbation theory is breaking down. We can estimate what $`p`$ is necessary by computing the contribution of gauge bosons to the self-energy. The result is not gauge invariant, but in Landau gauge the analogous contribution from gauge and ghost loops is In Feynman gauge the $`11`$ becomes $`14`$. The answer differs from that in , for instance, because we have left out the (heavy) $`A_0`$ field, treated as light there.
$$\mathrm{\Pi }_{ij}^{ab}(p)\mathrm{from}\mathrm{bosons}=\frac{11N_\mathrm{c}g^2T\delta ^{ab}}{64|p|}\left(\delta _{ij}p^2p_ip_j\right).$$
(18)
The sign is the opposite, indicating that gauge self-interactions lead to a more rapid onset of nonperturbatively large fields. The loop contribution comes on order the tree inverse propagator $`p^2\delta _{ij}p_ip_j`$, and the calculation therefore breaks down completely, for $`pg^2T/3`$. We take this as a fair estimate of the nonperturbative scale, though such an estimate certainly cannot be considered accurate to better than about a factor of 2.
For a heavy Higgs boson, then, we expect a correction of order $`(g^2T/m_H(T))/(48\pi )`$; but even for $`m_H(T)0`$ we only expect a correction from the Higgs boson to the propagator at the relevant momentum to be of order $`1/11`$. This corresponds to a rescaling by $`(1/11)`$ of the length scale where perturbation theory breaks down. However, since the fifth power of this length scale enters the sphaleron rate, the importance of the Higgs boson would be a little larger, as large as a $`(10/11)^50.6`$ reduction of $`\mathrm{\Gamma }`$. This analysis also implies that, when there is a condensate, it starts to significantly suppress baryon number violation when the induced gauge field mass is of order $`g^2T/3`$, which requires $`\varphi _02gT/3`$. Of course for such a small condensate, the perturbative notion of “condensate” is lost and what we are doing is unreliable; but this should give us some estimate of how strong the transition needs to be before the drop in the sphaleron rate becomes significant.
In the standard model and in the regime where the phase transition is strong, we can estimate the symmetric phase $`m_H(T_{\mathrm{eq}})`$; at leading order in $`\lambda /g^2`$, a standard calculation from the curvature of the one loop effective potential gives
$$m_H(T_{\mathrm{eq}})=g^2T/(16\pi \sqrt{\lambda /2g^2}).$$
(19)
Plugging this into Eq. (16) gives a self-energy correction of $`\sqrt{\lambda /18g^2}`$, and a correction to the sphaleron rate of
$$\frac{\delta \mathrm{\Gamma }}{\mathrm{\Gamma }}=5\sqrt{\frac{\lambda }{18g^2}}.$$
(20)
Below we will study the case where $`\lambda /g^2=0.036`$, for which the correction is $`22\%`$. However, this mass is not in the regime $`m_Hp`$ for $`p=g^2T/3`$, so we have computed Eq. (15) numerically for this $`m_H`$ and $`p`$; it yields $`0.037`$. A rough estimate is that we will see $`\mathrm{\Gamma }`$ reduced to $`(1.037)^5=.83`$ of the Yang-Mills theory value. This estimate is quite rough, but it gives the idea that the symmetric phase rate will be lower, but not much lower, than the pure Yang-Mills theory rate.
To conclude, at the thermodynamic level, one would expect the Higgs field induced correction to infrared gauge field behavior, and hence to the sphaleron rate, to be rather small in the symmetric phase, and of order 1 for condensates smaller than $`\varphi _0(2/3)gT`$. Naturally, for larger condensates the suppression becomes very substantial, see .
## III Effective Theory in the Presence of the Higgs
Now we will go about constructing appropriate effective theories for studying baryon number violation when there is a light Higgs boson. Our treatment will follow very closely that of . Indeed, most of the complications stem from the Yang-Mills sector and have been resolved in the literature cited; the Higgs boson will introduce comparatively minor new complications.
We will also leave out the U(1) field in what follows, even though it is light. This is a common and probably reasonable approximation. We make it partly because we expect the influence of the U(1) field to be weak; there is no direct interaction between the SU(2) and U(1) gauge fields, and the U(1) physics makes rather small modifications to the thermodynamics of the SU(2)-Higgs system . Also, including the U(1) physics would prove to be a significant complication, since at the $`kg^2T`$ scale the U(1) fields are overdamped but with a $`k`$ dependent damping which must be treated as a nonlocal effect even in the final effective theory. We will not try to address this problem here.
Since we are not concerned with cases where there is a very large condensate $`\varphi _0T/g`$, degrees of freedom with momentum $`kT`$ behave at leading order as massless free fields propagating in the background of the IR fields, and the degrees of freedom with $`kgT`$ have large occupation numbers and can be treated as classical fields. The infrared behavior of the theory is described by a classical effective theory in which the $`kT`$ degrees of freedom have been analytically integrated out. The resulting equations of motion are similar to those of the classical field theory. Defining
$`H`$ $`=`$ $`H_A+H_\mathrm{\Phi }+H_E+H_\mathrm{\Pi },`$ (21)
$`H_A`$ $`=`$ $`{\displaystyle d^3x\frac{1}{4g^2}F_{ij}^aF_{ij}^a(x)},F_{ij}[D_i,D_j],`$ (22)
$`H_\mathrm{\Phi }`$ $`=`$ $`{\displaystyle d^3x(D_i\mathrm{\Phi })^{}(D_i\mathrm{\Phi })(x)}+m_H^2\mathrm{\Phi }^{}\mathrm{\Phi }(x)+\lambda (\mathrm{\Phi }^{}\mathrm{\Phi })^2(x),`$ (23)
$`H_E`$ $`=`$ $`{\displaystyle d^3x\frac{1}{2g^2}E_i^aE_i^a(x)},E_i[D_t,D_i],`$ (24)
$`H_\mathrm{\Pi }`$ $`=`$ $`{\displaystyle d^3x\frac{1}{2}\mathrm{\Pi }^{}\mathrm{\Pi }(x)},\mathrm{\Pi }D_t\mathrm{\Phi },`$ (25)
the effective field equations will be the classical equations of motion derived from this Hamiltonian, supplemented by the hard thermal loop effects which arise from the integration over the heavy modes. The hard thermal loop (HTL) field equations are
$`(D_tE_i)^a(x)`$ $`=`$ $`g^2{\displaystyle \frac{H_A}{A_i^a(x)}}g^2{\displaystyle \frac{H_\mathrm{\Phi }}{A_i^a(x)}}m_\mathrm{D}{\displaystyle \frac{d\mathrm{\Omega }_v}{4\pi }v_iW^a(x,v)},`$ (26)
$`D_t\mathrm{\Pi }(x)`$ $`=`$ $`{\displaystyle \frac{H_\mathrm{\Phi }}{\mathrm{\Phi }^{}(x)}}m_{\mathrm{th}}^2\mathrm{\Phi }(x)`$ (27)
$`=`$ $`D_iD_i\mathrm{\Phi }(x)\left(m_H^2(T)2\lambda \mathrm{\Phi }^{}\mathrm{\Phi }(x)\right)\mathrm{\Phi }(x),`$ (28)
$`(D_tW)^a(x,v)`$ $`=`$ $`v_i(D_iW)^a(x,v)+m_\mathrm{D}v_iE_i^a(x),`$ (29)
$`(D_iE_i)^a(x)`$ $`=`$ $`g^2(\mathrm{\Pi }^{}iT^a\mathrm{\Phi }(x)+\mathrm{c}.\mathrm{c}.)+m_\mathrm{D}{\displaystyle }{\displaystyle \frac{d\mathrm{\Omega }_v}{4\pi }}W^a(x,v).`$ (30)
These equations determine the field evolution, given initial information for $`A`$, $`E`$, $`\mathrm{\Phi }`$, $`\mathrm{\Pi }`$, and $`W`$, up to the freedom to choose the time dependent gauge. The last equation is Gauss’ Law and needs to be applied as a constraint on the initial conditions; it commutes with the other equations so it remains valid at later times.
The only hard thermal loop effect on the soft scalar field is a thermal mass squared correction. We discuss this point at some length in Appendix A, where we demonstrate the absence of all other HTL effects involving scalar external lines. We write the correction as $`m_{\mathrm{th}}^2`$ (the thermally induced mass squared) and the sum $`m_H^2+m_{\mathrm{th}}^2=m_H^2(T)`$. In what follows we will absorb this shift in $`m_H^2`$ into $`H_\mathrm{\Phi }`$. This does not require any modification of Eq. (26) because $`m_{\mathrm{th}}^2\mathrm{\Phi }^{}\mathrm{\Phi }`$ is independent of $`A`$.
The HTLs for gauge fields are much more complicated, and are implemented here with auxiliary fields, the $`W`$ fields, which allow them to be written in a local way . The penalty is that the $`W`$ fields depend on direction $`v`$ as well as position $`x`$. Here $`v`$ is a unit vector, and $`d\mathrm{\Omega }_v`$ is an integral over directions (the unit sphere) normalized so $`𝑑\mathrm{\Omega }_v/4\pi =1`$. We normalize $`W`$ slightly differently than the references, absorbing a factor of $`m_D`$ into its normalization so it enters the Hamiltonian with the same weight as the gauge fields. Its listing in Eq. (21) would be
$$H_W=d^3x\frac{1}{2g^2}\frac{d\mathrm{\Omega }_v}{4\pi }W^a(x,v)W^a(x,v).$$
(31)
It is possible to derive the field equations from this (generalized) Hamiltonian, together with Eq. (21), but it requires rather nontrivial Lie-Poisson brackets . Note that the hard scalar contributions to the gauge HTL is identical in structure to that from hard gauge and fermionic degrees of freedom. This is most easily seen in the kinetic theory derivation of the gauge HTL’s , where one finds that the spin of a hard degree of freedom only matters at subleading order in $`g`$.
Eqs. (26)–(30) already give an effective theory which is amenable to numerical treatment along the lines of that given for pure Yang-Mills theory in . The added complication of including the Higgs field is much less than that involved in treating the hard thermal loops. However, such a numerical implementation is not ideal, because the system of equations presented is not UV finite. In particular the classical gauge and scalar degrees of freedom, with dynamics determined by Eqs. (26) and (27), generate UV divergent loop corrections which can be considered as extra contributions to the hard thermal loops. In a lattice regularization, these extra contributions are finite but grow linearly as the lattice spacing is made smaller, and are not rotationally invariant . So long as $`m_D`$ is kept suitably large this should be a subdominant effect, but it is not clear that it is safe to make $`m_D`$ as large as the inverse lattice spacing. Because of these problems it is not easy to extract high precision information from such a simulation, and if we expect a fairly small change to $`\mathrm{\Gamma }`$ due to the Higgs bosons, it may be problematic to isolate it from lattice spacing effects. Rather, we will follow Bödeker and integrate out more degrees of freedom to construct what ultimately proves to be a simpler and cleaner effective theory (though no longer valid to corrections parametrically suppressed by a full power of $`g`$), which will prove a better test-bed for studying the importance of Higgs physics.
With this in mind we consider integrating over the $`gT`$ scale, down to some intermediate scale $`\mu gT`$. We are also removing physics with a frequency scale $`\omega gT`$; however we cannot directly consider only degrees of freedom with $`\omega kgT`$; for now $`\omega `$ is permitted to be as large as $`\mu `$.
Here we will only treat the case in which there is not a Higgs boson condensate of order $`\varphi _0\mu /g`$. This either restricts how deeply into the broken phase our analysis remains valid, or how small we are permitted to make the scale $`\mu `$ which we integrate down to. We do not consider this restriction problematic because we are mostly interested in the symmetric phase, and when there is a large condensate we will need to use other tools to determine the rate anyway. We have already considered the broken phase problem in .
Under this assumption, the behavior of gauge bosons with $`k\mu `$ is not significantly changed by a Higgs condensate. It is also not significantly changed by the background of IR Higgs fluctuations, as can be verified by a loopwise analysis like the one in the last section; and the effects of hard Higgs excitations are already included by the HTL effective action. Therefore the integration over these degrees of freedom proceeds as it does in the case without Higgs fields. The integration over gauge and $`W`$ field degrees of freedom is very nontrivial and has been treated at length by Bödeker and by Arnold, Son, and Yaffe . They show that a collision term is induced for the $`W`$ fields, together with noise required by the fluctuation dissipation theorem. The integration over the $`gT`$ scale Higgs fields is much simpler. The Higgs field equation of motion, Eq. (27), is the same below the $`gT`$ scale as it is up to the $`T`$ scale. Provided that we keep $`\mu m_H(T)`$, the Higgs field still undergoes free relativistic propagation in the soft gauge and Higgs field background up to corrections $`O(g^2T/\mu )`$. Such propagation is precisely what generated the hard thermal loops. Standard power counting shows that contributions from the Higgs fields with momentum $`k>\mu `$ to more IR degrees of freedom are suppressed by $`g^2T/\mu `$ except for UV divergent contributions. But, as shown in , the structure of UV divergences in the classical field theory coincides exactly with the hard thermal loops. Hence the $`k>\mu `$ Higgs degrees of freedom induce HTL’s and parametrically suppressed additional effects. The HTL effects from the $`gT`$ Higgs fields are smaller by a factor of $`g`$ than those from $`T`$ physics, so we can neglect them next to the HTL effects already being included.
It is worth remarking why physically no collision integral is induced when we integrate out scalars with momentum $`kgT`$. The total scattering rate for hard modes by exchange of a soft gauge particle is $`g^2T\mathrm{log}(1/g)`$, with the logarithm arising from the momentum region $`gT`$ to $`g^2T`$. This large collision rate originates from the IR singular $`s^2/t^2`$ matrix elements in gauge boson exchange. When we integrate out $`gT`$ bosons we must explicitly include their contribution to scatterings via a collision integral. On the other hand, no scattering process between hard modes which is mediated by single scalar exchange has a scattering rate greater than $`g^4T\mathrm{log}(T/m_H)`$.This can be verified by looking at the matrix elements of tree level $`22`$ scattering processes mediated by a scalar, none of which contain $`s^2/t^2`$ type terms. At worst they go as $`s^2/tu`$ and are log IR divergent. Such rare scatterings are not important at leading order because we are eventually interested in physics at the $`1/g^2T`$ length scale.
On length scales more infrared than $`1/gT`$, Eqs. (26) and (29) describe overdamped evolution for the gauge fields. Further, the $`W`$ field equation of motion is linear in its slowly varying source $`v_iE_i`$. Therefore it is permissible at leading order in $`\mu /gT`$ to drop both the $`D_tE`$ term in Eq. (26) and the $`D_tW`$ term in Eq. (29) , leading to an effective theory for the $`\mu gT`$ scale physics, <sup>\**</sup><sup>\**</sup>\**For a discussion of the last equation, see .
$`m_\mathrm{D}{\displaystyle \frac{d\mathrm{\Omega }}{4\pi }v_iW^a(x,v)}`$ $`=`$ $`g^2{\displaystyle \frac{H_A}{A_i^a(x)}}g^2{\displaystyle \frac{H_\mathrm{\Phi }}{A_i^a(x)}},`$ (32)
$`D_t\mathrm{\Pi }(x)`$ $`=`$ $`{\displaystyle \frac{H_\mathrm{\Phi }}{\mathrm{\Phi }^{}(x)}},`$ (33)
$`v_i(D_iW)^a(x,v)`$ $`=`$ $`m_\mathrm{D}v_iE_i^a(x){\displaystyle \frac{d\mathrm{\Omega }_v^{}}{4\pi }C_{vv^{}}W^a(x,v^{})}+\zeta ^a(x,v),`$ (34)
$`m_D{\displaystyle \frac{d\mathrm{\Omega }_v}{4\pi }W^a(x,v)}`$ $`=`$ $`0.`$ (35)
The new features are the collision integral $`C_{vv^{}}`$, whose form and value is discussed in , and the noise $`\zeta `$. At leading order in $`\mathrm{log}(gT/\mu )`$ the collision integral is
$$C_{vv^{}}\gamma \left(\delta _{S^2}(vv^{})\frac{4}{\pi }\frac{(vv^{})^2}{\sqrt{1(vv^{})^2}}\right),\gamma \frac{N_cg^2T}{4\pi }\mathrm{log}\frac{m_D}{\mu }.$$
(36)
The noise is Gaussian and white with two-point correlator
$$\zeta ^a(x,v,t)\zeta ^b(y,v^{},t^{})=2g^2TC_{vv^{}}\delta ^{ab}\delta (xy)\delta (tt^{}).$$
(37)
Because of our nonstandard $`W`$ field normalization, the normalization of the noise differs from the references. Note that the size of the collision integral is parametrically $`C_{vv^{}}g^2T\mathrm{log}(gT/\mu )`$.
At the $`kg^2T`$ scale this effective theory has two natural time scales. First there is the time scale $`1/g^2T`$, on which Eq. (33) allows the Higgs fields to evolve. There is also a scale, $`1/g^4T\mathrm{log}(1/g)`$, on which the overdamped gauge and $`W`$ fields evolve. This reflects the fact that the gauge HTL’s include Landau damping and lead to overdamped evolution, while the Higgs HTL’s are just a mass correction and do not induce any $`g^2T^2`$ size damping, so the Higgs field is not overdamped.
If we view the system on length scales $`1/g^2T`$ and on the Higgs time scale $`1/g^2T`$, the gauge fields look “frozen” up to parametrically suppressed effects. On these time scales the Higgs field evolution is determined by Eq. (33) with the gauge field background frozen. What is the behavior of such a system? The Higgs propagates on an inhomogeneous background connection, interacting with itself via the nonlinear quartic coupling term. There is a widely (though not universally) believed conjecture that the evolution of a scalar field theory with quartic self-coupling in 3 dimensions should be ergodic, in which case it will randomize itself on the $`1/g^2T`$ time scale (since there is no other available time scale for its evolution). This view is supported, for instance, by the calculation of the damping rate of a scalar at rest, in a flat connection but with a quartic interaction; the damping rate is parametrically $`\lambda ^2T^2/m_Hg^2T`$, and the damping arises primarily from degrees of freedom with $`km_H`$ . We expect that the inhomogeneous connection should only make the randomization of the scalar field more efficient. In particular we speculate that the spectrum of the $`D^2`$ operator for a typical 3-D gauge field background exhibits Anderson localization at all frequencies.
On the $`1/g^4T`$ time scale on which the gauge field evolves, the Higgs field will thoroughly explore its fixed connection thermal ensemble. Therefore, on the time scale on which $`A`$ evolves, it will see a thermodynamic average of the possible Higgs field configurations. We emphasize that we are relying here on the conjectured ergodicity of 3-D scalar $`\varphi ^4`$ theory. It is also not clear that our treatment will remain true very near the endpoint of the electroweak phase transition, where the Higgs field correlation length grows to be $`1/g^2T`$, because in this regime it may take much longer for the Higgs field to explore its fixed gauge field ensemble. We will exclude that regime from consideration, although it is not clear to us that the effective theory we will derive cannot be used there as well.
When our assumption is valid, we should average the $`\mathrm{\Phi }`$ dependent part of Eq. (32) over the Higgs thermal ensemble,
$$g^2\frac{H_\mathrm{\Phi }}{A_i^a}g^2\frac{H_\mathrm{\Phi }}{A_i^a}=g^2T\frac{}{A_i^a}\mathrm{log}𝒟\mathcal{\Phi }\mathrm{exp}(H_\mathrm{\Phi }/T).$$
(38)
The RHS of Eq. (32) can be understood as $`g^2T`$ times a variation of a 3-D effective action describing the thermodynamics of the gauge fields. The sole modification from the inclusion of Higgs fields is that the effective action should include, besides the gauge part, a nonlocal piece arising from integrating over the Higgs fields. Hence the effective theory describing infrared gauge bosons is
$`m_\mathrm{D}{\displaystyle \frac{d\mathrm{\Omega }}{4\pi }v_iW^a(x,v)}`$ $`=`$ $`g^2T{\displaystyle \frac{}{A_i^a(x)}}{\displaystyle \frac{H_{\mathrm{eff}}(A)}{T}},`$ (39)
$`{\displaystyle \frac{H_{\mathrm{eff}}(A)}{T}}`$ $`=`$ $`\mathrm{log}{\displaystyle 𝒟\mathrm{\Phi }\mathrm{exp}\left(\frac{H_A+H_\mathrm{\Phi }}{T}\right)},`$ (40)
$`v_i(D_iW)^a(x,v)`$ $`=`$ $`m_\mathrm{D}v_iE_i^a(E){\displaystyle \frac{d\mathrm{\Omega }_v^{}}{4\pi }C_{vv^{}}W^a(x,v^{})}+\zeta ^a(x,v).`$ (41)
This is identical with the “theory 2” of reference , except for the Higgs field additions to $`H_{\mathrm{eff}}`$.
This theory is probably not well suited to numerical study. However, as Bödeker has shown, it has fairly simple behavior when studied at the length scale $`1/g^2T`$ . In this regime the collision integral dominates over the derivative term for the $`W`$ field<sup>††</sup><sup>††</sup>††What follows is a gross oversimplification of the argument, see ., since $`Cg^2T\mathrm{log}(gT/\mu )`$ with $`\mu g^2T`$, while the derivative term is $`vDg^2T`$; the collision term is therefore bigger by $`\mathrm{log}(1/g)`$. Roughly speaking, one may solve for $`W`$ in terms of $`E`$ in Eq. (41) and plug it into Eq. (39), yielding a local expression of form $`\sigma E=H/A`$.
At leading log order the Yang-Mills theory argument goes over directly to the case including a Higgs field, since it depends on manipulations of Eq. (41) only and this does not include the Higgs field. The next to leading log calculation of also still holds, provided there is no Higgs condensate of size $`\varphi _0\text{ }\stackrel{>}{}\text{ }gT/\mathrm{log}(1/g)`$. We discuss this point at more length in an appendix, where we show how the addition of the Higgs field does not modify the calculation presented in . Intuitively, the reason the Higgs field does not affect $`\sigma `$ at next to leading log order is as follows. The conductivity depends on the efficiency of collisions. The collisions are mediated by gauge excitations with momenta $`gT\text{ }\stackrel{>}{}\text{ }k\text{ }\stackrel{>}{}\text{ }g^2T\mathrm{log}(1/g)`$. At the low end of this range the Higgs fields modify the gauge field thermodynamics by $`O(1/\mathrm{log}(1/g))`$, see Section II, and this part of the range gives a contribution down by $`1/\mathrm{log}(1/g)`$ compared to the complete collision integral. Hence the influence of the Higgs field is suppressed by two powers of log.
Finally, we arrive at the effective theory
$`\sigma E_i^a(x)`$ $`=`$ $`g^2T{\displaystyle \frac{}{A_i^a(x)}}{\displaystyle \frac{H_{\mathrm{eff}}(A)}{T}}+\xi _i^a(x),`$ (42)
$`\xi _i^a(x,t)\xi _j^b(y,t^{})`$ $`=`$ $`2g^2T\sigma \delta _{ij}\delta _{ab}\delta ^3(xy)\delta (tt^{}),`$ (43)
$`\sigma ^1`$ $`=`$ $`{\displaystyle \frac{3}{m_D^2}}\gamma ,\gamma ={\displaystyle \frac{N_\mathrm{c}g^2T}{4\pi }}\left[\mathrm{ln}{\displaystyle \frac{m_D}{\gamma }}+3.041\right].`$ (44)
Here $`H_{\mathrm{eff}}`$ is as in Eq. (40) and the expression for $`\sigma ^1`$ is from with a particularly nice choice for the renormalization scale $`\mu `$. The sole modification the inclusion of the Higgs fields has made is in the form of $`H_{\mathrm{eff}}`$. Again we emphasize that the the derivation has assumed that the Higgs is light and that there is not a large Higgs condensate. If there is a large Higgs condensate $`\varphi _0\text{ }\stackrel{>}{}\text{ }gT/\mathrm{log}(1/g)`$ then the value of $`\sigma `$ changes, and if $`\varphi _0T`$ then the form of the effective theory changes as well. We will not discuss the latter case here. Also note that our effective theory is only valid for studying gauge field correlators; it cannot tell us much about unequal time Higgs field correlators.
## IV Numerics
Bödeker’s effective theory is Eq. (42), but with $`H_{\mathrm{eff}}`$ replaced with $`H_A`$. It makes an excellent starting point for numerical investigation of $`\mathrm{\Gamma }`$ in pure Yang-Mills theory for two reasons:
1. it is local, and
2. it is UV finite.
Neither is true for the effective theory we have derived. This potentially makes its study much more problematic than Bödeker’s effective theory.
In practice the second problem is not a substantial one. The theory is not UV finite because $`H_{\mathrm{eff}}`$ involves the 3 dimensional path integral for a scalar field, which contains linear and logarithmic mass divergences arising from one and two loop graphs. However, the path integral is still super-renormalizable, and the UV infinities are purely local and can be absorbed with a mass counterterm. Their value is known , and as we will discuss below, we could actually proceed even if they were not.
The real problem with implementing the effective theory is its nonlocality, which comes about because of the path integral in the expression for $`H_{\mathrm{eff}}`$, see Eq. (40). The solution is to think about how we would carry out such a path integral numerically. To evaluate Eq. (40) numerically, we would perform the path integral by Monte-Carlo, for instance by a Langevin equation,
$`{\displaystyle \frac{d\mathrm{\Phi }}{d\tau _\varphi }}(x)`$ $`=`$ $`{\displaystyle \frac{H_\mathrm{\Phi }}{\mathrm{\Phi }^{}(x)}}+\xi (x),`$ (45)
$`\xi (x,\tau _\varphi )\xi ^{}(x^{},\tau _\varphi ^{})`$ $`=`$ $`2T\mathbf{\hspace{0.33em}1}\delta (xx^{})\delta (\tau _\varphi \tau _\varphi ^{}).`$ (46)
Here the $`\mathrm{𝟏}`$ reminds us to make the noise diagonal in the components of the Higgs field. This Langevin equation must be evolved, and the result averaged, at every time step in the Langevin dynamics of the gauge fields.
This suggests that Eq. (42) can be replaced with the $`\eta \mathrm{}`$ limit of the following system of equations:
$`\sigma E_i^a(x)`$ $`=`$ $`g^2{\displaystyle \frac{}{A_i^a(x)}}(H_A(A)+H_\mathrm{\Phi }(A,\mathrm{\Phi }))+\xi _i^a(x,t),`$ (47)
$`\sigma D_t\mathrm{\Phi }(x)`$ $`=`$ $`\eta {\displaystyle \frac{}{\mathrm{\Phi }^{}(x)}}H_\mathrm{\Phi }(A,\mathrm{\Phi })+\xi _\mathrm{\Phi }(x,t),`$ (48)
$`\xi _\mathrm{\Phi }(x,t)\xi _\mathrm{\Phi }^{}(x^{},t^{})`$ $`=`$ $`2\eta \sigma T\mathbf{\hspace{0.33em}1}\delta (xx^{})\delta (tt^{}).`$ (49)
This is Langevin evolution but with a different rate for the evolution of the two fields, $`A`$ and $`\mathrm{\Phi }`$. In the limit that the Higgs field evolution is made infinitely fast, which is $`\eta \mathrm{}`$, the Higgs field evolution will perform the path integral in a much shorter time scale than the gauge fields evolve in, and we recover the desired equations of motion, Eq. (42).
Eq. (47) at a finite value of $`\eta `$ does make a good starting point for numerical work. The large $`\eta `$ limit must then be taken numerically. It is also convenient to rescale time to a Langevin time, $`\tau \sigma t`$. Note that the dimensions of $`\tau `$ are those of a length squared.
Our nonperturbative regularization for Eq. (47) will be the lattice. Since all of the numerical tools we use exist in the literature, we will only present the relevant references here rather than give complete details. The lattice discretization is standard, see . Our topological lattice definition of $`N_{\mathrm{CS}}`$ is the same as in . The relation between the couplings of the lattice and continuum systems has been worked out in , and we use the expressions there. These relations match all thermodynamic quantities at $`O(a)`$, leaving $`O(a^2)`$ errors, with two exceptions. We do not know the full $`O(a)`$ match for an additive piece of the $`\varphi ^2`$ operator insertion or for $`m_H^2(T)`$. If we were interested in determining $`\mathrm{\Gamma }`$ at a particular, fixed value of $`m_H^2(T)`$ then this could pose a problem. However, what we want is $`\mathrm{\Gamma }`$ when $`m_H^2(T)`$ is a fixed distance from the value which gives phase equilibrium, $`m_{\mathrm{eq}}^2`$. To determine this we do not need the absolute normalization of $`m_H^2(T)`$ to $`O(a)`$, but we will need to find $`m_{\mathrm{eq}}^2`$ numerically.
Now we briefly discuss algorithm. To determine $`m_{\mathrm{eq}}^2`$ we use the same algorithm and multicanonical techniques as . To perform Eq. (47), we may use whatever Higgs update we choose, provided that it is stochastic and that we take the $`\eta \mathrm{}`$ limit. We choose a mixture of heat bath Higgs updates and the x-y over-relaxation algorithm of . For the gauge field update we should use Langevin dynamics or any other strictly dissipative dynamics. We choose heat bath. For either algorithm, the effective infrared dynamics should be of the Langevin form, and the difference between the algorithms will be a radiative rescaling of the (Langevin) time scale and high dimension corrections, which first appear at $`O(a^2)`$<sup>‡‡</sup><sup>‡‡</sup>‡‡The question of what modifications may occur in the effective IR behavior has been addressed at some length by Arnold and Yaffe, see particularly Appendix A of . The radiative rescaling of the time scale is discussed at length in Appendix A of . That paper presents an analytic calculation of the lattice to continuum match for the Langevin algorithm, and the Langevin to heat-bath rescaling is found numerically. Within error it is equal to the coefficient $`Z_g^1`$ which appears in the $`O(a)`$ thermodynamical match detailed in . We will assume that this numerically determined relation holds analytically.
## V Results
With a numerical implementation of the effective theory in hand we will address two questions. First, how large is $`\mathrm{\Gamma }`$ in the symmetric phase, at $`m_{\mathrm{eq}}^2`$, for parameters where the phase transition is strong enough to preserve baryon number after its completion? Second, how does $`\mathrm{\Gamma }`$ vary as we go through the analytic crossover present for large vacuum Higgs masses?
To address the first question we consider Yang-Mills Higgs theory at $`\lambda /g^2=0.036`$, which is roughly the value at which the phase transition is barely strong enough to preserve baryon number cosmologically after its completion.<sup>\**</sup><sup>\**</sup>\**The value we use is slightly smaller than that found in because here we do not include the U(1) field. We determine $`m_{\mathrm{eq}}^2`$ at three lattice spacings, $`a=4/7g^2T`$, $`a=4/9g^2T`$, and $`a=4/12g^2T`$ (which in the notation which is sometimes customary are $`\beta =7`$, 9, and 12), by multicanonical Monte-Carlo. The probability distributions for $`\varphi ^2`$ at $`m_{\mathrm{eq}}^2`$ for two lattice volumes and $`a=4/9g^2T`$ are shown in Fig. 2.
First we consider the $`\eta \mathrm{}`$ limit at fixed lattice spacing $`a=4/9g^2T`$ in a cubic volume $`L^3`$ with periodic boundary conditions, for $`L=14.2/g^2T`$ (32 sites on a side). This volume is abundantly large enough to see the large volume value of $`\mathrm{\Gamma }`$ , and is large enough that strong metastability prevents tunneling to the broken phase at $`m_{\mathrm{eq}}^2`$, where we work. Table I presents $`\kappa ^{}`$, defined in Eq. (3), for various values of $`\eta `$, lattice spacing, and $`m_H^2`$. The $`\eta `$ dependence is weak and statistically compatible with zero.
Table I also shows that, as expected, there is almost no lattice spacing dependence in $`\mathrm{\Gamma }`$. The $`O(a)`$ match for the thermodynamic quantities and time scales is essential here, see . Finally, since in the usual electroweak baryogenesis scenario the symmetric phase undergoes supercooling, we should study $`\mathrm{\Gamma }`$ in the supercooled symmetric phase. Table I shows that the maximum supercooling compatible with strong metastability is still not enough to significantly change $`\mathrm{\Gamma }`$. We cannot increase the supercooling beyond what was used because the lattice system will nucleate to the broken phase. Indeed, one run used for the table ended with a nucleation to the broken phase, as is seen clearly from the time history for $`N_{\mathrm{CS}}`$ in that run, shown in Fig. 3. Naturally, the broken phase part of the evolution was not used in the analysis. Combining all the figures in the table, since all are statistically compatible and no trend (lattice spacing, supercooling, or $`\eta `$) is statistically significant, we get
$$\mathrm{\Gamma }_{\mathrm{symm}}\left[8.24\pm 0.10\right]\left(\frac{g^2T^2}{m_D^2}\right)\left(\mathrm{log}\frac{m_D}{g^2T}+C\right)\alpha ^5T^4,$$
(50)
for $`\lambda /g^2=0.036`$. We have computed the pure Yang-Mills theory value at the same lattice spacing with comparable precision to facilitate comparison; the symmetric phase value is $`0.832\pm .015`$ of the Yang-Mills theory value. (The Yang-Mills theory value found here is smaller than that quoted in , where we found $`\kappa ^{}=10.8\pm 0.7`$. This is a statistical fluctuation in the data there. The table includes a higher statistics redetermination of $`\kappa ^{}`$ at the same parameters used in that paper, with a result in agreement with the $`\beta =9`$ value found here, and statistically compatible at $`1\sigma `$ with the result determined in .)
Of course the above is academic because $`\lambda /g^2=0.036`$ converts to a vacuum Higgs mass $`m_H=44`$GeV at tree level. Including the large radiative top Yukawa corrections, this $`\lambda /g^2`$ does not correspond to any physical Higgs mass . In the standard model, all experimentally allowed Higgs masses fail to provide an electroweak phase transition at all. Rather, there is an analytic crossover . Although baryogenesis is probably impossible in that setting, it would still be interesting to see how $`\mathrm{\Gamma }`$ varies as we go through the crossover. Does it turn on suddenly or gradually, and does its turn-on point coincide with the peak in the susceptibility for $`\varphi ^2`$? We answer this question in Fig. 4, which shows how the $`\varphi ^2`$ susceptibility
$$\chi _{\varphi ^2}\frac{1}{V}\left[\left(\varphi ^2\right)^2\varphi ^2^2\right],$$
(51)
and the sphaleron rate $`\mathrm{\Gamma }`$ vary with $`m_H^2(T)`$. The data are for $`\lambda /g^2=5/16`$, corresponding at tree level to a physical vacuum $`m_H=130`$GeV, and were taken for a $`32^3`$ box with lattice spacing $`a=1/2g^2T`$. The sphaleron rate data were all taken using $`\eta =20`$. We see that the switch-on of $`\mathrm{\Gamma }`$, though smooth, is fairly rapid and occurs at slightly lower $`m_H^2(T)`$ than the peak susceptibility; that is, it is when conditions are a little more “broken phase-like” than when the susceptibility peaks. The sphaleron rate $`\mathrm{\Gamma }`$ proves a rather good order parameter to distinguish Higgs-like and symmetric-like phases. Although, like any order parameter must, it shows smooth behavior, the range where it is far from both its broken phase value $`\kappa ^{}0`$ and its symmetric phase value $`\kappa ^{}>5`$ is quite narrow, roughly as narrow as the peak in the $`\varphi ^2`$ susceptibility. However, we do not view the sphaleron rate as competitive with the susceptibility as a probe of where the crossover occurs. The main reason is that it is possible to perform a multicanonical reweighting which allows the susceptibility to be scanned in a wide range of $`m_H^2`$ (the data in the figure come from a single numerical run), and the statistics for the susceptibility improve more quickly. For instance, the $`N_{\mathrm{CS}}`$ diffusion data in Fig. 4 took more CPU time than the susceptibility data, but are substantially “dirtier”. (The data sets are plotted with their error bars. The $`N_{\mathrm{CS}}`$ diffusion errors are each independent, but there is very large cross-correlation in errors of neighboring points for the susceptibility because they were all computed from one data set.)
## VI Conclusion
Assuming that classical 3-D scalar $`\varphi ^4`$ theory is ergodic, the addition of a light Higgs degree of freedom replaces Bödeker’s effective theory for the evolution of infrared gauge fields with a slightly more complicated equation, Eq. (42). The sole change is in the thermodynamic potential for the gauge fields. Including the Higgs field makes this thermodynamic potential nonlocal. However, the effective theory is still a useful starting point for numerical work because we can use the limit of a sequence of local effective theories, namely, Langevin (or heat bath) evolution for the gauge fields and for the Higgs fields, but with much faster Langevin evolution for the Higgs degrees of freedom.
When the electroweak phase transition is strong, the sphaleron rate in the symmetric phase is reduced by around $`20\%`$ from its Yang-Mills theory value, roughly in accord with an estimate based on the thermodynamics. (The thermodynamic estimate for small $`\lambda /g^2`$ is $`(\delta \mathrm{\Gamma }/\mathrm{\Gamma })1.2\sqrt{\lambda /g^2}`$, see Eq. (20).) When there is no electroweak phase transition, but an analytic crossover, the sphaleron rate changes rather quickly from its symmetric phase value to nearly zero, roughly at the same value of $`m_H^2(T)`$ where the $`\varphi ^2`$ susceptibility peaks. The crossover region is of about the same width as the peak in the $`\varphi ^2`$ susceptibility and is displaced to slightly lower $`m_H^2(T)`$.
In this paper we have only studied the standard model, either for parameters which are ruled out experimentally or for which the electroweak phase transition cannot provide for baryogenesis. However it is fairly simple to see how to extend the work to more viable models like the MSSM (minimal supersymmetric standard model). In that case, the SU(2) and SU(3) gauge fields would each evolve under Langevin dynamics, with a thermodynamic potential arising from integrating over the Higgs and scalar top fields. However, from our results with a strong phase transition, it should be clear that, in all cases where the phase transition is strongly first order, $`\mathrm{\Gamma }`$ in the symmetric phase will be very close to its Yang-Mills theory value. This is because we have found that the suppression of $`\mathrm{\Gamma }`$ corresponds well to what we expect thermodynamically; and when the electroweak phase transition is strong in the MSSM, the symmetric phase Higgs mass is larger than in the standard model case. Therefore the SU(2) thermodynamics in the symmetric phase is closer to Yang-Mills theory in the MSSM than in the standard model, see Eq. (16). Hence it is almost certainly true that, in the MSSM and when the phase transition is strong, the symmetric phase sphaleron rate is lower than but within $`20\%`$ of the pure Yang-Mills theory value. In practice this means we can continue to quote the Yang-Mills theory result for all symmetric phase cases of physical interest, with modest error.
## Acknowledgments
I am grateful to Dam Son for a useful discussion in which he conjectured that the Higgs fields would enter only through $`H_{\mathrm{eff}}`$. I also acknowledge useful conversations with Larry Yaffe and Dietrich Bödeker. This work was partially supported by the DOE under contract DE-FGO3-96-ER40956.
## A Higgs fields and Hard Thermal Loops
In this appendix we show that the only hard thermal loop needed for the bosonic effective theory, which has soft external scalar field lines, is the scalar mass correction.
We need only consider HTL’s with bosonic external lines, since the appropriate IR effective theory we seek is bosonic. Hence we need in general to consider all diagrams with $`(n_s>0,\mathrm{even})`$ scalar and $`n_g`$ gauge boson external lines. We will use repeatedly the power counting rules derived in , which we repeat here for the reader’s convenience. A hard thermal loop is always an $`O(T^2)`$ contribution from the $`KT`$ momentum region of a one loop integral with soft ($`gT`$) external lines. To determine the largest power of $`T`$ possible from a diagram, follow these rules:
1. the loop integration contributes $`T^3`$;
2. the first propagator times the Matsubara sum contributes $`1/T`$;
3. every additional propagator contributes $`1/(PT)`$;
4. powers of $`K^\mu `$ in the numerator (from 3 point gauge vertices or fermionic propagators) each contribute $`T`$;
5. when there are two or more propagators, and all propagators are either bosonic or fermionic, there is an extra $`P/T`$ suppression.
If the result has a weaker power of $`T`$ than $`T^2`$ the diagram does not contribute an HTL; if it is $`T^2`$ the diagram will unless there is some cancellation.
Since we are only concerned with HTL’s with bosonic external legs, then either all the propagators in the loop will be bosons or all will be fermions. Any diagram with more than 1 propagator will get a $`P/T`$ from rule 5. and this will exclude diagrams with 4 point vertices (4 gluon, 2 gluon and 2 scalar, or 4 scalar) except for tadpoles, just as is the case in Yang-Mills theory. This leaves all diagrams of the general form of diagrams (c) and (d) in Figure 5.
The scalar self-energy diagrams are considered in (fermion loop) and (boson loops). The result is that the self-energy is a momentum independent mass correction. In particular it is useful to reproduce the argument for the gauge HTL from diagram (a) of Figure 5. In Feynman gauge this diagram contributes (following the notation of that paper, where capital letters are 4-vectors and lower case are spatial or temporal (0 subscript) components)
$`{\displaystyle \frac{N^21}{2N}}g^2T{\displaystyle \underset{k_0}{}}{\displaystyle \frac{d^3k}{(2\pi )^3}\frac{(2P+K)^\mu (2P+K)^\nu g_{\mu \nu }}{K^2(K+P)^2}}`$ (A1)
$`=`$ $`{\displaystyle \frac{N^21}{2N}}g^2TT{\displaystyle \underset{k_0}{}}{\displaystyle \frac{d^3k}{(2\pi )^3}\left[\frac{2(K+P)^2K^2+2P^2}{K^2(K+P)^2}\right]}.`$ (A2)
For the first term in the square brackets, the $`(K+P)^2`$ in the numerator cancels a propagator in the denominator, leaving one propagator and no powers of $`K`$ in the numerator. This gives an HTL effect only because there is a single propagator left, so 5. above is not invoked. The result is a momentum independent tadpole. The second term is similar. The last term has two propagators and no power of $`K`$ in the numerator; it is $`O(T^0)`$ and does not give an HTL.
Similarly, diagram (b) gives, in Feynman gauge and labeling the incoming scalar momentum $`P_1`$, the incoming gauge momentum $`P_2`$, and the outgoing scalar momentum $`P_3=P_1+P_2`$,
$$g^3T\underset{k_0}{}\frac{d^3k}{(2\pi )^3}\frac{(K+2P_1)(K+2P_3)(2K+P_1+P_3)^\mu }{K^2(K+P_1)^2(K+P_3)^2}.$$
(A3)
It is convenient to rewrite
$$(K+2P_1)(K+2P_3)=(K+P_1)^2+(K+P_3)^2K^2+(4P_1P_3P_1^2P_3^2).$$
(A4)
Each $`K^2`$ type term will cancel a propagator, leaving two propagators and one $`K^\mu `$ in the numerator. By the power counting rules this contributes at most $`T^3(1/T)(1/PT)T(P/T)T^1`$ and does not give a hard thermal loop. The $`P^2`$ term is $`O(T^0)`$, so it does not either.
This example illustrates why diagrams of form (c) and (d), with $`(n_s>0,\mathrm{even})`$ scalars and $`n_g`$ gauge bosons, $`(n_s+n_g)>2`$, do not give hard thermal loops. Such a diagram has $`(n_s+n_g)>2`$ propagators. It also has $`n_s+n_g`$ factors of $`K^\mu `$ in the numerator, either all from gauge vertices (bosonic loops) or all from fermionic propagators (fermionic loops). However, there are only $`n_g`$ external Lorentz indices. In Feynman gauge, the diagram will be a sum of terms, with at least $`n_s`$ of the $`K^\mu `$ contracted against each other in each term. Hence, each term from each diagram contains in its numerator, in Feynman gauge, a term of form $`(K+P_i)(K+P_j)`$, with $`P_i,P_j`$ some linear combinations of external momenta and $`K`$ the loop momentum. This can always be written as $`(\mathrm{sign})(K+P_l)^2+O(P^2)`$, with each $`(K+P_l)`$ a momentum on some propagator. Consider one of the $`(K+P_l)^2`$ terms. Cancelling the $`(K+P_l)^2`$ against the appropriate propagator leaves $`(n_s+n_g2)`$ $`K^\mu `$ factors in the numerator but $`(n_s+n_g1)>1`$ propagators. The power counting then gives at most $`T^3(1/T)(1/PT)^{n_s+n_g2}T^{n_s+n_g2}(P/T)T^1`$, too small to contribute an HTL, even without further cancellations. Meanwhile, the $`O(P^2)`$ term has $`(n_s+n_g)`$ propagators but $`(n_s+n_g2)`$ $`K^\mu `$ factors; it is at most $`P^2T^3(1/T)(1/PT)^{n_s+n_g1}T^{n_s+n_g2}(P/T)T^0`$ and also gives no HTL. Though we worked in Feynman gauge, we expect the hard thermal loops to be gauge invariant; if they vanish except a momentum independent scalar mass in Feynman gauge, they should in any gauge.
Hence the only bosonic HTL with scalar external lines is the gauge field mass insertion. Note that similar reasoning to what we give above also rules out HTL’s with both scalar and fermionic external lines; but this is unimportant for the purposes of this paper.
## B Higgs fields and next to leading log conductivity
In this appendix we briefly explain how the calculation presented in is modified by the presence of Higgs fields, and why the modification does not change the conductivity of the final effective theory at next to leading log order. The purpose of the appendix is to outline the argument, at times we will be sloppy with notations and with terms which are not relevant and have been discussed at much more length in .
The relevant part of the calculation in is the match between their “theory 2” and “theory 3”, only now “theory 2” is given by Eqs. (39-41), which for convenience we repeat here:
$`m_\mathrm{D}{\displaystyle \frac{d\mathrm{\Omega }}{4\pi }v_iW^a(x,v)}`$ $`=`$ $`g^2T{\displaystyle \frac{}{A_i^a(x)}}{\displaystyle \frac{H_{\mathrm{eff}}(A)}{T}},`$ (B1)
$`{\displaystyle \frac{H_{\mathrm{eff}}(A)}{T}}`$ $`=`$ $`\mathrm{log}{\displaystyle 𝒟\mathrm{\Phi }\mathrm{exp}\left(\frac{H_A+H_\mathrm{\Phi }}{T}\right)},`$ (B2)
$`v_i(D_iW)^a(x,v)`$ $`=`$ $`m_\mathrm{D}v_iE_i^a(E){\displaystyle \frac{d\mathrm{\Omega }_v^{}}{4\pi }C_{vv^{}}W^a(x,v^{})}+\zeta ^a(x,v).`$ (B3)
It is possible, at least formally, to invert Eq. (B3) to solve for $`W`$ in terms of $`E`$. Eq. (B1) then becomes
$$(\sigma _{ij}(D)E_j)(x)=g^2T\frac{}{A_i(x)}\frac{H_{\mathrm{eff}}(A)}{T}+\zeta _i^{},$$
(B4)
where $`\sigma (D)`$ is the nonlocal operator resulting from inverting Eq. (B3); it is discussed in . $`\zeta ^{}`$ is a noise with correlator $`\zeta ^{}\zeta ^{}=2T\sigma (D)`$. (Starting here we will aggressively suppress indices where we feel the meaning is clear.)
This Langevin equation only has a path integral representation if we are willing to accept $`(\delta H_{\mathrm{eff}}/\delta A)^2`$ inside the action of a path integral. It is not clear how to form a perturbation theory for such a path integral. However, if we rewrite $`H_{\mathrm{eff}}`$ as we did in Section IV, then it will become possible. The effective theory of interest is the $`\eta \mathrm{}`$ limit of
$`\sigma _{ij}(D)E_j`$ $`=`$ $`g^2T\left({\displaystyle \frac{H_A}{A_i}}+{\displaystyle \frac{H_\mathrm{\Phi }}{A_i}}\right),`$ (B5)
$`D_t\mathrm{\Phi }`$ $`=`$ $`\eta {\displaystyle \frac{H_\mathrm{\Phi }}{\mathrm{\Phi }^{}}}+\xi ,`$ (B6)
$`\xi (x,t)\xi ^{}(x^{},t^{})`$ $`=`$ $`2\eta T\mathrm{𝟏}\delta (xx^{})\delta (tt^{}).`$ (B7)
To write a path integral expression for this, we follow the standard trick of writing the Langevin equation and the average over the noise distribution as a path integral,
$`{\displaystyle 𝒟\zeta ^{}𝒟\xi 𝒟A}`$ $`𝒟\mathrm{\Phi }\mathrm{exp}\left({\displaystyle \zeta ^{}\frac{1}{4g^2\sigma (D)T}\zeta ^{}}\right)\mathrm{exp}\left({\displaystyle \xi \frac{1}{4\eta T}\xi }\right)`$ (B9)
$`\times \delta \left(\zeta ^{}+\sigma (D)E+g^2T\left[\delta (H_A+H_\mathrm{\Phi })/\delta A\right]\right)\delta \left(\xi +D_t\mathrm{\Phi }+\eta \left[\delta H_\mathrm{\Phi }/\delta \mathrm{\Phi }^{}\right]\right),`$
times Jacobians for each field, which are not important, see . Here the path integral performs the average over the noise, and the delta functions enforce the Langevin equations. Enforcing the delta function does the integrals over each noise, giving a path integral
$`{\displaystyle 𝒟A𝒟\mathrm{\Phi }\mathrm{exp}(L)},`$ (B10)
$`L`$ $`=`$ $`\left(\sigma (D)E+g^2T\left[{\displaystyle \frac{\delta (H_A+H_\mathrm{\Phi })}{\delta A}}\right]\right){\displaystyle \frac{1}{4g^2\sigma (D)T}}\left(\sigma (D)E+g^2T\left[{\displaystyle \frac{\delta (H_A+H_\mathrm{\Phi })}{\delta A}}\right]\right)`$ (B12)
$`+\left(D_t\mathrm{\Phi }+\eta \left[{\displaystyle \frac{\delta H_\mathrm{\Phi }}{\delta \mathrm{\Phi }^{}}}\right]\right)^{}{\displaystyle \frac{1}{4\eta T}}\left(D_t\mathrm{\Phi }+\eta \left[{\displaystyle \frac{\delta H_\mathrm{\Phi }}{\delta \mathrm{\Phi }^{}}}\right]\right).`$
Here we have not written the contributions from the Jacobians or from an extra regulation dependent term arising because of the nonlocality of $`\sigma `$, which ref. calls $`L_1[A]`$. These terms, and the reasons they can be dropped, are discussed at some length in .
Arnold and Yaffe have shown that $`\sigma `$ in Eq. (42) can be determined at next to leading log order by computing the $`\omega =0`$, $`k0`$ limit of the one loop, Coulomb gauge $`A_0A_0`$ self-energy for the theory with $`\sigma (D)`$, and finding what constant value of $`\sigma `$ is needed to get the same one loop result in the final effective theory. To account for Higgs contributions in this calculation, we must find all one loop Higgs field self-energy corrections to the $`A_0`$ field, and must determine whether their contributions survive in the $`\eta \mathrm{}`$ limit.
First we have to find what vertices couple $`\mathrm{\Phi }`$ to $`A_0`$. The only terms in the Lagrangian which contain $`A_0`$ at all are those which contain $`E`$ or $`D_t`$. They are
$$\frac{1}{4}E\sigma (D)E+\frac{1}{4\eta }(D_t\mathrm{\Phi })^{}(D_t\mathrm{\Phi })$$
(B13)
plus terms odd in $`D_t`$, which are of form
$$\frac{1}{2}E_i\frac{\delta H_A}{\delta A_i},\frac{1}{4}\left(2E_i\frac{\delta H_\mathrm{\Phi }}{\delta A_i}+(D_t\mathrm{\Phi })^{}\frac{\delta H_\mathrm{\Phi }}{\delta \mathrm{\Phi }^{}}+cc\right).$$
(B14)
The first term here is proportional to $`_tH_A`$. The second is proportional to $`_tH_\mathrm{\Phi }`$. Both are total spacetime derivatives, which can be converted to boundary terms at asymptotically early and late times and therefore neglected. Hence we only need to consider vertices which arise from the terms in Eq. (B13). These allow two new diagrams not considered in , which are the same as (a) and (b) in Figure 1 if the wavy lines are now taken to be $`A_0`$ propagators.
To show that these diagrams vanish in the small $`\eta `$ limit, we need only count powers of $`\eta `$ in the vertices and propagators. The $`A_0\mathrm{\Phi }^2`$ vertex carries a factor of $`\omega /\eta `$ and the $`A_0^2\mathrm{\Phi }^2`$ vertex carries a factor of $`1/\eta `$. The $`\mathrm{\Phi }`$ propagator is
$$\mathrm{\Phi }\mathrm{\Phi }=\frac{1}{\omega ^2/\eta +\eta k^4}.$$
(B15)
The contribution from diagram (a) then scales as
$$(\mathrm{a})\frac{1}{\eta }𝑑\omega d^3k\frac{1}{\omega ^2/\eta +\eta k^4}\eta ^{5/4},$$
(B16)
while diagram (b) behaves as
$$(\mathrm{b})𝑑\omega d^3k\frac{\omega ^2}{\eta ^2}\frac{1}{(\omega ^2/\eta +\eta k^4)^2}\eta ^{5/4},$$
(B17)
and we see that both vanish when we take $`\eta \mathrm{}`$. Therefore there are no new contributions at one loop from Higgs bosons, and the next to leading order calculation of $`\sigma `$ is unaffected. This will not be true in the case were there is a Higgs condensate of magnitude $`\varphi _0gT/\mathrm{log}(1/g)`$, because in that case the external Higgs field insertions on the $`A`$ field lines in the one loop diagrams considered in will change the gauge field propagators by order 1. The Higgs field will also be important beyond one loop, in the construction of an effective theory which can determine the $`O(1/\mathrm{log})`$ term in Eq. (3); but we will not investigate that problem here. |
warning/0001/hep-ph0001257.html | ar5iv | text | # Contents
## 1 Introduction
### 1.1 The Standard Model
The Standard Model (SM) of fundamental interactions describes strong, weak and electromagnetic interactions of elementary particles . It is based on a gauge principle, according to which all the forces of Nature are mediated by an exchange of the gauge fields of a corresponding local symmetry group. The symmetry group of the SM is
$$SU_{colour}(3)SU_{left}(2)U_{hypercharge}(1),$$
(1.1)
whereas the field content is the following:
Gauge sector : Spin = 1
The gauge bosons are spin 1 vector particles belonging to the adjoint representation of the group (1.1). Their quantum numbers with respect to $`SU(3)SU(2)U(1)`$ are:
$$\begin{array}{ccccc}\text{gluons}& G_\mu ^a:& (\underset{¯}{8},\underset{¯}{1},0)& SU_c(3)& g_s,\\ \begin{array}{cc}\text{intermediate}& \\ \text{weak bosons}& \end{array}& W_\mu ^i:& (\underset{¯}{1},\underset{¯}{3},0)& SU_L(2)& g,\\ \text{abelian boson}& B_\mu :& (\underset{¯}{1},\underset{¯}{1},0)& U_Y(1)& g^{},\end{array}$$
where the coupling constants are usually denoted by $`g_s`$, $`g`$ and $`g^{}`$, respectively.
Fermion sector : Spin = 1/2
The matter fields are fermions belonging to the fundamental representation of the gauge group. These are believed to be quarks and leptons of at least of three generations. The SM is left-right asymmetric. Left-handed and right-handed fermions have different quantum numbers:
$$\begin{array}{cccccccc}\text{quarks}& & & & & & & \\ Q_{\alpha L}^i=& \left(\begin{array}{c}U_\alpha ^i\\ D_\alpha ^i\end{array}\right)_L& =& \left(\begin{array}{c}u^i\\ d^i\end{array}\right)_L,& \left(\begin{array}{c}c^i\\ s^i\end{array}\right)_L,& \left(\begin{array}{c}t^i\\ b^i\end{array}\right)_L,& \mathrm{}& (\underset{¯}{3},\underset{¯}{2},1/3)\\ & U_{\alpha R}^i& =& u_{iR},& c_{iR},& t_{iR},& \mathrm{}& (\underset{¯}{3}^{},\underset{¯}{1},4/3)\\ & D_{\alpha R}^i& =& d_{iR},& s_{iR},& b_{iR},& \mathrm{}& (\underset{¯}{3}^{},\underset{¯}{1},2/3)\\ \text{leptons}& L_{\alpha L}& =& \left(\begin{array}{c}\nu _e\\ e\end{array}\right)_L,& \left(\begin{array}{c}\nu _\mu \\ \mu \end{array}\right)_L,& \left(\begin{array}{c}\nu _\tau \\ \tau \end{array}\right)_L,& \mathrm{}& (\underset{¯}{1},\underset{¯}{2},1)\\ & E_{\alpha R}& =& e_R,& \mu _R,& \tau _R,& \mathrm{}& (\underset{¯}{1},\underset{¯}{1},2)\end{array}$$
$`i=1,2,3`$ \- colour, $`\alpha =1,2,3,\mathrm{}`$ \- generation.
Higgs sector : Spin = 0
In the minimal version of the SM there is one doublet of Higgs scalar fields
$$H=\left(\begin{array}{c}H^0\\ H^{}\end{array}\right)(\underset{¯}{1},\underset{¯}{2},1),$$
(1.2)
which is introduced in order to give masses to quarks, leptons and intermediate weak bosons via spontaneous breaking of electroweak symmetry.
In Quantum Field Theory framework the SM is described by the following Lagrangian
$$=_{gauge}+_{Yukawa}+_{Higgs},$$
(1.3)
$`_{gauge}`$ $`=`$ $`{\displaystyle \frac{1}{4}}G_{\mu \nu }^aG_{\mu \nu }^a{\displaystyle \frac{1}{4}}W_{\mu \nu }^iW_{\mu \nu }^i{\displaystyle \frac{1}{4}}B_{\mu \nu }B_{\mu \nu }`$
$`+i\overline{L}_\alpha \gamma ^\mu D_\mu L_\alpha +i\overline{Q}_\alpha \gamma ^\mu D_\mu Q_\alpha +i\overline{E}_\alpha \gamma ^\mu D_\mu E_\alpha `$
$`+i\overline{U}_\alpha \gamma ^\mu D_\mu U_\alpha +i\overline{D}_\alpha \gamma ^\mu D_\mu D_\alpha +(D_\mu H)^{}(D_\mu H),`$
where
$`G_{\mu \nu }^a`$ $`=`$ $`_\mu G_\nu ^a_\nu G_\mu ^a+g_sf^{abc}G_\mu ^bG_\nu ^c,`$
$`W_{\mu \nu }^i`$ $`=`$ $`_\mu W_\nu ^i_\nu W_\mu ^i+gϵ^{ijk}W_\mu ^jW_\nu ^k,`$
$`B_{\mu \nu }`$ $`=`$ $`_\mu B_\nu _\nu B_\mu ,`$
$`D_\mu L_\alpha `$ $`=`$ $`(_\mu i{\displaystyle \frac{g}{2}}\tau ^iW_\mu ^i+i{\displaystyle \frac{g^{}}{2}}B_\mu )L_\alpha ,`$
$`D_\mu E_\alpha `$ $`=`$ $`(_\mu +ig^{}B_\mu )E_\alpha ,`$
$`D_\mu Q_\alpha `$ $`=`$ $`(_\mu i{\displaystyle \frac{g}{2}}\tau ^iW_\mu ^ii{\displaystyle \frac{g^{}}{6}}B_\mu i{\displaystyle \frac{g_s}{2}}\lambda ^aG_\mu ^a)Q_\alpha ,`$
$`D_\mu U_\alpha `$ $`=`$ $`(_\mu i{\displaystyle \frac{2}{3}}g^{}B_\mu i{\displaystyle \frac{g_s}{2}}\lambda ^aG_\mu ^a)U_\alpha ,`$
$`D_\mu D_\alpha `$ $`=`$ $`(_\mu +i{\displaystyle \frac{1}{3}}g^{}B_\mu i{\displaystyle \frac{g_s}{2}}\lambda ^aG_\mu ^a)D_\alpha .`$
$$_{Yukawa}=y_{\alpha \beta }^L\overline{L}_\alpha E_\beta H+y_{\alpha \beta }^D\overline{Q}_\alpha D_\beta H+y_{\alpha \beta }^U\overline{Q}_\alpha U_\beta \stackrel{~}{H}+h.c.,$$
(1.5)
where $`\stackrel{~}{H}=i\tau _2H^{}`$.
$$_{Higgs}=V=m^2H^{}H\frac{\lambda }{2}(H^{}H)^2.$$
(1.6)
Here $`\{y\}`$ are the Yukawa and $`\lambda `$ is the Higgs coupling constants, both dimensionless, and $`m`$ is the only dimensional mass parameter<sup>1</sup><sup>1</sup>1We use the usual for particle physics units $`c=\mathrm{}=1`$.
The Lagrangian of the SM contains the following set of free parameters:
* 3 gauge couplings $`g_s,g,g^{}`$;
* 3 Yukawa matrices $`y_{\alpha \beta }^L,y_{\alpha \beta }^D,y_{\alpha \beta }^U`$;
* Higgs coupling constant $`\lambda `$;
* Higgs mass parameter $`m^2`$;
* number of matter fields (generations).
All the particles obtain their masses due to spontaneous breaking of $`SU_{left}(2)`$ symmetry group via a non-zero vacuum expectation value (v.e.v.) of the Higgs field
$$<H>=\left(\begin{array}{c}v\\ 0\end{array}\right),v=m/\sqrt{\lambda }.$$
(1.7)
As a result the gauge group of the SM is spontaneously broken down to
$$SU_c(3)SU_L(2)U_Y(1)SU_c(3)U_{EM}(1).$$
The physical weak intermediate bosons are the linear combinations of the gauge ones
$$W_\mu ^\pm =\frac{W_\mu ^1iW_\mu ^2}{\sqrt{2}},Z_\mu =\mathrm{sin}\theta _WB_\mu +\mathrm{cos}\theta _WW_\mu ^3$$
(1.8)
with masses
$$m_W=\frac{1}{\sqrt{2}}gv,m_Z=m_W/\mathrm{cos}\theta _W,\mathrm{tan}\theta _W=g^{}/g,$$
(1.9)
while the photon field
$$\gamma _\mu =\mathrm{cos}\theta _WB_\mu +\mathrm{sin}\theta _WW_\mu ^3$$
(1.10)
remains massless.
The matter fields acquire masses proportional to the corresponding Yukawa couplings:
$$M_{\alpha \beta }^u=y_{\alpha \beta }^uv,M_{\alpha \beta }^d=y_{\alpha \beta }^dv,M_{\alpha \beta }^l=y_{\alpha \beta }^lv,m_H=\sqrt{2}m.$$
(1.11)
Explicit mass terms in the Lagrangian are forbidden because they are not $`SU_{left}(2)`$ symmetrical and would destroy the renormalizability of the Standard Model.
The SM has been constructed as a result of numerous efforts both theoretical and experimental. At present the SM is extraordinary successful, the achieved accuracy of its predictions corresponds to the experimental data within 5 % . All the particles except for the Higgs boson have been discovered experimentally.
However the SM has its natural drawbacks and unsolved problems. Among them are:
* large number of free parameters,
* formal unification of strong and electroweak interactions,
* the Higgs boson has not yet been observed and it is not clear whether it is fundamental or composite,
* the problem of CP-violation is not well understood including CP-violation in strong interaction,
* flavour mixing and the number of generations are arbitrary,
* the origin of the mass spectrum is unclear.
The answer to these problems lies beyond the SM.
### 1.2 RG flow in the Standard Model
The question is: how to go beyond the SM? Apparently we are talking about new particles, new structures, new interactions and new symmetries. The answer is not obvious. We describe below one of the options and assume that at high energies (small distances) fundamental interactions possess wider symmetries, in particular, new kind of symmetry, symmetry between bosons and fermions, called sypersymmetry. This is a wide subject by itself, however, we would like to look at it from the point of view of renormalization group. Let us try to go along the road offered by the renormalization group flow.
Let us take the Lagrangian of the SM and see what happens with its parameters (the couplings) when the energy scale increases. As it follows from eq.(1.1) one has three gauge couplings corresponding to $`SU(3)_c\times SU(2)_L\times U(1)_Y`$ gauge groups, $`g_s,g`$ and $`g^{}`$, respectively. In what follows it will be more useful to consider the set $`\{g_1,g_2,g_3\}=\{\sqrt{3/5}g^{},g,g_s\}`$. Besides, there are three Yukawa couplings $`y_U,y_D`$ and $`y_L`$, which are $`3\times 3`$ matrices in the generation space and one Higgs coupling $`\lambda `$.
To simplify the picture we consider only the third generation Yukawa couplings replacing the Yukawa matrices by their $`(3,3)`$ elements. Then the RG equations in the leading one-loop order are :
$`{\displaystyle \frac{d\text{a}_i}{dt}}`$ $`=`$ $`b_i\text{a}_i^2,t=\mathrm{log}(Q^2/\mu ^2),`$ (1.12)
$`{\displaystyle \frac{dY_t}{dt}}`$ $`=`$ $`Y_t\left(8\text{a}_3+{\displaystyle \frac{9}{4}}\text{a}_2+{\displaystyle \frac{17}{20}}\text{a}_1{\displaystyle \frac{9}{2}}Y_t\right),`$
$`{\displaystyle \frac{dY_b}{dt}}`$ $`=`$ $`Y_b\left(8\text{a}_3+{\displaystyle \frac{9}{4}}\text{a}_2+{\displaystyle \frac{1}{4}}\text{a}_1{\displaystyle \frac{3}{2}}Y_t\right),`$ (1.13)
$`{\displaystyle \frac{dY_\tau }{dt}}`$ $`=`$ $`Y_\tau \left({\displaystyle \frac{9}{4}}\text{a}_2+{\displaystyle \frac{9}{4}}\text{a}_13Y_t\right),`$
$`{\displaystyle \frac{d\lambda }{dt}}`$ $`=`$ $`{\displaystyle \frac{9}{8}}\text{a}_2^2+{\displaystyle \frac{9}{20}}\text{a}_2\text{a}_1+{\displaystyle \frac{27}{200}}\text{a}_1^2{\displaystyle \frac{9}{2}}\text{a}_2\lambda {\displaystyle \frac{9}{10}}\text{a}_1\lambda +6\lambda Y_t6Y_t^2+6\lambda ^2,`$ (1.14)
where to simplify the formulas we have used the notation
$$\text{a}_i\frac{\alpha _i}{4\pi }\frac{g_i^2}{16\pi ^2},(i=1,2,3),Y_k\frac{(y_{33}^k)^2}{16\pi ^2},(k=t,b,\tau ).$$
For the SM the coefficients $`b_i`$ are:
$$b_i=\left(\begin{array}{c}\hfill b_1\\ \hfill b_2\\ \hfill b_3\end{array}\right)=\left(\begin{array}{c}\hfill 0\\ \hfill 22/3\\ \hfill 11\end{array}\right)+N_{Fam}\left(\begin{array}{c}\hfill 4/3\\ \hfill 4/3\\ \hfill 4/3\end{array}\right)+N_{Higgs}\left(\begin{array}{c}\hfill 1/10\\ \hfill 1/6\\ \hfill 0\end{array}\right).$$
(1.15)
Here $`N_{Fam}`$ is the number of generations of matter multiplets and $`N_{Higgs}`$ is the number of Higgs doublets. We use $`N_{Fam}=3`$ and $`N_{Higgs}=1`$ for the minimal SM.
The initial conditions for the couplings can be taken at some scale which is experimentally favorable. Thus, for the gauge couplings one has precise measurement at the Z-boson mass scale obtained at LEP accelerator
$$\alpha _1(M_Z)=0.017,\alpha _2(M_Z)=0.034,\alpha _3(M_Z)=0.118\pm 0.005.$$
(1.16)
As for the Yukawa couplings, they are related to the running quark masses by eq.(1.11), where $`v`$ is the vacuum expectation value of the Higgs field. It can be calculated, for instance, from the Z-boson mass according to eq.(1.9) and is equal to $`v=174.1`$ GeV. Thus, knowing the quark masses one can find the values of the Yukawa couplings. One should however distinguish between the running and the pole quark masses which are determined experimentally.
Having all this in mind and solving eqs.(1.12-1.13) one has the following qualitative picture (see Fig.1). The behaviour of the Higgs quartic coupling $`\lambda `$ strongly depends on initial conditions which are unknown in this case. We return to this subject later.
The qualitative picture presented in Fig.1 contains an obvious unification pattern. The three gauge couplings are seem to unify at energy of the order of $`10^{15}10^{16}`$ GeV and so do the Yukawa couplings $`Y_b`$ and $`Y_\tau `$. What does it mean? The usual answer is given in the framework of the Grand Unification hypothesis : three gauge interactions are the three branches of a single gauge interaction described by a simple gauge group with a single coupling. All quarks and leptons belong to some representation of this group. This explains the equality of gauge and (some) Yukawa couplings at the unification scale.
The GUT hypothesis has many far reaching consequences, however, one can see that the unification scale is very high. This is not only difficult to check experimentally, but creates a big problem, called the hierarchy problem.
The point is that in a theory with two so very different scales: $`M_W10^2`$ GeV and $`M_{GUT}10^{16}`$ GeV, it is very difficult both to achieve this hierarchy of $`10^{14}`$ in a natural way and to preserve it against the radiative corrections.
Indeed, due to modern point of view, the mass scales in the SM and in GUT are defined by vacuum expectation values of the scalar fields, called the Higgs fields. Non-zero v.e.v.’s of these fields lead to spontaneous breaking of the corresponding gauge symmetry and provide masses to all the particles. So, we have at least two scalar particles with the masses of the order of $`10^2`$ and $`10^{16}`$ GeV. However, these masses obtain the radiative corrections proportional to the masses of the interacting particles. Due to inevitable interaction between the light and heavy fields the radiative corrections to the light Higgs mass are proportional to the heavy one
$$\delta m^2g^2M^2,$$
where $`g`$ is some coupling. Assuming $`m10^2GeV,M10^{16}GeV,g0.1`$, one gets the radiative correction which is $`10^{13}`$ times bigger than the mass itself. This correction obviously spoils the hierarchy unless it is canceled. A cancellation with a precision $`10^{13}`$ needs a very accurate fine tuning of the coupling constants.
Solution to the fine-tuning problem has been found in the framework of a revolutionary hypothesis: the existence of a new type of symmetry, the symmetry between bosons and fermions, called supersymmetry.
## 2 Supersymmetry
### 2.1 Motivations of SUSY
Supersymmetry or fermion-boson symmetry has not yet been observed in Nature. This is a purely theoretical invention . Its validity in particle physics follows from common belief in unification. The general idea is a unification of all forces of Nature. It defines the strategy : increasing unification towards smaller distances up to $`l_{Pl}10^{33}`$ cm including quantum gravity. However, the graviton has spin 2, while the other gauge bosons (photon, gluons, $`W`$ and $`Z`$ weak bosons) have spin 1. Unification of spin 2 and spin 1 gauge forces within unique algebra is forbidden due to the no-go theorems for any symmetry but SUSY.
If $`Q`$ is a generator of SUSY algebra, then
$$Q|boson>=|fermion>\text{and}Q|fermion>=|boson>.$$
Hence starting with the graviton spin 2 state and acting by SUSY generators we get the following chain of states
$$spin2spin3/2spin1spin1/2spin0.$$
Thus, a partial unification of matter (fermions) with forces (bosons) naturally arises out of an attempt to unify gravity with the other interactions.
The uniqueness of SUSY is due to a strict mathematical statement that algebra of SUSY is the only graded (i.e. containing anticommutators as well as commutators) Lie algebra possible within relativistic field theory .
The other motivation of SUSY is the solution of the hierarchy problem mentioned above. At the moment supersymmetry is the only known way to achieve the cancellation of quadratic terms in radiative corrections (also known as the cancellation of the quadratic divergences). Moreover, SUSY automatically cancels quadratic corrections in all orders of perturbation theory .
### 2.2 Global SUSY: algebra and representations
As can be easily seen, supersymmetry transformations differ from ordinary global transformations as far as they convert bosons into fermions and vice versa. Indeed if we symbolically write SUSY transformation as
$$\delta B=\epsilon f,$$
where $`B`$ and $`f`$ are boson and fermion fields, respectively, and $`\epsilon `$ is an infinitesimal transformation parameter, then from the usual (anti)commutation relations for (fermions) bosons
$$\{f,f\}=0,[B,B]=0$$
we immediately find
$$\{\epsilon ,\epsilon \}=0.$$
This means that all the generators of SUSY must be fermionic, i.e. they must change the spin by a half-odd amount and change the statistics.
Combined with the usual Poincaré and internal symmetry algebra the Super-Poincaré Lie algebra contains additional generators
$$\begin{array}{c}[Q_\alpha ^i,P_\mu ]=[\overline{Q}_{\dot{\alpha }}^i,P_\mu ]=0,\hfill \\ [Q_\alpha ^i,M_{\mu \nu }]=\frac{1}{2}(\sigma _{\mu \nu })_\alpha ^\beta Q_\beta ^i,[\overline{Q}_{\dot{\alpha }}^i,M_{\mu \nu }]=\frac{1}{2}\overline{Q}_{\dot{\beta }}^i(\overline{\sigma }_{\mu \nu })_{\dot{\alpha }}^{\dot{\beta }},\hfill \\ [Q_\alpha ^i,B_r]=(b_r)_j^iQ_\alpha ^j,[\overline{Q}_{\dot{\alpha }}^i,B_r]=\overline{Q}_{\dot{\alpha }}^j(b_r)_j^i,\hfill \\ \{Q_\alpha ^i,\overline{Q}_{\dot{\beta }}^j\}=2\delta ^{ij}(\sigma ^\mu )_{\alpha \dot{\beta }}P_\mu ,\hfill \\ \{Q_\alpha ^i,Q_\beta ^j\}=2ϵ_{\alpha \beta }Z^{ij},Z_{ij}=a_{ij}^rb_r,Z^{ij}=Z_{ij}^+,\hfill \\ \{\overline{Q}_{\dot{\alpha }}^i,\overline{Q}_{\dot{\beta }}^j\}=2ϵ_{\dot{\alpha }\dot{\beta }}Z^{ij},[Z_{ij},anything]=0,\hfill \\ \alpha ,\dot{\alpha },\beta ,\dot{\beta }=1,2i,j=1,2,\mathrm{},N.\hfill \end{array}$$
(2.1)
Here $`P_\mu `$ and $`M_{\mu \nu }`$ are four-momentum and angular momentum operators respectively, $`B_r`$ are internal symmetry generators, $`Q^i`$ and $`\overline{Q}^i`$ are spinorial SUSY generators and $`Z_{ij}`$ are the so-called central charges. $`\alpha ,\dot{\alpha },\beta ,\dot{\beta }`$ are spinorial indices. In the simplest case one has one spinor generator $`Q_\alpha `$ (and the conjugated one $`\overline{Q}_{\dot{\alpha }}`$) that corresponds to an ordinary or N=1 sypersymmetry. When $`N>1`$ one has an extended sypersymmetry.
An elegant formulation of supersymmetry transformations and invariants can be achieved in the framework of superspace . Superspace differs from the ordinary Euclidean (Minkowski) space by addition of two new coordinates, $`\theta _\alpha `$ and $`\overline{\theta }_{\dot{\alpha }}`$, which are grassmannian, i.e. anticommuting, variables
$$\{\theta _\alpha ,\theta _\beta \}=0,\{\overline{\theta }_{\dot{\alpha }},\overline{\theta }_{\dot{\beta }}\}=0,\theta _\alpha ^2=0,\overline{\theta }_{\dot{\alpha }}^2=0,\alpha ,\beta ,\dot{\alpha },\dot{\beta }=1,2.$$
Thus, we go from space to superspace
$$\begin{array}{cc}Space& Superspace\\ x_\mu & x_\mu ,\theta _\alpha ,\overline{\theta }_{\dot{\alpha }}\end{array}$$
A SUSY group element can be constructed in superspace in the same way as an ordinary translation in the usual space
$$G(x,\theta ,\overline{\theta })=e^{i\left(x^\mu P_\mu +\theta Q+\overline{\theta }\overline{Q}\right)}.$$
It leads to the supertranslation in superspace
$$\begin{array}{ccc}x_\mu & & x_\mu +i\theta \sigma _\mu \overline{\epsilon }i\epsilon \sigma _\mu \overline{\theta },\hfill \\ \theta & & \theta +\epsilon ,\hfill \\ \overline{\theta }& & \overline{\theta }+\overline{\epsilon },\hfill \end{array}$$
(2.2)
where $`\epsilon `$ and $`\overline{\epsilon }`$ are grassmannian transformation parameters. Taking them to be local or space-time dependent one gets local translation. And the theory that is invariant under local translations is general relativity. Thus local supersymmetry is just the theory of gravity or supergravity . This way following the gauge principle one gets a unified theory of all four interactions known as SUGRA theory.
To define the fields on a superspace consider representations of the Super-Poincaré group (2.1) . The simplest one is a scalar superfield $`F(x,\theta ,\overline{\theta })`$ which is SUSY invariant. Its Taylor expansion in $`\theta `$ and $`\overline{\theta }`$ has only several terms due to the nilpotent character of grassmannian parameters. However, this superfield is a reducible representation of SUSY. To get an irreducible one, we define a chiral superfield which obeys the equation
$$\overline{D}F=0,\text{where}\overline{D}=\frac{}{\overline{\theta }}i\theta \sigma ^\mu _\mu .$$
(2.3)
Its Taylor expansion looks like ($`y=x+i\theta \sigma \overline{\theta }`$)
$`\mathrm{\Phi }(y,\theta )`$ $`=`$ $`A(y)+\sqrt{2}\theta \psi (y)+\theta \theta F(y)`$ (2.4)
$`=`$ $`A(x)+i\theta \sigma ^\mu \overline{\theta }_\mu A(x)+{\displaystyle \frac{1}{4}}\theta \theta \overline{\theta }\overline{\theta }\mathrm{}A(x)`$
$`+`$ $`\sqrt{2}\theta \psi (x){\displaystyle \frac{i}{\sqrt{2}}}\theta \theta _\mu \psi (x)\sigma ^\mu \overline{\theta }+\theta \theta F(x)`$
The coefficients are ordinary functions of $`x`$ being the usual fields. They are called the components of a superfield. In eq.(2.4) one has 2 bosonic (complex scalar field $`A`$) and 2 fermionic (Weyl spinor field $`\psi `$) degrees of freedom. The componet fields $`A`$ and $`\psi `$ are called the superpartners. The field $`F`$ is an auxiliary field, it has the “wrong” dimension and has no physical meaning. It is needed to close the algebra (2.1). One can get rid of the auxiliary fields with the help of equations of motion.
Thus a superfield contains an equal number of bosonic and fermionic degrees of freedom. Under SUSY transformation they convert one into another.
The product of chiral superfields $`\mathrm{\Phi }^2,\mathrm{\Phi }^3`$, etc is also a chiral superfield, while the product of chiral and antichiral ones $`\mathrm{\Phi }^+\mathrm{\Phi }`$ is a general superfield.
To construct the gauge invariant interactions, we will need a real vector superfield $`V=V^+`$. It is not chiral but rather a general superfield. Its expansion over $`\theta `$ and $`\overline{\theta }`$ looks like
$`V(x,\theta ,\overline{\theta })`$ $`=`$ $`C(x)+i\theta \chi (x)i\overline{\theta }\overline{\chi }(x)`$ (2.5)
$`+`$ $`{\displaystyle \frac{i}{2}}\theta \theta [M(x)+iN(x)]{\displaystyle \frac{i}{2}}\overline{\theta }\overline{\theta }[M(x)iN(x)]`$
$``$ $`\theta \sigma ^\mu \overline{\theta }v_\mu (x)+i\theta \theta \overline{\theta }[\overline{\lambda }(x)+{\displaystyle \frac{i}{2}}\overline{\sigma }^\mu _\mu \chi (x)]`$
$``$ $`i\overline{\theta }\overline{\theta }\theta [\lambda +{\displaystyle \frac{i}{2}}\sigma ^\mu _\mu \overline{\chi }(x)]+{\displaystyle \frac{1}{2}}\theta \theta \overline{\theta }\overline{\theta }[D(x)+{\displaystyle \frac{1}{2}}\mathrm{}C(x)].`$
The physical degrees of freedom corresponding to a real vector superfield are the vector gauge field $`v_\mu `$ and its superpartner the Majorana four component spinor field made of two Weyl spinors $`\lambda `$ and $`\overline{\lambda }`$. All other components are unphysical and can be eliminated. Thus we again have an equal number of bosonic and fermionic degrees of freedom.
One can choose a gauge (Wess-Zumino gauge) where $`C=\chi =M=N=0`$, leaving us with the physical degrees of freedom except for the auxiliary field $`D`$. In this gauge
$`V`$ $`=`$ $`\theta \sigma ^\mu \overline{\theta }v_\mu (x)+i\theta \theta \overline{\theta }\overline{\lambda }(x)i\overline{\theta }\overline{\theta }\theta \lambda (x)+{\displaystyle \frac{1}{2}}\theta \theta \overline{\theta }\overline{\theta }D(x),`$
$`V^2`$ $`=`$ $`{\displaystyle \frac{1}{2}}\theta \theta \overline{\theta }\overline{\theta }v_\mu (x)v^\mu (x),`$
$`V^3`$ $`=`$ $`0,etc.`$ (2.6)
One can define also a field strength tensor (as analog of $`F_{\mu \nu }`$ in gauge theories)
$$W_\alpha =\frac{1}{4}\overline{D}^2e^VD_\alpha e^V,\overline{W}_{\dot{\alpha }}=\frac{1}{4}D^2e^V\overline{D}_\alpha e^V,$$
(2.7)
which is needed to construct gauge invariant Lagrangians.
### 2.3 SUSY Lagrangians
In the superfield notation SUSY invariant Lagrangians are the polynomials of superfields. Having in mind that for component fields we should have the ordinary terms, the general SUSY invariant Lagrangian has the form
$``$ $`=`$ $`{\displaystyle d^2\theta d^2\overline{\theta }\mathrm{\Phi }_i^+\mathrm{\Phi }_i}+{\displaystyle d^2\theta [\lambda _i\mathrm{\Phi }_i+\frac{1}{2}m_{ij}\mathrm{\Phi }_i\mathrm{\Phi }_j+\frac{1}{3}y_{ijk}\mathrm{\Phi }_i\mathrm{\Phi }_j\mathrm{\Phi }_k]}+h.c.`$ (2.8)
where the first part is a kinetic term and the second one is a superpotential $`𝒲`$. Here instead of taking the proper components we use an integration over the superspace according to the rules of grassmannian integration
$$𝑑\theta _\alpha =0,\theta _\alpha 𝑑\theta _\beta =\delta _{\alpha \beta }.$$
Performing this integration we get in components
$``$ $`=`$ $`i_\mu \overline{\psi }_i\overline{\sigma }^\mu \psi _i+A_i^{}\mathrm{}A_i+F_i^{}F_i`$
$`+`$ $`[\lambda _iF_i+m_{ij}(A_iF_j{\displaystyle \frac{1}{2}}\psi _i\psi _j)+y_{ijk}(A_iA_jF_k\psi _i\psi _jA_k)+h.c.].`$
or solving the constraints
$``$ $`=`$ $`i_\mu \overline{\psi }_i\overline{\sigma }^\mu \psi _i+A_i^{}\mathrm{}A_i{\displaystyle \frac{1}{2}}m_{ij}\psi _i\psi _j{\displaystyle \frac{1}{2}}m_{ij}^{}\overline{\psi }_i\overline{\psi }_j`$ (2.10)
$`y_{ijk}\psi _i\psi _jA_ky_{ijk}^{}\overline{\psi }_i\overline{\psi }_jA_k^{}V(A_i,A_j),`$
where $`V=F_k^{}F_k`$. Note that because of the renormalizability constraint $`VA^4`$ the superpotential should be limited by $`𝒲\mathrm{\Phi }^3`$ as in eq.(2.8).
The gauge field part of a Lagrangian is
$``$ $`=`$ $`{\displaystyle \frac{1}{4}}{\displaystyle d^2\theta W^\alpha W_\alpha }+{\displaystyle \frac{1}{4}}{\displaystyle d^2\overline{\theta }\overline{W}^{\dot{\alpha }}\overline{W}_{\dot{\alpha }}}`$ (2.11)
$`=`$ $`{\displaystyle \frac{1}{2}}D^2{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }i\lambda \sigma ^\mu D_\mu \overline{\lambda }.`$
To obtain a gauge-invariant interaction with matter chiral superfields, consider their gauge transformation
$$\mathrm{\Phi }e^{ig\mathrm{\Lambda }}\mathrm{\Phi },\mathrm{\Phi }^+\mathrm{\Phi }^+e^{ig\mathrm{\Lambda }^+},VV+i(\mathrm{\Lambda }\mathrm{\Lambda }^+),$$
where $`\mathrm{\Lambda }`$ is a gauge parameter (chiral superfield).
It is clear now how to construct both SUSY and gauge invariant interaction which is equivalent to transition from the usual to covariant derivatives
$$d^2\theta d^2\overline{\theta }\mathrm{\Phi }_i^+\mathrm{\Phi }_id^2\theta d^2\overline{\theta }\mathrm{\Phi }_i^+e^{gV}\mathrm{\Phi }_i.$$
Thus, the form of the Lagrangian is practically fixed by symmetry requirements. The only freedom is the field content, the value of the gauge coupling $`g`$ , the Yukawa couplings $`y_{ijk}`$ and the masses. This allows one to construct a SUSY generalization of the SM.
## 3 Minimal Supersymmetric Standard Model
As has been already mentioned, in SUSY theories the number of bosonic degrees of freedom equals that of fermionic. In the SM one has 28 bosonic and 90 fermionic degrees of freedom. So the SM is in great deal non-supersymmetric. Trying to add some new particles to supersymmetrize the SM, one should take into account the following observations:
1. There are no fermions with quantum numbers of the gauge bosons;
2. Higgs fields have a non-zero v.e.v.s, hence they cannot be superpartners of quarks and leptons since this would induce a spontaneous violation of baryon and lepton numbers;
3. One needs at least two complex chiral Higgs multiplets to give masses to Up and Down quarks.
The latter is due to the form of a superpotential and chirality of matter superfields. Indeed, the superpotential should be invariant under $`SU(3)\times SU(2)\times U(1)`$ gauge group. If one looks at the Yukawa interaction in the Standard Model, eq.(1.5), one finds that it is indeed $`U(1)`$ invariant since the sum of hypercharges in each vertex equal zero. In the last term this is achieved by taking the conjugated Higgs doublet $`\stackrel{~}{H}=i\tau _2H^{}`$ instead of $`H`$. However, in SUSY $`H`$ is a chiral superfield and hence a superpotential, which is constructed out of chiral fields, can contain only $`H`$ but not $`\stackrel{~}{H}`$, which is an antichiral superfield.
Another reason for the second Higgs doublet is related to chiral anomalies. It is known that chiral anomalies spoil the gauge invariance and, hence, the renormalizability of the theory. They are canceled in the SM between quarks and leptons in each generation. However, if one introduces a chiral Higgs superfield, it contains higgsinos, which are chiral fermions, and contain anomalies. To cancel them one has to add the second Higgs doublet with the opposite hypercharge.
Therefore the Higgs sector in SUSY models is inevitably enlarged, it contains an even number of doublets.
Conclusion: In SUSY models supersymmetry associates known bosons with new fermions and known fermions with new bosons.
### 3.1 The field content
Consider the particle content of the Minimal Supersymmetric Standard Model . According to the previous discussion in the minimal version we double the number of particles (introducing a superpartner to each particle) and add another Higgs doublet (with its superpartner). The particle content of the MSSM then looks as follows :
Particle Content of the MSSM
| Superfield | Bosons | Fermions | $`SU_c(3)`$ | $`SU_L(2)`$ | $`U_Y(1)`$ |
| --- | --- | --- | --- | --- | --- |
| Gauge | | | | | |
| $`𝐆^𝐚`$ | gluon $`g^a`$ | gluino$`\stackrel{~}{g}^a`$ | 8 | 0 | 0 |
| $`𝐕^𝐤`$ | Weak $`W^k`$ $`(W^\pm ,Z)`$ | wino, zino $`\stackrel{~}{w}^k`$ $`(\stackrel{~}{w}^\pm ,\stackrel{~}{z})`$ | 1 | 3 | 0 |
| $`𝐕^{}`$ | Hypercharge $`B(\gamma )`$ | bino $`\stackrel{~}{b}(\stackrel{~}{\gamma })`$ | 1 | 1 | 0 |
| Matter | | | | | |
| $`\begin{array}{c}𝐋_𝐢\\ 𝐄_𝐢\end{array}`$ | sleptons $`\{\begin{array}{c}\stackrel{~}{L}_i=(\stackrel{~}{\nu },\stackrel{~}{e})_L\hfill \\ \stackrel{~}{E}_i=\stackrel{~}{e}_R\hfill \end{array}`$ | leptons $`\{\begin{array}{c}L_i=(\nu ,e)_L\hfill \\ E_i=e_R\hfill \end{array}`$ | $`\begin{array}{c}1\hfill \\ 1\hfill \end{array}`$ | $`\begin{array}{c}2\hfill \\ 1\hfill \end{array}`$ | $`\begin{array}{c}\hfill 1\\ \hfill 2\end{array}`$ |
| $`\begin{array}{c}𝐐_𝐢\\ 𝐔_𝐢\\ 𝐃_𝐢\end{array}`$ | squarks $`\{\begin{array}{c}\stackrel{~}{Q}_i=(\stackrel{~}{u},\stackrel{~}{d})_L\hfill \\ \stackrel{~}{U}_i=\stackrel{~}{u}_R\hfill \\ \stackrel{~}{D}_i=\stackrel{~}{d}_R\hfill \end{array}`$ | quarks $`\{\begin{array}{c}Q_i=(u,d)_L\hfill \\ U_i=u_R\hfill \\ D_i=d_R\hfill \end{array}`$ | $`\begin{array}{c}3\hfill \\ 3^{}\hfill \\ 3^{}\hfill \end{array}`$ | $`\begin{array}{c}2\hfill \\ 1\hfill \\ 1\hfill \end{array}`$ | $`\begin{array}{c}\hfill 1/3\\ \hfill 4/3\\ \hfill 2/3\end{array}`$ |
| Higgs | | | | | |
| $`\begin{array}{c}𝐇_\mathrm{𝟏}\\ 𝐇_\mathrm{𝟐}\end{array}`$ | Higgses $`\{\begin{array}{c}H_1\hfill \\ H_2\hfill \end{array}`$ | higgsinos $`\{\begin{array}{c}\stackrel{~}{H}_1\hfill \\ \stackrel{~}{H}_2\hfill \end{array}`$ | $`\begin{array}{c}1\hfill \\ 1\hfill \end{array}`$ | $`\begin{array}{c}2\hfill \\ 2\hfill \end{array}`$ | $`\begin{array}{c}\hfill 1\\ \hfill 1\end{array}`$ |
where $`a=1,2,\mathrm{},8`$ and $`k=1,2,3`$ are $`SU(3)`$ and $`SU(2)`$ indices, respectively, and $`i=1,2,3`$ is the generation index. Hereafter tilde denotes a superpartner of an ordinary particle.
Thus, the characteristic feature of any supersymmetric generalization of the SM is the presence of superpartners. If supersymmetry is exact, superpartners of ordinary particles should have the same masses and have to be observed. The absence of them at modern energies is believed to be explained by the fact that their masses are very heavy, that means that supersymmetry should be broken. Hence, if the energy of accelerators is high enough, the superpartners will be created.
The presence of an extra Higgs doublet in SUSY model is a novel feature of the theory. In the MSSM one has two doublets with the quantum numbers (1,2,-1) and (1,2,1), respectively:
$$H_1=\left(\begin{array}{c}H_1^0\\ H_1^{}\end{array}\right)=\left(\begin{array}{c}v_1+\frac{S_1+iP_1}{\sqrt{2}}\\ H_1^{}\end{array}\right),H_2=\left(\begin{array}{c}H_2^+\\ H_2^0\end{array}\right)=\left(\begin{array}{c}H_2^+\\ v_2+\frac{S_2+iP_2}{\sqrt{2}}\end{array}\right),$$
(3.1)
where $`v_i`$ are the vacuum expectation values of the neutral components.
Hence, one has 8=4+4=5+3 degrees of freedom. As in the case of the SM, 3 degrees of freedom can be gauged away, and one is left with 5 physical states compared to 1 state in the SM.
Thus, in the MSSM, as actually in any two Higgs doublet model, one has five physical Higgs bosons: two CP-even neutral, one CP-odd neutral and two charged. We consider the mass eigenstates below.
### 3.2 Largangian of the MSSM
The Lagrangian of the MSSM consists of two parts; the first part is SUSY generalization of the Standard Model, while the second one represents the SUSY breaking as mentioned above.
$$=_{SUSY}+_{Breaking},$$
(3.2)
where
$$_{SUSY}=_{Gauge}+_{Yukawa}$$
(3.3)
and
$`_{Gauge}`$ $`=`$ $`{\displaystyle \underset{SU(3),SU(2),U(1)}{}}{\displaystyle \frac{1}{4}}\left({\displaystyle d^2\theta TrW^\alpha W_\alpha }+{\displaystyle d^2\overline{\theta }Tr\overline{W}^{\dot{\alpha }}\overline{W}_{\dot{\alpha }}}\right)`$ (3.4)
$`+{\displaystyle \underset{Matter}{}}{\displaystyle d^2\theta d^2\overline{\theta }\mathrm{\Phi }_i^{}e^{g_3\widehat{V}_3+g_2\widehat{V}_2+g_1\widehat{V}_1}\mathrm{\Phi }_i},`$
$`_{Yukawa}`$ $`=`$ $`{\displaystyle d^2\theta (𝒲_R+𝒲_{NR})}+h.c.`$ (3.5)
The index $`R`$ in a superpotential refers to the so-called $`R`$-parity which adjusts a ”$`+`$” charge to all the ordinary particles and a ”$``$” charge to their superpartners . The first part of $`𝒲`$ is R-symmetric
$$W_R=ϵ_{ij}(y_{ab}^UQ_a^jU_b^cH_2^i+y_{ab}^DQ_a^jD_b^cH_1^i+y_{ab}^LL_a^jE_b^cH_1^i+\mu H_1^iH_2^j),$$
(3.6)
where $`i,j=1,2,3`$ are $`SU(2)`$ and $`a,b=1,2,3`$ are the generation indices; colour indices are suppressed. This part of the Largangian almost exactly repeats that of the SM except that the fields are now the superfields rather than the ordinary fields of the SM. The only difference is the last term which describes the Higgs mixing. It is absent in the SM since we have only one Higgs field there.
The second part is R-nonsymmetric
$`W_{NR}`$ $`=`$ $`ϵ_{ij}(\lambda _{abd}^LL_a^iL_b^jE_d^c+\lambda _{abd}^LL_a^iQ_b^jD_d^c+\mu _a^{}L_a^iH_2^j)`$ (3.7)
$`+`$ $`\lambda _{abd}^BU_a^cD_b^cD_d^c.`$
These terms are absent in the SM. The reason is very simple: one can not replace the superfields in eq.(3.7) by the ordinary fields like in eq.(3.6) because of the Lorentz invariance. These terms have the other property, they violate either lepton (the first line in eq.(3.7)) or baryon number (the second line). Since both effects are not observed in Nature, these terms must be suppressed or be excluded. In the minimal version of the MSSM these terms are not included, they are forbidden by R-parity conservation .
### 3.3 Soft SUSY breaking
To introduce supersymmetry breaking as required by the absence of superpartners at modern energies, one has to be careful not to spoil the cancellation of divergencies which allows to solve the hierarchy problem. This is achieved by spontaneous breaking of SUSY in the same way as spontaneous electroweak symmetry breaking. One introduces the field whose vacuum expectation value breaks supersymmetry. However, due to a special character of SUSY, this should be a superfield, whose auxiliary $`F`$ or $`D`$ components acquire non-zero v.e.v.’s. This leads to appearance of the so-called soft breaking terms. In the simplest version the soft supersymmetry breaking terms are :
$`_{Breaking}`$ $`=`$ $`m_0^2{\displaystyle \underset{i}{}}|\phi _i|^2+({\displaystyle \frac{1}{2}}m_{1/2}{\displaystyle \underset{\alpha }{}}\stackrel{~}{\lambda }_\alpha \stackrel{~}{\lambda }_\alpha `$
$`+`$ $`A[y_{ab}^U\stackrel{~}{Q}_a\stackrel{~}{U}_b^cH_2+y_{ab}^D\stackrel{~}{Q}_a\stackrel{~}{D}_b^cH_1+y_{ab}^L\stackrel{~}{L}_a\stackrel{~}{E}_b^cH_1]+B[\mu H_1H_2]+h.c.),`$
where we have suppressed $`SU(2)`$ indices. Here $`\phi _i`$ are all scalar fields, $`\stackrel{~}{\lambda }_\alpha `$ are the gaugino fields, $`\stackrel{~}{Q},\stackrel{~}{U},\stackrel{~}{D}`$ and $`\stackrel{~}{L},\stackrel{~}{E}`$ are the squark and slepton fields, respectively, and $`H_{1,2}`$ are the SU(2) doublet Higgs fields.
These terms are obtained via supergravity mechanism and are usually introduced at the GUT scale. We have assumed in eq.(3.3) the so-called universality of the soft terms, namely, we put all the spin 0 particle masses to be equal to the universal value $`m_0`$, all the spin 1/2 particle (gaugino) masses to be equal to $`m_{1/2}`$ and all the cubic and quadratic terms, proportional to $`A`$ and $`B`$, to repeat the structure of the Yukawa superpotential (3.6). This is an additional requirement motivated by the supergravity mechanism of SUSY breaking as mentioned earlier .
Universality is not a necessary requirement and one may consider non-universal soft terms as well. However, it will not change the qualitative picture presented below, so for simplicity in what follows we consider the universal boundary conditions.
It should be noted that supergravity induced universality of the soft terms is more likely to be valid at the Planck scale, rather than at the GUT one. This is because a natural scale for gravity is $`M_{Planck}`$, while $`M_{GUT}`$ is the scale for the gauge interactions. However, due to a small difference between these two scales, it is usually ignored in the first approximation resulting in minor uncertainties in the low-energy predictions .
The soft terms explicitly break supersymmetry. As will be shown later they lead to the mass spectrum of superpartners different from that of the ordinary particles. Remind that the masses of quarks and leptons remain zero until $`SU(2)`$ invariance is spontaneously broken.
### 3.4 Masses
With given values of $`m_0,m_{1/2},\mu ,Y_t,Y_b,Y_\tau ,A`$, and $`B`$ at the GUT scale, one can solve the corresponding RG equations thus linking the values at the GUT and electroweak scales. Substituting these parameters into the mass matrices one can predict the mass spectrum of superpartners .
#### 3.4.1 Gaugino-higgsino mass terms
The mass matrix for the gauginos, the superpartners of the gauge bosons, and for higgsinos, the superpartners of the Higgs bosons, is non-diagonal, thus leading to their mixing. The mass terms look like
$$_{GauginoHiggsono}=\frac{1}{2}M_3\overline{\lambda }_a\lambda _a\frac{1}{2}\overline{\chi }M^{(0)}\chi (\overline{\psi }M^{(c)}\psi +h.c.),$$
(3.9)
where $`\lambda _a,a=1,2,\mathrm{},8,`$ are the Majorana gluino fields and
$$\chi =\left(\begin{array}{c}\stackrel{~}{B}^0\\ \stackrel{~}{W}^3\\ \stackrel{~}{H}_1^0\\ \stackrel{~}{H}_2^0\end{array}\right),\psi =\left(\begin{array}{c}\stackrel{~}{W}^+\\ \stackrel{~}{H}^+\end{array}\right)$$
(3.10)
are, respectively, the Majorana neutralino and Dirac chargino fields. The neutralino mass matrix is:
$$M^{(0)}=\left(\begin{array}{cccc}M_1& 0& M_Z\mathrm{cos}\beta \mathrm{sin}_W& M_Z\mathrm{sin}\beta \mathrm{sin}_W\\ 0& M_2& M_Z\mathrm{cos}\beta \mathrm{cos}_W& M_Z\mathrm{sin}\beta \mathrm{cos}_W\\ M_Z\mathrm{cos}\beta \mathrm{sin}_W& M_Z\mathrm{cos}\beta \mathrm{cos}_W& 0& \mu \\ M_Z\mathrm{sin}\beta \mathrm{sin}_W& M_Z\mathrm{sin}\beta \mathrm{cos}_W& \mu & 0\end{array}\right).$$
(3.11)
The physical neutralino masses $`M_{\stackrel{~}{\chi }_i^0}`$ are obtained as eigenvalues of this matrix after diagonalization. For charginos one has:
$$M^{(c)}=\left(\begin{array}{cc}M_2& \sqrt{2}M_W\mathrm{sin}\beta \\ \sqrt{2}M_W\mathrm{cos}\beta & \mu \end{array}\right).$$
(3.12)
This matrix has two chargino eigenstates $`\stackrel{~}{\chi }_{1,2}^\pm `$ with mass eigenvalues
$$M_{1,2}^2=\frac{1}{2}\left[M_2^2+\mu ^2+2M_W^2\sqrt{(M_2^2\mu ^2)^2+4M_W^4\mathrm{cos}^22\beta +4M_W^2(M_2^2+\mu ^2+2M_2\mu \mathrm{sin}2\beta )}\right].$$
(3.13)
#### 3.4.2 Squark and slepton masses
The non-negligible Yukawa couplings cause a mixing between the electroweak eigenstates and the mass eigenstates of the third generation particles. The mixing matrices for the $`\stackrel{~}{m}_t^2,\stackrel{~}{m}_b^2`$ and $`\stackrel{~}{m}_\tau ^2`$ are:
$$\left(\begin{array}{cc}\stackrel{~}{m}_{tL}^2& m_t(A_t\mu \mathrm{cot}\beta )\\ m_t(A_t\mu \mathrm{cot}\beta )& \stackrel{~}{m}_{tR}^2\end{array}\right),$$
$$\left(\begin{array}{cc}\stackrel{~}{m}_{bL}^2& m_b(A_b\mu \mathrm{tan}\beta )\\ m_b(A_b\mu \mathrm{tan}\beta )& \stackrel{~}{m}_{bR}^2\end{array}\right),$$
$$\left(\begin{array}{cc}\stackrel{~}{m}_{\tau L}^2& m_\tau (A_\tau \mu \mathrm{tan}\beta )\\ m_\tau (A_\tau \mu \mathrm{tan}\beta )& \stackrel{~}{m}_{\tau R}^2\end{array}\right)$$
with
$`\stackrel{~}{m}_{tL}^2`$ $`=`$ $`\stackrel{~}{m}_Q^2+m_t^2+{\displaystyle \frac{1}{6}}(4M_W^2M_Z^2)\mathrm{cos}2\beta ,`$
$`\stackrel{~}{m}_{tR}^2`$ $`=`$ $`\stackrel{~}{m}_U^2+m_t^2{\displaystyle \frac{2}{3}}(M_W^2M_Z^2)\mathrm{cos}2\beta ,`$
$`\stackrel{~}{m}_{bL}^2`$ $`=`$ $`\stackrel{~}{m}_Q^2+m_b^2{\displaystyle \frac{1}{6}}(2M_W^2+M_Z^2)\mathrm{cos}2\beta ,`$
$`\stackrel{~}{m}_{bR}^2`$ $`=`$ $`\stackrel{~}{m}_D^2+m_b^2+{\displaystyle \frac{1}{3}}(M_W^2M_Z^2)\mathrm{cos}2\beta ,`$
$`\stackrel{~}{m}_{\tau L}^2`$ $`=`$ $`\stackrel{~}{m}_L^2+m_\tau ^2{\displaystyle \frac{1}{2}}(2M_W^2M_Z^2)\mathrm{cos}2\beta ,`$
$`\stackrel{~}{m}_{\tau R}^2`$ $`=`$ $`\stackrel{~}{m}_E^2+m_\tau ^2+(M_W^2M_Z^2)\mathrm{cos}2\beta `$
and the mass eigenstates are the eigenvalues of these mass matrices.
#### 3.4.3 The Higgs potential
As has been already mentioned, the Higgs potential in MSSM is totally defined by superpotential (and the soft terms). Due to the structure of $`𝒲`$ the Higgs self-interaction is given by the $`D`$-terms, while the $`F`$-terms contribute only to the mass matrix. The tree level potential is:
$`V_{tree}(H_1,H_2)`$ $`=`$ $`m_1^2|H_1|^2+m_2^2|H_2|^2m_3^2(H_1H_2+h.c.)`$ (3.14)
$`+`$ $`{\displaystyle \frac{g^2+g^{}_{}{}^{}2}{8}}(|H_1|^2|H_2|^2)^2+{\displaystyle \frac{g^2}{2}}|H_1^+H_2|^2,`$
where $`m_1^2=m_{H_1}^2+\mu ^2,m_2^2=m_{H_2}^2+\mu ^2`$. At the GUT scale $`m_1^2=m_2^2=m_0^2+\mu _0^2,m_3^2=B\mu _0`$. Notice, that the Higgs self-interaction coupling in eq.(3.14) is fixed and is defined by the gauge interactions as opposite to the SM.
The potential (3.14), in accordance with supersymmetry, is positively definite and stable. It has no non-trivial minimum different from zero. Indeed, let us write the minimization condition for the potential (3.14)
$`{\displaystyle \frac{1}{2}}{\displaystyle \frac{\delta V}{\delta H_1}}`$ $`=`$ $`m_1^2v_1m_3^2v_2+{\displaystyle \frac{g^2+g^2}{4}}(v_1^2v_2^2)v_1=0,`$ (3.15)
$`{\displaystyle \frac{1}{2}}{\displaystyle \frac{\delta V}{\delta H_2}}`$ $`=`$ $`m_2^2v_2m_3^2v_1+{\displaystyle \frac{g^2+g^2}{4}}(v_1^2v_2^2)v_2=0,`$ (3.16)
where we have introduced the notation
$$<H_1>v_1=v\mathrm{cos}\beta ,<H_2>v_2=v\mathrm{sin}\beta ,v^2=v_1^2+v_2^2,\mathrm{tan}\beta \frac{v_2}{v_1}.$$
Solution of eqs.(3.15),(3.16) can be expressed in terms of $`v^2`$ and $`\mathrm{sin}\beta `$:
$$v^2=\frac{4(m_1^2m_2^2\mathrm{tan}^2\beta )}{(g^2+g^2)(\mathrm{tan}^2\beta 1)},\mathrm{sin}2\beta =\frac{2m_3^2}{m_1^2+m_2^2}.$$
(3.17)
One can easily see from eq.(3.17) that if $`m_1^2=m_2^2=m_0^2+\mu _0^2`$, $`v^2`$ happens to be negative, i.e. the minimum does not exist. In fact, real positive solutions to eqs.(3.15),(3.16) exist only if the following conditions are satisfied :
$$m_1^2+m_2^2>2m_3^2,m_1^2m_2^2<m_3^4,$$
(3.18)
which is not the case at the GUT scale. This means that spontaneous breaking of the $`SU(2)`$ gauge invariance, which is needed in the SM to give masses for all the particles, does not take place in the MSSM.
This strong statement is valid, however, only at the GUT scale. Indeed, going down with energy the parameters of the potential (3.14) are renormalized. They become the “running” parameters with the energy scale dependence given by the RG equations. The running of the parameters leads to a remarkable phenomenon known as a radiative spontaneous symmetry breaking which we discuss below.
Provided conditions (3.18) are satisfied the mass matrices at the tree level are
CP-odd components $`P_1`$ and $`P_2`$ :
$$^{odd}=\frac{^2V}{P_iP_j}|_{H_i=v_i}=\left(\begin{array}{cc}\mathrm{tan}\beta & 1\\ 1& \mathrm{cot}\beta \end{array}\right)m_3^2,$$
(3.19)
CP-even neutral components $`S_1`$ and $`S_2`$:
$$^{even}=\frac{^2V}{S_iS_j}|_{H_i=v_i}=\left(\begin{array}{cc}\mathrm{tan}\beta & 1\\ 1& \mathrm{cot}\beta \end{array}\right)m_3^2+\left(\begin{array}{cc}\mathrm{cot}\beta & 1\\ 1& \mathrm{tan}\beta \end{array}\right)M_Z\mathrm{cos}\beta \mathrm{sin}\beta ,$$
(3.20)
Charged components $`H^{}`$ and $`H^+`$:
$$^{charged}=\frac{^2V}{H_i^+H_j^{}}|_{H_i=v_i}=\left(\begin{array}{cc}\mathrm{tan}\beta & 1\\ 1& \mathrm{cot}\beta \end{array}\right)(m_3^2+M_W\mathrm{cos}\beta \mathrm{sin}\beta ).$$
(3.21)
Diagonalising the mass matrices one gets the mass eigenstates :
$$\begin{array}{c}\{\begin{array}{cccc}G^0\hfill & =\hfill & \mathrm{cos}\beta P_1+\mathrm{sin}\beta P_2,\hfill & \hfill GoldstonebosonZ_0,\\ A\hfill & =\hfill & \mathrm{sin}\beta P_1+\mathrm{cos}\beta P_2,\hfill & \hfill NeutralCP=1Higgs,\end{array}\hfill \\ \\ \{\begin{array}{cccc}G^+\hfill & =\hfill & \mathrm{cos}\beta (H_1^{})^{}+\mathrm{sin}\beta H_2^+,\hfill & \hfill GoldstonebosonW^+,\\ H^+\hfill & =\hfill & \mathrm{sin}\beta (H_1^{})^{}+\mathrm{cos}\beta H_2^+,\hfill & \hfill ChargedHiggs,\end{array}\hfill \\ \\ \{\begin{array}{cccc}h\hfill & =\hfill & \mathrm{sin}\alpha S_1+\mathrm{cos}\alpha S_2,\hfill & \hfill SMHiggsbosonCP=1,\\ H\hfill & =\hfill & \mathrm{cos}\alpha S_1+\mathrm{sin}\alpha S_2,\hfill & \hfill ExtraheavyHiggsboson,\end{array}\hfill \end{array}$$
where the mixing angle $`\alpha `$ is given by
$$\mathrm{tan}2\alpha =\mathrm{tan}2\beta \left(\frac{m_A^2+M_Z^2}{m_A^2M_Z^2}\right).$$
The physical Higgs bosons acquire the following masses :
$`\text{CP-odd neutral Higgs}A:`$ $`m_A^2=m_1^2+m_2^2,`$
$`\text{Charge Higgses}H^\pm :`$ $`m_{H^\pm }^2=m_A^2+M_W^2,`$ (3.22)
$`\text{CP-even neutral Higgses}H,h:`$
$$m_{H,h}^2=\frac{1}{2}\left[m_A^2+M_Z^2\pm \sqrt{(m_A^2+M_Z^2)^24m_A^2M_Z^2\mathrm{cos}^22\beta }\right],$$
(3.23)
where as usual
$$M_W^2=\frac{g^2}{2}v^2,M_Z^2=\frac{g^2+g^2}{2}v^2.$$
This leads to the once celebrated SUSY mass relations:
$$\begin{array}{c}m_{H^\pm }M_W,\\ m_hm_AM_H,\\ m_hM_Z|\mathrm{cos}2\beta |M_Z,\\ m_h^2+m_H^2=m_A^2+M_Z^2.\end{array}$$
Thus, the lightest neutral Higgs boson happens to be lighter than $`Z`$ boson, that clearly distinguishes it from the SM one. Though we do not know the mass of the Higgs boson in the SM, there are several indirect constraints leading to the lower boundary of $`m_h^{SM}135`$ GeV . After including the radiative corrections the mass of the lightest Higgs boson in the MSSM, $`m_h`$, increases. We consider it in more detail below.
### 3.5 RG flow in the MSSM
If one compares the RG flow in the SM and the MSSM, one finds additional contributions from superpartners to the RG equations.
Consider the gauge couplings. In the SM the RG flow is given by eqs.(1.12). We have mentioned already in Sec.1 that it offers the unification pattern supporting the GUT hypothesis. However, if one looks at the curves more attentively, one finds that the situation is not that good. Indeed, let us consider the solution to the RG equations in more detail. The result is demonstrated in the left part of Fig.2, which shows the evolution of an inverse of the couplings as function of a logarithm of energy . In this presentation the evolution becomes a straight line in first order. The second order corrections are small and do not cause any visible deviation from a straight line. Fig.2 clearly demonstrates that within the SM the coupling constants unification at a single point is impossible. It is excluded by more than 8 standard deviations . This result means that the unification can only be obtained if new physics enters between the electroweak and the Planck scales.
In the MSSM the slopes of the RG evolution curves are modified. The coefficients $`b_i`$ in eq.(1.12) now are :
$$b_i=\left(\begin{array}{c}\hfill b_1\\ \hfill b_2\\ \hfill b_3\end{array}\right)=\left(\begin{array}{c}\hfill 0\\ \hfill 6\\ \hfill 9\end{array}\right)+N_{Fam}\left(\begin{array}{c}\hfill 2\\ \hfill 2\\ \hfill 2\end{array}\right)+N_{Higgs}\left(\begin{array}{c}\hfill 3/10\\ \hfill 1/2\\ \hfill 0\end{array}\right),$$
(3.24)
where use $`N_{Fam}=3`$ and $`N_{Higgs}=2`$, which corresponds to the MSSM.
It turns out that within the SUSY model perfect unification can be obtained if the SUSY masses are of the order of 1 TeV. This is shown on the right part of Fig.2; the SUSY particles are assumed to contribute effectively to the running of the coupling constants only for energies above the typical SUSY mass scale, which causes the change in the slope of the lines near 1 TeV. From the fit requiring unification one finds for the breakpoint $`M_{SUSY}`$ and the unification point $`M_{GUT}`$ :
$`M_{SUSY}`$ $`=`$ $`10^{3.4\pm 0.9\pm 0.4}GeV,`$
$`M_{GUT}`$ $`=`$ $`10^{15.8\pm 0.3\pm 0.1}GeV,`$ (3.25)
$`\alpha _{GUT}^1`$ $`=`$ $`(26.3\pm 1.9\pm 1.0).`$
The first error originates from the uncertainty in the coupling constant, while the second one is due to the uncertainty in the mass splittings between the SUSY particles. For SUSY models, the dimensional reduction $`\overline{DR}`$ scheme is used .
This unification of the gauge couplings was considered as the first “evidence” for supersymmetry, especially since $`M_{SUSY}`$ was found in the range preferred by the fine-tuning arguments.
It should be noted, that the unification of the three curves at a single point is not that trivial as it may seem from the existence of three free parameters ($`M_{SUSY},M_{GUT}`$ and $`\alpha _{GUT}`$). The reason is simple: when introducing new particles one influences all three curves simultaneously, thus giving rise to strong correlations between the slopes of the three lines. For example, adding new generations and/or new Higgs doublets never yield unification.
## 4 Renormalization of Softly Broken SUSY Theories
To find the RG flow for the soft terms one has to know how they are renormalized. Remarkably that the renormalizations in softly broken SUSY theories follow a simple pattern which is completely defined by an unbroken theory .
The main idea is that a softly broken supersymmetric gauge theory can be considered as a rigid SUSY theory imbeded into external space-time independent superfield, so that all couplings and masses become external superfields. The crucial statement is that the singular part of effective action depends on external superfield, but not on its derivatives, so that one can calculate it when the external field is a constant, i.e. in a rigid theory . This approach to a softly broken sypersymmetric theory allows one to use remarkable mathematical properties of $`N=1`$ SUSY theories such as non-renormalization theorems, cancellation of quadratic divergences, etc. The renormalization procedure in a softly broken SUSY gauge theory can be performed in the following way :
One takes renormalization constants of a rigid theory, calculated in some massless scheme, substitutes instead of the rigid couplings (gauge and Yukawa) their modified expressions, which depend on a Grassmannian variable, and expand over this variable.
This gives renormalization constants for the soft terms. Differentiating them with respect to a scale one can find corresponding renormalization group equations.
In fact as it has been shown in this procedure works at all stages. One can make the above mentioned substitution on the level of the renormalization constants, RG equations, solutions to these equations, approximate solutions, fixed points, finiteness conditions, etc. Expanding then over a Grassmannian variable one obtains corresponding expressions for the soft terms.
We demonstrate now how this procedure works in the MSSM. Using notation introduced above the modified couplings in the MSSM are ($`\eta =\theta ^2,\overline{\eta }=\overline{\theta }^2`$)
$`\stackrel{~}{\text{a}}_i`$ $`=`$ $`\text{a}_i(1+M_i\eta +\overline{M}_i\overline{\eta }+(M_i\overline{M}_i+\mathrm{\Sigma }_{\alpha _i})\eta \overline{\eta }),`$ (4.1)
$`\stackrel{~}{Y}_k`$ $`=`$ $`Y_k(1A_k\eta \overline{A}_k\overline{\eta }+(A_k\overline{A}_k+\mathrm{\Sigma }_k)\eta \overline{\eta }),`$ (4.2)
where $`M_i`$ are the gaugino masses, $`A_k`$ are the trilinear scalar couplings, $`\mathrm{\Sigma }_k`$ are the certain combinations of soft squark and slepton masses entering the Yukawa vertex and $`\mathrm{\Sigma }_{\alpha _i}`$ are the SUSY ghost soft terms
$$\mathrm{\Sigma }_t=\stackrel{~}{m}_{Q3}^2+\stackrel{~}{m}_{U3}^2+m_{H2}^2,\mathrm{\Sigma }_b=\stackrel{~}{m}_{Q3}^2+\stackrel{~}{m}_{D3}^2+m_{H1}^2,\mathrm{\Sigma }_\tau =\stackrel{~}{m}_{L3}^2+\stackrel{~}{m}_{E3}^2+m_{H1}^2,\mathrm{\Sigma }_{\alpha _i}=M_i^2+\stackrel{~}{m}_{gh_i}^2$$
and $`\stackrel{~}{m}_{gh}^2`$ is the soft scalar ghost mass, which is eliminated by solving the RG equation. In one-loop order $`\stackrel{~}{m}_{gh}^2=0`$.
To get now the RG equations for the soft terms one just has to take the corresponding RG equations for the rigid couplings and perform the Grassmannian expansion. The one-loop RG equations for the MSSM couplings are :
$`{\displaystyle \frac{d\text{a}_i}{dt}}`$ $`=`$ $`b_i\text{a}_i^2,`$
$`{\displaystyle \frac{dY_U}{dt}}`$ $`=`$ $`Y_L\left({\displaystyle \frac{16}{3}}\text{a}_3+3\text{a}_2+{\displaystyle \frac{13}{15}}\text{a}_16Y_UY_D\right),`$
$`{\displaystyle \frac{dY_D}{dt}}`$ $`=`$ $`Y_D\left({\displaystyle \frac{16}{3}}\text{a}_3+3\text{a}_2+{\displaystyle \frac{7}{15}}\text{a}_1Y_U6Y_DY_L\right),`$
$`{\displaystyle \frac{dY_L}{dt}}`$ $`=`$ $`Y_L\left(3\text{a}_2+{\displaystyle \frac{9}{5}}\text{a}_13Y_D4Y_L\right).`$ (4.3)
Performing the Grassmannian expansion one finds:
$`{\displaystyle \frac{dM_i}{dt}}`$ $`=`$ $`b_i\text{a}_iM_i.`$
$`{\displaystyle \frac{dA_U}{dt}}`$ $`=`$ $`{\displaystyle \frac{16}{3}}\text{a}_3M_3+3\text{a}_2M_2+{\displaystyle \frac{13}{15}}\text{a}_1M_1+6Y_UA_U+Y_DA_D,`$
$`{\displaystyle \frac{dA_D}{dt}}`$ $`=`$ $`{\displaystyle \frac{16}{3}}\text{a}_3M_3+3\text{a}_2M_2+{\displaystyle \frac{7}{15}}\text{a}_1M_1+6Y_DA_D+Y_UA_U+Y_LA_L,`$
$`{\displaystyle \frac{dA_L}{dt}}`$ $`=`$ $`3\text{a}_2M_2+{\displaystyle \frac{9}{5}}\text{a}_1M_1+3Y_DA_D+4Y_LA_L,`$
$`{\displaystyle \frac{dB}{dt}}`$ $`=`$ $`3\text{a}_2M_2+{\displaystyle \frac{3}{5}}\text{a}_1M_1+3Y_UA_U+3Y_DA_D+Y_LA_L.`$
$`{\displaystyle \frac{d\stackrel{~}{m}_Q^2}{dt}}`$ $`=`$ $`[({\displaystyle \frac{16}{3}}\text{a}_3M_3^2+3\text{a}_2M_2^2+{\displaystyle \frac{1}{15}}\text{a}_1M_1^2)Y_U(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_U^2+m_{H_2}^2+A_U^2)`$
$`Y_D(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_D^2+m_{H_1}^2+A_D^2)],`$
$`{\displaystyle \frac{d\stackrel{~}{m}_U^2}{dt}}`$ $`=`$ $`\left[({\displaystyle \frac{16}{3}}\text{a}_3M_3^2+{\displaystyle \frac{16}{15}}\text{a}_1M_1^2)2Y_U(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_U^2+m_{H_2}^2+A_U^2)\right],`$
$`{\displaystyle \frac{d\stackrel{~}{m}_D^2}{dt}}`$ $`=`$ $`\left[({\displaystyle \frac{16}{3}}\text{a}_3M_3^2+{\displaystyle \frac{4}{15}}\text{a}_1M_1^2)2Y_D(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_D^2+m_{H_1}^2+A_D^2)\right],`$
$`{\displaystyle \frac{d\stackrel{~}{m}_L^2}{dt}}`$ $`=`$ $`\left[3(\text{a}_2M_2^2+{\displaystyle \frac{1}{5}}\text{a}_1M_1^2)Y_L(\stackrel{~}{m}_L^2+\stackrel{~}{m}_E^2+m_{H_1}^2+A_L^2)\right],`$
$`{\displaystyle \frac{d\stackrel{~}{m}_E^2}{dt}}`$ $`=`$ $`\left[({\displaystyle \frac{12}{5}}\text{a}_1M_1^2)2Y_L(\stackrel{~}{m}_L^2+\stackrel{~}{m}_E^2+m_{H_1}^2+A_L^2)\right],`$
$`{\displaystyle \frac{d\mu ^2}{dt}}`$ $`=`$ $`\mu ^2\left[3(\text{a}_2+{\displaystyle \frac{1}{5}}\text{a}_1)(3Y_U+3Y_D+Y_L)\right],`$ (4.4)
$`{\displaystyle \frac{dm_{H_1}^2}{dt}}`$ $`=`$ $`[3(\text{a}_2M_2^2+{\displaystyle \frac{1}{5}}\text{a}_1M_1^2)3Y_D(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_D^2+m_{H_1}^2+A_D^2)`$
$`Y_L(\stackrel{~}{m}_L^2+\stackrel{~}{m}_E^2+m_{H_1}^2+A_L^2)],`$
$`{\displaystyle \frac{dm_{H_2}^2}{dt}}`$ $`=`$ $`\left[3(\text{a}_2M_2^2+{\displaystyle \frac{1}{5}}\text{a}_1M_1^2)3Y_U(\stackrel{~}{m}_Q^2+\stackrel{~}{m}_U^2+m_{H_2}^2+A_U^2)\right],`$
where we have already substituted the solution $`\stackrel{~}{m}_{gh}^2=0`$ in the one-loop order. (Note that to get the RG equation for the individual squark and slepton masses one needs to know the anomalous dimensions for the corresponding fields.)
## 5 RG Flow for the Soft Terms
Having all the RG equations, one can find now the RG flow for the soft terms. To see what happens at lower scales one has to run the RG equations for the mass parameters from GUT to the EW scale. Let us take some initial values of the soft masses at the GUT scale in the interval between $`10^2÷10^3`$ GeV consistent with SUSY scale suggested by unification of the gauge couplings (3.25). This leads to the following RG flow of the soft terms shown in Fig.3 (note that we perform the running of soft paramerters in the opposite direction, from GUT to EW scale)
One should mention the following general features common to any choice of initial conditions:
i) The gaugino masses follow the running of the gauge couplings and split at low energies. The gluino mass is running faster than the others and is usually the heaviest due to the strong interaction.
ii) The squark and slepton masses also split at low energies, the stops (and sbottoms) being the lightest due to relatively big Yukawa couplings of the third generation.
iii) The Higgs masses (or at least one of them) are running down very quickly and may even become negative.
To calculate the masses one has also to take into account the mixing between various states (see eqs.(3.11,3.12, 3.4.2-3.4.2).
### 5.1 Radiative Electroweak symmetry breaking
The running of the Higgs masses leads to the phenomenon known as a radiative electroweak symmetry breaking. By this we mean the following: At the GUT energy scale both the Higgs mass parameters $`m_1^2`$ and $`m_2^2`$ are positive and the Higgs potential has no non-trivial minima. However, when running down to the EW scale due to the radiative corrections they may change sign so that the potential develops a non-trivial minimum. At this minimum the electroweak symmetry happens to be spontaneously broken. Thus, contrary to the SM where one has to choose the negative sign of the Higgs mass squared ”by hand”, in the MSSM the effect of spontaneous symmetry breaking is triggered by the radiative corrections.
Indeed, one can see in Fig.3 is that $`m_2^2`$ (or both $`m_1^2`$ and $`m_2^2`$) decreases when going down from the GUT scale to the $`M_Z`$ scale and can even become negative. This is the effect of the large top (and bottom) Yukawa couplings in the RG equations. As a result, at some value of $`Q^2`$ the conditions (3.18) are satisfied, so that the non-trivial minimum appears. This triggers spontaneous breaking of the $`SU(2)`$ gauge invariance. The vacuum expectations of the Higgs fields acquire non-zero values and provide masses to the quarks, leptons, $`SU(2)`$ gauge bosons, and additional masses to their superpartners.
This way one obtains also the explanation of why the two scales are so much different. Due to the logariphmic running of the parameters one needs a long ”running time” to get $`m_1^2`$ to be negative when starting from a positive value of the order of $`M_{SUSY}`$ scale $`10^2÷10^3`$ GeV.
## 6 Infrared Quasi-fixed Points
Examining the RG equations for the Yukawa couplings one finds that they possess the infrared fixed points. This is a very typical behaviour for RG equations. In this section we give a short description of the infrared quasi-fixed points (IRQFP) in the MSSM. They play an important role in predictions of the mass spectrum.
As in the previous section we consider the RG flow in the direction from GUT to EW scale, the running parameter being $`t=\mathrm{log}M_{GUT}^2/Q^2`$. This corresponds to the opposite sign in RG eqs.(4.3,4.4).
### 6.1 Low tan$`\beta `$ regime
Consider first the low $`\mathrm{tan}\beta `$ regime. In this case, the only important Yukawa coupling is the top-quark one, all the others can be put equal to zero and the RG equations can be solved analytically
$$\text{a}_i(t)=\frac{\text{a}_0}{1+\text{a}_0b_it},Y_t(t)=\frac{Y_0E_t(t)}{1+6Y_0F_t(t)},$$
(6.1)
where
$`E_t(t)`$ $`=`$ $`{\displaystyle \underset{i}{}}(1+b_i\text{a}_0t)^{c_{ti}/b_i},c_{ti}=({\displaystyle \frac{13}{15}},3,{\displaystyle \frac{16}{3}}),`$
$`F_t(t)`$ $`=`$ $`{\displaystyle _0^t}E_t(t^{})𝑑t^{}.`$
In the IR regime solution (6.1) possesses a quasi-fixed point. Indeed taking the limit $`Y_0=Y_t(0)\mathrm{}`$ one can drop 1 in the denominator of eq.(6.1) and obtain the IRQFP
$$Y(t)Y_t^{FP}=\frac{E_t(t)}{6F_t(t)},$$
(6.2)
which is independent of the initial condition .
Though perturbation theory is not valid for $`Y_t>1`$, it does not prevent us from using the fixed point (6.2) since it attracts any solution with $`Y_0>\text{a}_0`$ or, numerically, for $`Y_0>0.1/4\pi `$ . Thus, for a wide range of initial values $`Y_t`$ is driven to the IR quasi-fixed point given by eq. (6.2) which numerically corresponds to $`y_t(M_Z)1.125`$. It is useful to introduce the ratio $`\rho _tY_t/\text{a}_3`$ since the strong coupling is the leading one in the IR regime. At the fixed point $`\rho _t(M_Z)0.84`$ and is approached in the IR regime when $`Q^2`$ decreases. The behaviour of $`\rho (t)`$ is shown in Fig.4 .
To get the solutions for the soft terms it is enough to perform the substitution $`\text{a}\stackrel{~}{\text{a}}`$ and $`Y\stackrel{~}{Y}`$ and expand over $`\eta `$ and $`\overline{\eta }`$. Expanding the gauge coupling in (6.1) up to $`\eta `$ one has (hereafter we assume $`M_{0i}=m_{1/2}`$)
$$M_i(t)=\frac{m_{1/2}}{1+b_i\text{a}_0t}.$$
(6.3)
Performing the same expansion for the Yukawa coupling one finds
$$A_t(t)=\frac{A_0}{1+6Y_0F_t}m_{1/2}\left(\frac{t}{E_t}\frac{dE_t}{dt}\frac{6Y_0}{1+6Y_0F_t}(tE_tF_t)\right).$$
(6.4)
To get the solution for the $`\mathrm{\Sigma }`$ term one has to make expansion over $`\eta `$ and $`\overline{\eta }`$. This leads to
$$\mathrm{\Sigma }_t(t)=\frac{\mathrm{\Sigma }_0A_0^2}{1+6Y_0F_t}+\frac{(A_0+m_{1/2}6Y_0(tE_tF_t))^2}{(1+6Y_0F_t)^2}+m_{1/2}^2\left[\frac{d}{dt}\left(\frac{t^2}{E_t}\frac{dE_t}{dt}\right)\frac{6Y_0}{1+6Y_0F_t}t^2\frac{dE_t}{dt}\right],$$
(6.5)
With analytic solutions (6.4,6.5) one can analyze asymptotics and, in particular, find the infrared quasi-fixed points which correspond to $`Y_0\mathrm{}`$
$`A_t^{FP}`$ $`=`$ $`m_{1/2}\left({\displaystyle \frac{t}{E_t}}{\displaystyle \frac{dE_t}{dt}}{\displaystyle \frac{tE_tF_t}{F_t}}\right),`$ (6.6)
$`\mathrm{\Sigma }_t^{FP}`$ $`=`$ $`m_{1/2}^2\left[\left({\displaystyle \frac{tE_tF_t}{F_t}}\right)^2+{\displaystyle \frac{d}{dt}}\left({\displaystyle \frac{t^2}{E_t}}{\displaystyle \frac{dE_t}{dt}}\right){\displaystyle \frac{t^2}{F_t}}{\displaystyle \frac{dE_t}{dt}}\right].`$ (6.7)
The FP solutions (6.6,6.7) can be directly obtained from a fixed point for the rigid Yukawa coupling (6.2) by Grassmannian expansion. This explains, in particular, why fixed point solutions for the soft couplings exist if they exist for the rigid ones and with the same stability properties .
The behaviour of $`\rho _A=A_t/M_3`$ as a function of $`\text{a}_3`$ for a fixed ratio $`Y_0/\text{a}_0=5`$ is shown in Fig. 4b. One can observe the strong attraction to the IR stable quasi-fixed point $`\rho _A0.62`$ .
One can also write down solutions for the individual masses. This can be obtained using the Grassmannian expansion of solutions for the corresponding superfield propagators. For the first two generations one has
$`\stackrel{~}{m}_{Q_L}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}({\displaystyle \frac{16}{3}}f_3+3f_2+{\displaystyle \frac{1}{15}}f_1),`$
$`\stackrel{~}{m}_{U_R}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}({\displaystyle \frac{16}{3}}f_3+{\displaystyle \frac{16}{15}}f_1),`$
$`\stackrel{~}{m}_{D_R}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}({\displaystyle \frac{16}{3}}f_3+{\displaystyle \frac{4}{15}}f_1),`$
$`m_{H_1}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}(3f_2+{\displaystyle \frac{3}{5}}f_1),`$
$`\stackrel{~}{m}_{L_L}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}(3f_2+{\displaystyle \frac{3}{15}}f_1),`$
$`\stackrel{~}{m}_{E_R}^2`$ $`=`$ $`m_0^2+{\displaystyle \frac{m_{1/2}^2}{2}}({\displaystyle \frac{12}{5}}f_1),`$
where
$$f_i=\frac{1}{b_i}\left(1\frac{1}{(1+b_i\text{a}_0t)^2}\right).$$
The third generation masses get the contribution from the top Yukawa coupling
$`\stackrel{~}{m}_{b_R}^2`$ $`=`$ $`\stackrel{~}{m}_{D_R}^2,`$
$`\stackrel{~}{m}_{b_L}^2`$ $`=`$ $`\stackrel{~}{m}_{Q_L}^2+\mathrm{\Delta }/6,`$
$`\stackrel{~}{m}_{t_R}^2`$ $`=`$ $`\stackrel{~}{m}_{U_R}^2+\mathrm{\Delta }/3,`$
$`\stackrel{~}{m}_{t_L}^2`$ $`=`$ $`\stackrel{~}{m}_{Q_L}^2+\mathrm{\Delta }/6,`$
$`m_{H_2}^2`$ $`=`$ $`m_{H_1}^2+\mathrm{\Delta }/2,`$
where $`\mathrm{\Delta }`$ is related to $`\mathrm{\Sigma }_t`$ (6.5) by
$$\mathrm{\Delta }=\mathrm{\Sigma }_t\mathrm{\Sigma }_0m_{1/2}^2\frac{d}{dt}\left(\frac{t^2}{E_t}\frac{dE_t}{dt}\right).$$
There is no obvious infrared attractive fixed point for $`m_{H_1}^2`$. However, one can take the linear combination $`m_+^2=m_{H_1}^2+2m_{H_2}^2`$ which together with $`m_{H_2}^2`$ shows the IR fixed point behaviour in the limit $`Y_0\mathrm{}`$.
$$\frac{m_+^{2FP}}{M_3^2}0.73,\frac{m_{H_2}^{2FP}}{M_3^2}0.12\left(\frac{1}{2}\frac{m_0^2}{m_{1/2}^2}+3.4\right).$$
(6.8)
In eq.(6.8) one has only weak dependence on the ratio $`m_0^2/m_{1/2}^2`$. One can find the IR quasi-fixed point $`m_{H_2}^2/M_3^20.40`$ which corresponds to $`m_0^2/m_{1/2}^2=0`$. As for the combination $`m_+^2`$, the dependence on initial conditions disappears completely, as it follows from (6.8). The situation is illustrated in Fig.5 .
Consider now the squark masses. In the limit $`Y_t\mathrm{}`$ these solutions are driven to the IRQFP’s
$$\frac{\stackrel{~}{m}_U^{2FP}}{M_3^2}0.48,\frac{\stackrel{~}{m}_Q^{2FP}}{M_3^2}0.12\left(\frac{1}{2}\frac{m_0^2}{m_{1/2}^2}+5.8\right)$$
(6.9)
As it follows from eq.(6.9), the solution for $`\stackrel{~}{m}_U^2/M_3^2`$ becomes independent of the initial conditions $`m_0/m_{1/2}`$ and $`A_0/m_{1/2}`$, when the top-quark Yukawa coupling is initially large enough. As a result, the solutions of RGE’s are driven to the fixed point (6.9) for a wide range of $`m_0^2/m_{1/2}^2`$ (Fig.6b). As for $`m_Q^2`$, the dependence on initial conditions does not completely disappear; however, it is rather weak like in the case of $`m_{H_2}^2`$ and approaches the value $`\stackrel{~}{m}_Q^2/M_3^20.69`$ (Fig.6a) .
The bilinear SUSY breaking parameter B does not exhibit a fixed-point behaviour in the limit $`Y_0\text{a}_0`$. To see this, consider the solution in the aforementioned limit. One has
$$\frac{B}{M_3}0.35\left(\frac{B_0}{m_{1/2}}\frac{1}{2}\frac{A_0}{m_{1/2}}0.8\right).$$
(6.10)
It is clear that neither $`B_0/m_{1/2}`$ nor $`A_0/m_{1/2}`$ may be neglected. As a consequence, no fixed point behaviour for the ratio $`B/M_3`$ is observed.
We have considered the one-loop RGE’s. It is interesting to see, however, how our results are modified when two-loop RGE’s are used. For comparison we present the two-loop IRQFP values together with our one-loop results in Table 1.
As one can see from this table, the difference between the one-loop and two-loop results is negligible for $`A_t/M_3`$, $`\stackrel{~}{m}_U^2/M_3^2`$ and $`m_+^2/M_3^2`$. As for $`m_{H_2}^2/M_3^2`$ and $`\stackrel{~}{m}_Q^2/M_3^2`$, the two-loop corrections to the fixed points are about two times as small as deviations from them. As it was mentioned above, such corrections have a negligible impact on our main results.
### 6.2 Large tan$`\beta `$ regime
We now give a short description of the infrared behaviour of the RGE’s in the MSSM for the large $`\mathrm{tan}\beta `$ regime. While with a single Yukawa coupling the analytical solution to the one-loop RG equations has been known for long, for increasing number of Yukawa couplings it has been obtained quite recently in the form that allows iterative representation.
One can write down the one-loop RG equations (4.3) as
$`{\displaystyle \frac{d\text{a}_i}{dt}}`$ $`=`$ $`b_i\text{a}_i^2,`$ (6.11)
$`{\displaystyle \frac{dY_k}{dt}}`$ $`=`$ $`Y_k({\displaystyle \underset{i}{}}c_{ki}\text{a}_i{\displaystyle \underset{l}{}}a_{kl}Y_l),`$ (6.12)
where
$`b_i`$ $`=`$ $`\{33/5,1,3\},`$
$`c_{ti}`$ $`=`$ $`\{13/15,3,16/3\},c_{bi}=\{7/15,3,16/3\},c_{\tau i}=\{9/5,3,0\},`$
$`a_{tl}`$ $`=`$ $`\{6,1,0\},a_{bl}=\{1,6,1\},a_{\tau l}=\{0,3,4\}.`$
Then the general solution to eqs.(6.11,6.12) can be written in the form
$`\text{a}_i`$ $`=`$ $`{\displaystyle \frac{\text{a}_0}{1+b_i\text{a}_0t}},`$ (6.13)
$`Y_k`$ $`=`$ $`{\displaystyle \frac{Y_k^0u_k}{1+a_{kk}Y_k^0_0^tu_k}},`$ (6.14)
where the functions $`u_k`$ obey the integral system of equations
$$u_t=\frac{E_t}{(1+6Y_b^0_0^tu_b)^{1/6}},u_b=\frac{E_b}{(1+6Y_t^0_0^tu_t)^{1/6}(1+4Y_\tau ^0_0^tu_\tau )^{1/4}},u_\tau =\frac{E_\tau }{(1+6Y_b^0_0^tu_b)^{1/2}},$$
(6.15)
and the functions $`E_k`$ are given by
$$E_k=\underset{i=1}{\overset{3}{}}(1+b_i\text{a}_0t)^{c_{ki}/b_i}.$$
(6.16)
Let us stress that eqs.(6.13,6.14) give the exact solution to eqs.(6.11,6.12), while the $`u_k`$’s in eqs.(6.15), although solved formally in terms of the $`E_k`$’s and $`Y_k^0`$’s as continued integrated fractions, should in practice be solved iteratively.
Let us now perform the substitution (4.1,4.2) in (6.13-6.15) and expand over $`\eta `$ and $`\overline{\eta }`$. Then the linear term in $`\eta `$ will give us the solution for $`M_i`$ and $`A_k`$ and the $`\eta \overline{\eta }`$ terms the ones for $`\mathrm{\Sigma }_k`$. The resulting exact solutions look similar to those for the rigid couplings (6.136.15)
$`M_i`$ $`=`$ $`{\displaystyle \frac{M_i^0}{1+b_i\text{a}_i^0t}},`$ (6.17)
$`A_k`$ $`=`$ $`e_k+{\displaystyle \frac{A_k^0/Y_k^0+a_{kk}u_ke_k}{1/Y_k^0+a_{kk}u_k}},`$ (6.18)
$`\mathrm{\Sigma }_k`$ $`=`$ $`\xi _k+A_k^2+2e_kA_k{\displaystyle \frac{(A_k^0)^2/Y_k^0\mathrm{\Sigma }_k^0/Y_k^0+a_{kk}u_k\xi _k}{1/Y_k^0+a_{kk}u_k}},`$ (6.19)
where the new functions $`e_k`$ and $`\xi _k`$ have been introduced which obey the iteration equations
$`e_t`$ $`=`$ $`{\displaystyle \frac{1}{E_t}}{\displaystyle \frac{d\stackrel{~}{E}_t}{d\eta }}+{\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}},`$
$`e_b`$ $`=`$ $`{\displaystyle \frac{1}{E_b}}{\displaystyle \frac{d\stackrel{~}{E}_b}{d\eta }}+{\displaystyle \frac{A_t^0u_tu_te_t}{1/Y_t^0+6u_t}}+{\displaystyle \frac{A_\tau ^0u_\tau u_\tau e_\tau }{1/Y_\tau ^0+4u_\tau }},`$
$`e_\tau `$ $`=`$ $`{\displaystyle \frac{1}{E_\tau }}{\displaystyle \frac{d\stackrel{~}{E}_\tau }{d\eta }}+3{\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}},`$
$`\xi _t`$ $`=`$ $`{\displaystyle \frac{1}{E_t}}{\displaystyle \frac{d^2\stackrel{~}{E}_t}{d\eta d\overline{\eta }}}+2{\displaystyle \frac{1}{E_t}}{\displaystyle \frac{d\stackrel{~}{E}_t}{d\eta }}{\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}}+7\left({\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}}\right)^2`$
$`\left((\mathrm{\Sigma }_b^0+(A_b^0)^2){\displaystyle u_b}2A_b^0{\displaystyle u_be_b}+{\displaystyle u_b\xi _b}\right)/\left({\displaystyle \frac{1}{Y_b^0}}+6{\displaystyle u_b}\right),`$
$`\xi _b`$ $`=`$ $`{\displaystyle \frac{1}{E_b}}{\displaystyle \frac{d^2\stackrel{~}{E}_b}{d\eta d\overline{\eta }}}+2{\displaystyle \frac{1}{E_b}}{\displaystyle \frac{d\stackrel{~}{E}_b}{d\eta }}\left[{\displaystyle \frac{A_t^0u_tu_te_t}{1/Y_t^0+6u_t}}+{\displaystyle \frac{A_\tau ^0u_\tau u_\tau e_\tau }{1/Y_\tau ^0+4u_\tau }}\right]`$
$`+7\left({\displaystyle \frac{A_t^0u_tu_te_t}{1/Y_t^0+6u_t}}\right)^2+5\left({\displaystyle \frac{A_\tau ^0u_\tau u_\tau e_\tau }{1/Y_\tau ^0+4u_\tau }}\right)^2`$
$`+2\left({\displaystyle \frac{A_t^0u_tu_te_t}{1/Y_t^0+6u_t}}\right)\left({\displaystyle \frac{A_\tau ^0u_\tau u_\tau e_\tau }{1/Y_\tau ^0+4u_\tau }}\right)`$
$`\left((\mathrm{\Sigma }_t^0+(A_t^0)^2){\displaystyle u_t}2A_t^0{\displaystyle u_te_t}+{\displaystyle u_t\xi _t}\right)/\left({\displaystyle \frac{1}{Y_t^0}}+6{\displaystyle u_t}\right)`$
$`\left((\mathrm{\Sigma }_\tau ^0+(A_\tau ^0)^2){\displaystyle u_\tau }2A_\tau ^0{\displaystyle u_\tau e_\tau }+{\displaystyle u_\tau \xi _\tau }\right)/\left({\displaystyle \frac{1}{Y_\tau ^0}}+4{\displaystyle u_\tau }\right),`$
$`\xi _\tau `$ $`=`$ $`{\displaystyle \frac{1}{E_\tau }}{\displaystyle \frac{d^2\stackrel{~}{E}_\tau }{d\eta d\overline{\eta }}}+6{\displaystyle \frac{1}{E_\tau }}{\displaystyle \frac{d\stackrel{~}{E}_\tau }{d\eta }}{\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}}+27\left({\displaystyle \frac{A_b^0u_bu_be_b}{1/Y_b^0+6u_b}}\right)^2`$ (6.20)
$`3\left((\mathrm{\Sigma }_b^0+(A_b^0)^2){\displaystyle u_b}2A_b^0{\displaystyle u_be_b}+{\displaystyle u_b\xi _b}\right)/\left({\displaystyle \frac{1}{Y_b^0}}+6{\displaystyle u_b}\right).`$
Here the variations of $`\stackrel{~}{E}_k`$ should be taken at $`\eta =\overline{\eta }=0`$. When solving eqs.(6.15) and (6.20) in the $`n`$-th iteration one has to substitute in the r.h.s. the $`(n1)`$-th iterative solution for all the corresponding functions.
The solutions for the individual soft masses are linearly expressed through $`\mathrm{\Sigma }`$’s
$`\stackrel{~}{m}_{Q_3}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{128f_3+87f_211f_1}{122}}+{\displaystyle \frac{17(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)+20(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)5(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`\stackrel{~}{m}_{U_3}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{144f_3108f_2+144/5f_1}{122}}+{\displaystyle \frac{42(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)8(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)+2(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`\stackrel{~}{m}_{D_3}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{112f_384f_2+112/5f_1}{122}}+{\displaystyle \frac{8(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)+48(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)12(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`m_{H_1}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{240f_33f_257/5f_1}{122}}+{\displaystyle \frac{9(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)+54(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)+17(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`m_{H_2}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{272f_3+21f_289/5f_1}{122}}+{\displaystyle \frac{63(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)12(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)+3(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`\stackrel{~}{m}_{L_3}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{80f_3+123f_2103/5f_1}{122}}+{\displaystyle \frac{3(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)18(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)+35(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}},`$
$`\stackrel{~}{m}_{E_3}^2`$ $`=`$ $`m_0^2+m_{1/2}^2{\displaystyle \frac{160f_3120f_2+32f_1}{122}}+{\displaystyle \frac{6(\mathrm{\Sigma }_t\mathrm{\Sigma }_t^0)36(\mathrm{\Sigma }_b\mathrm{\Sigma }_b^0)+70(\mathrm{\Sigma }_\tau \mathrm{\Sigma }_t^0)}{122}}.`$
The solutions (6.136.15, 6.176.20) have a nice property since they contain explicit dependence on initial conditions and one can trace this dependence in the final results. This is of special importance for the non-universal case since one can see which of the parameters is essential and which is washed out during the evolution. In particular the solution for the Yukawa couplings exhibit the fixed point behaviour when the initial values are large enough. More precisely, in the regime $`Y_t^0,Y_b^0,Y_\tau ^0\mathrm{}`$ with fixed finite ratios $`Y_t^0/Y_b^0=r_1,Y_b^0/Y_\tau ^0=r_2`$, it is legitimate to drop $`1`$ in the denominators of eqs.(6.14, 6.15) in which case the exact Yukawa solutions go to the IRQFP defined by
$$Y_k^{FP}=\frac{u_k^{FP}}{a_{kk}u_k^{FP}}$$
(6.21)
with
$$u_t^{FP}=\frac{E_t}{(u_b^{FP})^{1/6}},u_b^{FP}=\frac{E_b}{(u_t^{FP})^{1/6}(u_\tau ^{FP})^{1/4}},u_\tau ^{FP}=\frac{E_\tau }{(u_b^{FP})^{1/2}}$$
(6.22)
extending the IRQFP (6.2) to three Yukawa couplings. What is worth stressing here is that both the dependence on the initial condition for each Yukawa as well as the effect of Yukawa non-unification, ($`r_1,r_2`$), have completely dropped out of the runnings.
This in turn leads to the IRQFPs for the soft terms. Disappearance of $`Y_k^0`$ in the FP solution naturally leads to the disappearance of $`A_k^0`$ and $`\mathrm{\Sigma }_k^0`$ in the soft term fixed points.
Below we present the result of numerical analysis. We begin with Yukawa couplings and assume the equality of the Yukawa couplings of the third generation at the GUT scale: $`Y_t(M_{GUT})=Y_b(M_{GUT})=Y_\tau (M_{GUT})`$.
In Figs. 4a,b,c the numerical solutions of the RGE’s are shown for a wide range of initial values of $`\rho _t(M_{GUT})=\rho _b(M_{GUT})=\rho _\tau (M_{GUT})`$ from the interval $`<0.2,5>`$, where $`\rho _i=Y_i/\text{a}_3`$. One can clearly see the IRQFP type behaviour when the parameter $`\rho _i`$ at the GUT scale is big enough .
We have found the following values of the Yukawa couplings $`y_i`$ at the $`M_Z`$ scale
$$y_t<0.787,1.048>,y_b<0.758,0.98>,y_{tau}<0.375,0.619>.$$
Comparing $`y_t`$ and $`y_b`$ one can see that the ratio belongs to a very narrow interval $`y_t/y_b<1.039,1.069>`$.
Now we proceed with the discussion of RGE’s for trilinear scalar couplings, $`A_i,i=(t,b,\tau )`$. The results are shown in Figs. 4d,e for the following quantities $`\rho _{A_i}=A_i/M_3,i=(t,b)`$ for different initial values at the GUT scale and for $`\rho _i(M_{GUT})=5`$. One can see the strong attraction to the fixed points .
The question of stability of these IRQFPs becomes important for further consideration. Analyzing their stability under the change of the initial conditions for $`\rho _i(M_{GUT})`$ one finds remarkable stability, which allows to use them as fixed parameters at the $`M_Z`$ scale. In Fig. 7f a particular example of stability of IRQFP for $`A_t`$ is shown. As a result one has the following IRQFP values for the parameters $`\rho _{A_i}`$:
$$\rho _{A_t}0.619,\rho _{A_b}0.658,\rho _{A_\tau }0.090.$$
The last step in the investigation of the RGE’s is the calculation of the soft mass parameters.
As one can see from Figs. 8a,b there exist IRQFP’s
$$m_{H_1}^2/M_3^20.306,m_{H_2}^2/M_3^20.339.$$
The numbers correspond to the initial condition $`m_0^2/m_{1/2}^2=0`$. Later the initial values for the ratio $`m_0^2/m_{1/2}^2`$ belonging to the following interval $`m_0^2/m_{1/2}^2<0,2>`$ are considered.
In Figs. 8c,d,e the infrared behaviour of the soft SUSY breaking squark masses is shown. One can immediately see that all masses have IRQFPs which are used in the next section to find the mass spectrum. For further analysis only the squark masses are important. As for sleptons they also have an attractive infrared behaviour but it does not influence the mass spectrum of the Higgs bosons in which we are interested in below and we do not show them explicitly.
Numerical values of the ratios are the following :
$$m_Q^2/M_3^20.58,m_U^2/M_3^20.52,m_D^2/M_3^20.53,$$
obtained for $`m_0^2/m_{1/2}^2=0`$. One again finds a very week dependence on initial values of the Yukawa couplings.
The behaviour of the bilinear SUSY breaking parameter $`B`$ is the same as in low $`\mathrm{tan}\beta `$ case. The ratio $`B/M_3`$ does not exhibit the infrared quasi fixed point behaviour.
Thus, one can see that solutions of RGE’s for all MSSM SUSY breaking parameters (the only exception is the parameter $`B`$) are driven to the infrared attractive fixed points if the Yukawa couplings at the GUT scale are large enough.
Our analysis is constrained by the one-loop RG equations. The difference between one-loop and two-loop IRQFPs is similar to the low $`\mathrm{tan}\beta `$ case (see Table 1) and is less than 10 per cent. At the same time the deviations from the IRQFPs obtained by one-loop RGEs are also of the same order which defines the accuracy of our predictions. The only place where it really matters is the prediction of the lightest Higgs boson mass where all the proper corrections are taken into account.
## 7 Higgs Boson Mass Prediction in the SM and MSSM
### 7.1 The Higgs boson mass in the SM
The last unobserved particle from the Standard Model is the Higgs boson. Its discovery would allow one to complete the SM paradigm and confirm the mechanism of spontaneous symmetry breaking. On the contrary, the absence of the Higgs boson would awake doubts about the whole picture and would require new concepts.
Experimental limits on the Higgs boson mass come from a direct search at LEP II and Tevatron and from indirect fits of electroweak precision data, first of all from the radiative corrections to the W and top quark masses. A combined fit of modern experimental data gives
$$m_h=78_{47}^{+86}\mathrm{GeV},$$
(7.1)
which at the 95% confidence level leads to the upper bound of 260 GeV (see Fig.9). At the same time, recent direct searches at LEP II for the c.m. energy of 189 GeV give the lower limit of almost 95 GeV<sup>2</sup><sup>2</sup>2The last run of LEP II at 200 GeV c.m. energy has increased this bound up to 103 GeV . From theoretical point of view low Higgs mass could be a hint for physics beyond the SM, in particular for the supersymmetric extension of the SM.
Within the Standard Model the value of the Higgs mass $`m_h`$ is not predicted. However, one can get the bounds on the Higgs mass . They follow from the behaviour of the quartic coupling which obeys the RG equation (1.14).
Since the quartic coupling grows with rising energy indefinitely, an upper bound on $`m_h`$ follows from the requirement that the theory be valid up to the scale $`M_{Planck}`$ or up to a given cut-off scale $`\mathrm{\Lambda }`$ below $`M_{Planck}`$ . The scale $`\mathrm{\Lambda }`$ could be identified with the scale at which a Landau pole develops. The upper bound on $`m_h`$ depends mildly on the top-quark mass through the impact of the top-quark Yukawa coupling on the running of the quartic coupling $`\lambda `$.
On the other hand, the requirement of vacuum stability in the SM (positivity of $`\lambda `$) imposes a lower bound on the Higgs boson mass, which crucially depends on the top-quark mass as well as on the cut-off $`\mathrm{\Lambda }`$ . Again, the dependence of this lower bound on $`m_t`$ is due to the effect of the top-quark Yukawa coupling on the quartic coupling in eq.(1.14), which drives $`\lambda `$ to negative values at large scales, thus destabilizing the standard electroweak vacuum.
From the point of view of LEP and Tevatron physics, the upper bound on the SM Higgs boson mass does not pose any relevant restriction. The lower bound on $`m_h`$, instead, is particularly important in view of search for the Higgs boson at LEPII and Tevatron. For $`m_t174`$ GeV and $`\alpha _s(M_Z)=0.118`$ the results at $`\mathrm{\Lambda }=10^{19}`$ GeV or at $`\mathrm{\Lambda }=1`$ TeV can be given by the approximate formulae
$`m_h`$ $`>`$ $`135+2.1[m_t174]4.5\left[{\displaystyle \frac{\alpha _s(M_Z)0.118}{0.006}}\right],\mathrm{\Lambda }=10^{19}GeV,`$ (7.2)
$`m_h`$ $`>`$ $`72+0.9[m_t174]1.0\left[{\displaystyle \frac{\alpha _s(M_Z)0.118}{0.006}}\right],\mathrm{\Lambda }=1TeV,`$ (7.3)
where the masses are in units of GeV.
Fig.10 shows the perturbativity and stability bounds on the Higgs boson mass of the SM for different values of the cut-off $`\mathrm{\Lambda }`$ at which new physics is expected.
We see from Fig.10 and eqs.(7.2,7.3) that indeed for $`m_t174`$ GeV the discovery of a Higgs particle at LEPII would imply that the Standard Model breaks down at a scale $`\mathrm{\Lambda }`$ well below $`M_{GUT}`$ or $`M_{Planck}`$, smaller for lighter Higgs. Actually, if the SM is valid up to $`\mathrm{\Lambda }M_{GUT}`$ or $`M_{Planck}`$, for $`m_t174`$ GeV only a small range of values is allowed: $`134<m_h<200`$ GeV. For $`m_t`$ = 174 GeV and $`m_h<100`$ GeV \[i.e. in the LEPII range\] new physics should appear below the scale $`\mathrm{\Lambda }`$ a few to 100 TeV. The dependence on the top-quark mass however is noticeable. A lower value, $`m_t`$ 170 GeV, would relax the previous requirement to $`\mathrm{\Lambda }10^3`$ TeV, while a heavier value $`m_t`$ 180 GeV would demand new physics at an energy scale as low as 10 TeV.
### 7.2 The Higgs boson mass in the MSSM
It has been already mentioned that in the MSSM the mass of the lightest Higgs boson is predicted to be less than the $`Z`$-boson mass. This is, however, the tree level result and the masses acquire the radiative corrections.
With account of the radiative corrections the effective Higgs bosons potential is
$$V_{Higgs}^{eff}=V_{tree}+\mathrm{\Delta }V,$$
(7.4)
where $`V_{tree}`$ is given by eq.(3.14) and in one loop order
$$\mathrm{\Delta }V_{1loop}=\underset{k}{}\frac{1}{64\pi ^2}(1)^{J_k}(2J_k+1)c_km_k^4\left(\mathrm{log}\frac{m_k^2}{Q^2}\frac{3}{2}\right).$$
(7.5)
Here the sum is taken over all the particles in the loop, $`J_k`$ is the spin and $`m_k`$ is the field dependent mass of a particle at the scale $`Q`$. These radiative corrections vanish when supersymmetry is not broken and are positive in softly broken case. The leading contribution comes from (s)top loops
$$\mathrm{\Delta }V_{1loop}^{stop}=\frac{3}{32\pi ^2}\left[\stackrel{~}{m}_{t_1}^4(\mathrm{log}\frac{\stackrel{~}{m}_{t_1}^2}{Q^2}\frac{3}{2})+\stackrel{~}{m}_{t_2}^4(\mathrm{log}\frac{\stackrel{~}{m}_{t_2}^2}{Q^2}\frac{3}{2})2m_t^4(\mathrm{log}\frac{m_t^2}{Q^2}\frac{3}{2})\right].$$
(7.6)
Contributions from the other particles are much smaller .
These corrections lead to the following modification of the tree level relation for the lightest Higgs mass
$$m_h^2M_Z^2\mathrm{cos}^22\beta +\frac{3g^2m_t^4}{16\pi ^2M_W^2}\mathrm{log}\frac{\stackrel{~}{m}_{t_1}^2\stackrel{~}{m}_{t_2}^2}{m_t^4}.$$
(7.7)
One finds that the one loop correction is positive and increases the mass value. Two loop corrections have the opposite effect but are smaller and result in slightly lower value of the Higgs mass.
To find out numerical values of these corrections one has to determine the masses of all superpartners. This means that one has to know the initial conditions for the soft papameters. Fortunately due to the IRQFP solutions the dependence on initial conditions may disappear at low energies. This allows one to reduce the number of unknown parameters and make predictions.
Due to extreme importance of the Higgs mass predictions for experimental searches this problem has been the subject of intense investigation. There is a considerable amount of papers devoted to this topic (see e.g. ). And though the initial assumptions and the strategy are different, the general conclusions are very similar.
In what follows we accept the strategy advocated in Refs. : As input parameters one takes the known values of the top-quark, bottom-quark and $`\tau `$-lepton masses, the experimental values of the gauge couplings, and the mass of the Z-boson . To reduce the arbitrariness of the soft terms we use the fixed-point values for the Yukawa couplings and SUSY breaking parameters. The value of $`\mathrm{tan}\beta `$ is determined from the relations between the running quark masses and the Higgs v.e.v.’s in the MSSM
$`m_t`$ $`=`$ $`y_tv\mathrm{sin}\beta ,`$ (7.8)
$`m_b`$ $`=`$ $`y_bv\mathrm{cos}\beta ,`$ (7.9)
$`m_\tau `$ $`=`$ $`y_\tau v\mathrm{cos}\beta .`$ (7.10)
The Higgs mixing parameter $`\mu `$ is defined from the minimization conditions for the Higgs potential and requirement of radiative EWSB.
As an output we determine the mass spectrum of superpartners and of the Higgs bosons as functions of the only free parameter, namely $`m_{1/2}`$, which is directly related to the gluino mass $`M_3`$. Varying this parameter within the experimentally allowed range, one gets all the masses as functions of this parameter (see Table2).
For low $`tan\beta `$ the value of $`\mathrm{sin}\beta `$ is determined from eq.(7.8), while for high $`\mathrm{tan}\beta `$ it is more convenient to use the relation $`\mathrm{tan}\beta =\frac{m_t}{m_b}\frac{y_b}{y_t}`$, since the ratio $`y_t/y_b`$ is almost a constant in the range of possible values of $`y_t`$ and $`y_b`$.
For the evaluation of $`\mathrm{tan}\beta `$ one first needs to determine the running top- and bottom-quark masses. One can find them using the well-known relations to the pole masses (see e.g. ), including both QCD and SUSY corrections. For the top-quark one has:
$$m_t(m_t)=\frac{m_t^{pole}}{1+\left(\frac{\mathrm{\Delta }m_t}{m_t}\right)_{QCD}+\left(\frac{\mathrm{\Delta }m_t}{m_t}\right)_{SUSY}}.$$
(7.11)
Then, the following procedure is used to evaluate the running top mass. First, only the QCD correction is taken into account and $`m_t(m_t)`$ is found in the first approximation. This allows one to determine both the stop masses and the stop mixing angle. Next, having at hand the stop and gluino masses, one takes into account the stop/gluino corrections.
For the bottom quark the situation is more complicated because the mass of the bottom quark $`m_b`$ is essentially smaller than the scale $`M_Z`$ and so one has to take into account the running of this mass from the scale $`m_b`$ to the scale $`M_Z`$. The procedure is the following : one starts with the bottom-quark pole mass, $`m_b^{pole}=4.94\pm 0.15`$ and finds the SM bottom-quark mass at the scale $`m_b`$ using the two-loop $`QCD`$ corrections
$$m_b(m_b)^{SM}=\frac{m_b^{pole}}{1+\left(\frac{\mathrm{\Delta }m_b}{m_b}\right)_{QCD}}.$$
(7.12)
Then, evolving this mass to the scale $`M_Z`$ and using a numerical solution of the two-loop SM RGEs with $`\alpha _3(M_Z)=0.12`$ one obtains $`m_b(M_Z)_{SM}=2.91`$ GeV. Using this value one can calculate the sbottom masses and then return back to take into account the SUSY corrections from massive SUSY particles
$$m_b(M_Z)=\frac{m_b(M_Z)^{SM}}{1+\left(\frac{\mathrm{\Delta }m_b}{m_b}\right)_{SUSY}}.$$
(7.13)
When calculating the stop and sbottom masses one needs to know the Higgs mixing parameter $`\mu `$. For determination of this parameter one uses the relation between the $`Z`$-boson mass and the low-energy values of $`m_{H_1}^2`$ and $`m_{H_2}^2`$ which comes from the minimization of the Higgs potential:
$$\frac{M_Z^2}{2}+\mu ^2=\frac{m_{H_1}^2+\mathrm{\Sigma }_1(m_{H_2}^2+\mathrm{\Sigma }_2)\mathrm{tan}^2\beta }{\mathrm{tan}^2\beta 1},$$
(7.14)
where $`\mathrm{\Sigma }_1`$ and $`\mathrm{\Sigma }_2`$ are the radiative corrections . Large contributions to these functions come from stops and sbottoms. This equation allows one to obtain the absolute value of $`\mu `$, the sign of $`\mu `$ remains a free parameter.
Whence the quark running masses and the $`\mu `$ parameter are found, one can determine the corresponding values of $`\mathrm{tan}\beta `$ with the help of eqs.(7.8,7.9). This gives in low and high $`\mathrm{tan}\beta `$ cases, respectively
$`\mathrm{tan}\beta `$ $`=`$ $`1.47\pm 0.15\pm 0.05for\mu >0,`$
$`\mathrm{tan}\beta `$ $`=`$ $`1.56\pm 0.15\pm 0.05for\mu <0,`$
$`\mathrm{tan}\beta `$ $`=`$ $`69.3\pm 0.6\pm 0.3for\mu >0,`$
$`\mathrm{tan}\beta `$ $`=`$ $`38.1\pm 0.9\pm 0.4for\mu <0.`$
The deviations from the central value are connected with the experimental uncertainties of the top-quark mass, $`\alpha _3(M_Z)`$ and uncertainty due to the fixed point values of $`y_t(M_Z)`$ and $`y_b(M_Z)`$.
Having all the relevant parameters at hand it is possible to estimate the masses of the Higgs bosons. With the fixed point behaviour the only dependence left is on $`m_{1/2}`$ or the gluino mass $`M_3`$. It is restricted only experimentally: $`M_3>144`$ GeV for arbitrary values of the squarks masses.
Let us start with low $`\mathrm{tan}\beta `$ case. The masses of CP-odd, charged and CP-even heavy Higgses increase almost linearly with $`M_3`$. The main restriction comes from the experimental limit on the lightest Higgs boson mass. It excludes $`\mu <0`$ case and for $`\mu >0`$ requires the heavy gluino mass $`M_3750`$ GeV. Subsequently one obtains
$$m_A>844GeV,m_{H^\pm }>846GeV,m_H>848GeV,for\mu >0,$$
i.e. these particles are too heavy to be detected in the nearest experiments.
For high $`\mathrm{tan}\beta `$ already the requirement of positivity of $`m_A^2`$ excludes the region with small $`M_3`$. In the most promising region $`M_3>1`$ TeV ($`m_{1/2}>300`$ GeV) for the both cases $`\mu >0`$ and $`\mu <0`$ the masses of CP-odd, charged and CP-even heavy Higgses are also too heavy to be detected in the near future
$$m_A>1100\text{GeV for}\mu >0,m_A>570\text{GeV for}\mu <0,$$
$$m_{H^\pm }>1105\text{GeV for}\mu >0,m_{H^\pm }>575\text{GeV for}\mu <0.$$
$$m_H>1100\text{GeV for}\mu >0,m_H>570\text{GeV for}\mu <0.$$
The situation is different for the lightest Higgs boson $`h`$, which is much lighter. Radiative corrections in this case are crucial and increase the value of the Higgs mass substantially . They have been calculated up to the second loop order . As can be seen from eq.(7.7) the one loop correction is positive increasing the tree level value almost up to 100% and the second one is negative thus decreasing it a little bit. We use in our analysis the leading two-loop contributions evaluated in ref.. As has been already mentioned these corrections depend on the masses of squarks and the other superpartners for which we substitute the values obtained above from the IRQFP’s.
The results depend on the sign of parameter $`\mu `$. It is not fixed since the requirement of EWSB defines only the value of $`\mu ^2`$. However, for low $`\mathrm{tan}\beta `$ negative values of $`\mu `$ lead to a very small Higgs mass which is already excluded by modern experimental data, so further on we consider only the positive values of $`\mu `$. Fot high $`\mathrm{tan}\beta `$ both signs of $`\mu `$ are allowed.
Consider first the low $`\mathrm{tan}\beta `$ regime. At the upper part of Fig.11 it is shown the value of $`m_h`$ for $`\mu >0`$ as a function of the geometrical mean of stop masses - this parameter is often identified with a supersymmetry breaking scale $`M_{SUSY}`$. One can see that the value of $`m_h`$ quickly saturates close to $``$ 100 GeV. For $`M_{SUSY}`$ of the order of 1 TeV the value of the lightest Higgs mass is
$$m_h=(94.3+1.6+0.6\pm 5\pm 0.4)\text{GeV},\text{ for}M_{SUSY}=1TeV,$$
(7.15)
where the first uncertainty comes from the deviations from the IRQFPs for the mass parameters, the second one is related to that of the top-quark Yukawa coupling, the third reflects the uncertainty of the top-quark mass of 5 GeV, and the last one comes from that of the strong coupling.
One can see that the main source of uncertainty is the experimental error in the top-quark mass. As for the uncertainties connected with the fixed points, they give much smaller errors of the order of 1 GeV.
The obtained result (7.15) is very close to the upper boundary, $`m_h=97`$ GeV, obtained in Refs. . Note, however, that the uncertainties mentioned above as well as the upper bound are valid for the universal boundary conditions. Loosing these requirement leads to increase of the upper bound values of the Higgs mass in case of low $`\mathrm{tan}\beta `$ up to $`102`$ GeV .
For the high $`\mathrm{tan}\beta `$ case the lightest Higgs is slightly heavier, but the difference is crucial for LEP II. The mass of the lightest Higgs boson as a function of $`M_{SUSY}`$ is shown in the lower part of Fig.11 . One has the following values of $`m_h`$ at a typical scale $`M_{SUSY}=1`$ TeV ($`M_31.3`$TeV) :
$`m_h`$ $`=`$ $`128.20.47.1\pm 5\text{GeV, for}\mu >0,`$
$`m_h`$ $`=`$ $`120.60.13.8\pm 5\text{GeV, for}\mu <0.`$
The first uncertainty is connected with the deviations from the IRQFPs for mass parameters, the second one with the Yukawa coupling IRQFPs, and the third one is due to the experimental uncertainty in the top-quark mass. One can immediately see that the deviations from the IRQFPs for mass parameters are negligible and only influence the steep fall of the function on the left, which is related to the restriction on the CP-odd Higgs boson mass $`m_A`$. In contrast with the low $`\mathrm{tan}\beta `$ case, where the dependence on the deviations from Yukawa fixed points was about $`1`$ GeV, in the present case it is much stronger. The experimental uncertainty in the strong coupling constant $`\alpha _s`$ is not included because it is negligible compared to those of the top-quark mass and the Yukawa couplings and is not essential here contrary to the low $`\mathrm{tan}\beta `$ case.
One can see that for large $`\mathrm{tan}\beta `$ the masses of the lightest Higgs boson are typically around 120 GeV that is too heavy for observation at LEP II. Note, however, that the uncertainties increase if one considers the non-universal boundary conditions which decreases the lower boundary for the Higss mass<sup>3</sup><sup>3</sup>3Lep II is now increasing its energy and may possibly reach the lower bound of high $`\mathrm{tan}\beta `$ scenario predictions.. At the same time recent experimental data has practically excluded the low $`\mathrm{tan}\beta `$ scenario and in the next year the situation will be completely clarified.
Thus, one can see that in the IRQFP approach all the Higgs bosons except for the lightest one are found to be too heavy to be accessible in the nearest experiments. This conclusion essentially coincides with the results of more sophisticated analyses. The lightest neutral Higgs boson, on the contrary is always light. The situation may improve a bit if one considers more complicated models with enlarged Higgs structure . However, it does not change the generic feature of SUSY theories, the presence of light Higgs boson.
In the near future with the operation of the Tevatron and LHC proton-antiproton accelerators with the c.m energy of 2 and 15 TeV, respectively, the energy range from 100 to 1000 GeV will be scanned and supersymmetry will be either discovered or the minimal version will be abandoned in particle physics. The Higgs boson might be the first target in search of SUSY.
## 8 Conclusion
We have attempted to show how following the pattern of RG flow one can explore physics lying beyond the Standard Model of fundamental interactions. The methods are essentially based on renormalization group technique in the framework of quantum field theory. The running of parameters is the key ingredient in attempts to go beyond the range of energies accessible to modern accelerators, to find manifestation of new physics. In the absence of solid experimental facts the RG flow happens to be the only sauce of definite predictions. The peculiarity of the moment is that these predictions are the subject of experimental tests today, and in the near future, so one may hope to check the ideas described above at the turn of the Millennium.
Acknowledgments
The author is grateful to S.Codoban, A.Gladyshev, M.Jurčišin, V.Velizhanin, K.Ter-Martirossian, R.Nevzorov, W. de Boer and G.Moultaka for useful discussions and to S.Codoban for help in preparing the manuscript. Special thanks to the organizers of the conference D.O’Connor and C.Stephens. Financial support from RFBR grants # 99-02-16650 and # 96-15-96030 is kindly acknowledged. |
warning/0001/quant-ph0001060.html | ar5iv | text | # Universal Quantum Entanglement Concentration Gate
## Abstract
We construct a Universal Quantum Entanglement Concentration Gate (QEC-Gate). Special times operations of QEC-Gate can transform a pure 2-level bipartite entangled state to nearly maximum entanglement. The transformation can attain any required fidelity with optimal probability by adjusting concentration step. We also generate QEC-Gate to the Schmidt decomposable multi-partite system.
PACS numbers: 03.67.-a, 03.65.Bz, 89.70.+c
The entanglement nature is a key source to distinguish quantum and classical information theory. Some authors $`[13]`$ have been successful in identifying many fundamental properties of entanglement. Because many important quantum processes, such as teleportation $`[4]`$, superdense code $`[5]`$ and quantum computation $`[6]`$, require maximally entangled states, the methods of entanglement enhancement and entanglement concentration was frequently discussed $`[13,79]`$. Such work is closely related to the fundamental problems of state transformation $`[1018]`$. Two main protocols, probabilistic $`[7,11]`$ and approximate $`[16]`$, has been applied in the optimal process. An unavoidable problem of these work is that the transformation greatly depends on the input states.
In this report, we construct a Universal Quantum Entanglement Concentration Gate (QEC-Gate), multi-operation of which can transform a $`2`$-level bipartite partial pure entangled state to a nearly maximally entangled state with optimal probability. Its fidelity can be entirely controlled by the parameter concentration step. Any required fidelity can be approached by adjusting concentration step and enhancing operation times. Additionally, each operation of QEC-Gate can always process with the optimal probability.
In general, a $`2`$-level bipartite entangled state can be represented as the following according to Schmidt decomposition,
$$|\psi =\mathrm{cos}\theta |11+\mathrm{sin}\theta |00,$$
(1)
where $`0\theta \frac{\pi }{4}`$. Its entanglement, $`E\left(|\psi \right)=\mathrm{cos}^2\theta \mathrm{log}_2\mathrm{cos}^2\theta \mathrm{sin}^2\theta \mathrm{log}_2\mathrm{sin}^2\theta `$ can be compared directly by $`\theta `$.
Suppose the pure entangled state is shared by two distant observers, Alice and Bob. Since any operation in quantum mechanics can be represented by a unitary evolution together with a measurement, a probe $`P`$ with a Hilbert space, $`|P_0|P_1`$ ($`|P_0`$ and $`|P_1`$ are orthogonal), is introduced. Our gate operation is a local unitary evolution (Alice side) together with the postselection of the measurement results which can be generally represented as
$$(\widehat{U}_{AP}I_B)|\psi |P_0=\sqrt{\gamma }|\stackrel{~}{\psi }|P_0+\sqrt{1\gamma }|\varphi _{AB}|P_1,$$
(2)
where $`\gamma `$ is the probability of successful state transformation of $`|\psi |\stackrel{~}{\psi }`$, $`|\varphi _{AB}`$ is an arbitrary state in the composite system of Alice and Bob. If the measurement of the probe results in $`|P_0`$, this transformation is successful.
The QEC-Gate $`G(\xi ,\eta )`$ executes the unitary-reduction operation $`U_{AP}`$ which is described as the following with two parameters $`\xi `$ and $`\eta `$, $`0<\xi <\eta <\frac{\pi }{4}`$,
$`\widehat{U}_{AP}|1_A|P_0`$ $`=`$ $`\sqrt{\gamma _0}{\displaystyle \frac{\mathrm{cos}\eta }{\mathrm{cos}\xi }}|1_A|P_0`$ (3)
$`+\sqrt{1\gamma _0}{\displaystyle \frac{1}{\mathrm{cos}\xi }}|1_A|P_1,`$
$$\widehat{U}_{AP}|0_A|P_0=\sqrt{\gamma _0}\frac{\mathrm{sin}\eta }{\mathrm{sin}\xi }|0_A|P_0,$$
(4)
where $`\gamma _0=\frac{\mathrm{sin}^2\xi }{\mathrm{sin}^2\eta }`$ is exactly the optimal probability to transform entangled state $`\mathrm{cos}\xi |11_{AB}+\mathrm{sin}\xi |00_{AB}`$ to $`\mathrm{cos}\eta |11_{AB}+\mathrm{sin}\eta |00_{AB}`$ $`[11]`$. In fact, $`U_{AP}`$ can be implemented by qubit $`A`$ controlling $`P`$ rotation. It can be expressed as a generalized Toffoli gate $`\mathrm{\Lambda }_1\left(U\right)`$ $`\left[19\right]`$ that applies $`U`$ to qubit $`P`$ if and only if $`A`$ on $`|1`$, where $`U=\left(\begin{array}{cc}\delta & \sqrt{1\delta ^2}\\ \sqrt{1\delta ^2}& \delta \end{array}\right)`$ with $`\delta =\frac{\mathrm{tan}\xi }{\mathrm{tan}\eta }`$.
We set $`\theta =\theta _00,\frac{\pi }{4}`$ in Eq.(1) as the partial entangled initial input state (it is reasonable because if the initial input state’s Schmidt decomposition is represented on another base, an unitary operation $`V_AV_B`$ is needed to transform it to the form of Eq.(1)). One operation of $`G(\xi ,\eta )`$ on the input can be represented as
$`\left(\widehat{U}_{AP}I_B\right)\left(\mathrm{cos}\theta _0|11_{AB}+\mathrm{sin}\theta _0|00_{AB}\right)|P_0`$ ()
$`=`$ $`\sqrt{\gamma _0}\left(\mathrm{cos}\theta _0{\displaystyle \frac{\mathrm{cos}\eta }{\mathrm{cos}\xi }}|11_{AB}+\mathrm{sin}\theta _0{\displaystyle \frac{\mathrm{sin}\eta }{\mathrm{sin}\xi }}|00_{AB}\right)|P_0`$ (5)
$`+\sqrt{1\gamma _0}{\displaystyle \frac{\mathrm{cos}\theta _0}{\mathrm{cos}\xi }}|11_{AB}|P_1.`$
The normalization of output state in Eq.(5) can yield the transformation probability $`\gamma _1=\frac{\mathrm{sin}^2\theta _0}{\mathrm{sin}^2\theta _1}`$. This probability is also optimal according to Vidal Theorem $`\left[11\right]`$.
Denote the output entangled state as $`\mathrm{cos}\theta _1|11_{AB}+\mathrm{sin}\theta _1|00_{AB}`$, we get
$$\mathrm{tan}\theta _1=\frac{1}{\delta }\mathrm{tan}\theta _0.$$
(6)
Define Concentration Step $`\mathrm{\Delta }(\xi ,\eta )=\frac{1}{\delta }=\frac{\mathrm{tan}\eta }{\mathrm{tan}\xi }`$. Notice that if $`\xi `$ and $`\eta `$ are set properly, which means that $`\mathrm{\Delta }(\xi ,\eta )1`$ has a small value, we can promise the enhancement of entanglement. A natural application is to operate $`G(\xi ,\eta )`$ on the initial input state more than one times. For each operation, the entanglement get an increase. Surely this situation can not continue infinitely and the converting operation indicates the approach of maximum. Another obvious property of QEC-Gate is that the enhancement of entanglement is discrete. So in general cases, the final concentrated state is not exactly the maximum entangled state. Here we can determine the optimal operation times $`T`$, the final output state $`|\mathrm{\Phi }_{final}`$, total probability $`\mathrm{\Gamma }`$ and fidelity $`F`$.
$$T=\left[\frac{\mathrm{ln}\mathrm{tan}\theta _0}{\mathrm{ln}\mathrm{\Delta }(\xi ,\eta )}\right],$$
(6)
where $`\left[x\right]`$ is the Gauss Functor which gives the integer part of real number $`x`$. Notice that this result is based on the limit of $`\theta _T(0,\frac{\pi }{4}]`$. If we permit $`\theta _T>\frac{\pi }{4}`$, when $`\theta _i\frac{\pi }{4}<\theta _{i+1}`$, we set $`T=i`$ if $`\left|\theta _i\frac{\pi }{4}\right|\left|\theta _{i+1}\frac{\pi }{4}\right|`$, otherwise, $`T=i+1`$. This choice can promise the optimal output.
$`|\mathrm{\Phi }_{final}`$ ()
$`=`$ $`G^T\left(\mathrm{cos}\theta _0|11_{AB}+\mathrm{sin}\theta _0|00_{AB}\right)`$ (7)
$`=`$ $`A\left(\mathrm{cos}\theta _0\left({\displaystyle \frac{\mathrm{cos}\eta }{\mathrm{cos}\xi }}\right)^T|11_{AB}+\mathrm{sin}\theta _0\left({\displaystyle \frac{\mathrm{sin}\eta }{\mathrm{sin}\xi }}\right)^T|00_{AB}\right),`$ (8)
where $`A`$ is a normalizing coefficient which is
$$A\left(\theta _0\right)=\left(\left(\mathrm{cos}^2\theta _0+\mathrm{\Delta }^{2T}\mathrm{sin}^2\theta _0\right)\left(\frac{1+\mathrm{tan}^2\xi }{1+\mathrm{tan}^2\eta }\right)^T\right)^{\frac{1}{2}}.$$
(9)
The probability and fidelity are
$$\mathrm{\Gamma }=\gamma _i=\frac{\mathrm{sin}^2\theta _0}{\mathrm{sin}^2\theta _T}2\mathrm{sin}^2\theta _0,$$
(9)
$$F=\frac{1}{2}\left(\frac{1+\mathrm{tan}\theta _0\mathrm{\Delta }^T(\xi ,\eta )}{\sqrt{1+\mathrm{tan}^2\theta _0\mathrm{\Delta }^{2T}(\xi ,\eta )}}\right)^2.$$
(10)
From the results above, we can see that the total probability is only related to the initial input state. The probability of each transformation approaches $`1`$ when $`\mathrm{\Delta }(\xi ,\eta )1`$ is small enough. But with the increasing operation times, $`\mathrm{\Gamma }`$ remains approximately unchanged.
Because operation times $`T`$ is not a continuous variable, the fidelity represented by Eq.(10) changes periodically with $`\theta _0`$. The period is determined by concentration step $`\mathrm{\Delta }`$, that is $`F\left(\theta _0\right)=F\left(\theta _0\mathrm{arctan}\mathrm{\Delta }\right)`$. For some special $`\theta _0\left\{\theta _0^{\left(k\right)}=\mathrm{arctan}\frac{1}{\mathrm{\Delta }^k},k=1,2,\mathrm{}\right\}`$, $`F`$ can exactly attain 1. But the average fidelity, or even the minimum, is more important. To examine the relationship of $`F`$ and concentration step, we firstly calculate the minimum of $`F\left(\theta _0\right)`$. In this situation, $`\frac{\pi }{4}\theta _T=\theta _{T+1}\frac{\pi }{4}`$, so we get
$$F_{\mathrm{min}}\left(\mathrm{\Delta }\right)=\mathrm{min}F\left(\theta _0\right)=\frac{1}{2}\left(\frac{1+\sqrt{\mathrm{\Delta }}}{\sqrt{1+\mathrm{\Delta }}}\right)^2.$$
(11)
The differential coefficient shows that $`F_{\mathrm{min}}\left(\mathrm{\Delta }\right)`$ is a monotonous decreasing function of $`\mathrm{\Delta }`$. We can increase the fidelity by reducing $`\mathrm{\Delta }(\xi ,\eta )`$. When $`\mathrm{\Delta }1`$, $`F_{\mathrm{min}}\left(\mathrm{\Delta }\right)1`$. Obviously, it is realized at the price of increasing operation times.
We (three of us) $`[18]`$ have shown that two different entangled states can be concentrated by same local operations and classical communication if and only if they share the same marginal density operator on one side. Although the design of QEC-Gate makes it possible to attain any required fidelity, it doesn’t contradict with the previous result. For different initial states, the final states and the operation times, which determine the practical operation of QEC-Machine, may be different. Actually, the discrete strategy which is adopted to simulate the continuous transformation space can create an identical unit to approximately measure the space. The finer the separation is, the better the measure is. This is exactly the reason why a universal gate can be constructed.
For the complicated situation of $`n`$-partite entanglement, the QEC-Gate can not be directly generated. But it is still efficient in transforming some special entanglements, such as the Schmidt decomposable entangled states $`\left\{\mathrm{cos}\theta |11\mathrm{}1+\mathrm{sin}\theta |00\mathrm{}0,0<\theta <\frac{\pi }{4}\right\}`$. Since the QEC-Gate operates only on the composite system of one of the $`n`$ parties and the probe, the result is the same as that of $`2`$-partite state transformation.
In summery, we have constructed a Universal Quantum Entanglement Concentration Gate, multi-operation of which can transform a partial entanglement to a nearly maximum entanglement with optimal probability. Any required fidelity can be approached by decreasing concentration step.
This work was supported by National Nature Science Foundation of China. |
warning/0001/astro-ph0001230.html | ar5iv | text | # 7Li in Metal-Poor Stars: The Spread of the Li Plateau
## 1. Introduction
The first indication that the old stars of the Galaxy exhibited an almost uniform Li abundance emerged at IAU Coll. 68, when Spite & Spite (1981, 1982) presented their first observations of warm halo dwarfs. Almost two decades later, IAU Symp. 198 met to consider progress in studies of this and other light elements.
Studies by many workers in the decade following the Spite & Spite discoveries resulted in mounting evidence that the warm halo dwarfs exhibited a unique Li abundance (e.g. Rebolo, Molaro, & Beckman 1988). The interpretation of this abundance as the primordial one reflecting big bang nucleosynthesis, at worst “hardly altered” (Spite & Spite 1982), hinged on the importance of possible depletion of Li from a higher initial abundance. While ample evidence existed of Li destruction in some stars, the lack of a significant spread in halo dwarf Li abundances provided empirical evidence that destruction may have been minimal in these objects. (See Boesgaard & Steigman 1985 for a review of that period.) Classical stellar evolution models (e.g. Deliyannis, Demarque, & Kawaler 1990) fitted this interpretation, showing that Li destruction in metal-deficient dwarfs with shallow surface convective zones would be minimal ($`{}_{}{}^{}{}_{}{}^{<}`$ 0.05 dex). However, this same class of stellar models failed to explain numerous Population I star observations, and an alternative class of models invoking extra mixing implied that considerable Li depletion (as high as 1 dex; Pinsonneault, Deliyannis, & Demarque 1992) could have occurred in the halo stars.
In the next decade, several dissenting voices were heard. Deliyannis, Pinsonneault, & Duncan (1993) argued that there was a non-negligible spread in the Li abundances of the halo dwarfs that would not be consistent with a perfectly primordial composition. Depending on the sample selected, they found a Li spread of $`\sigma 0.04`$ dex. Thorburn (1994) found an even greater intrinsic spread $`\sigma 0.10`$ dex, and moreover claimed, as did Norris, Ryan, & Stringfellow (1994), that the abundances depended on both $`T_{\mathrm{eff}}`$ and \[Fe/H\]. Such dependences were contrary to the notion of a unique Li abundance in the halo stars, and thus undermined the association of the observed Li abundance(s) with the primordial one. The efforts of Ryan et al. (1996) to bring all previous observations onto a uniform temperature and abundance scale did not eliminate the cited dependences.
One of the largest uncertainties in abundance analyses is errors in the effective temperature scales. Spite & Spite (1993) and Bonifacio & Molaro (1997) discussed the possible role of such errors in distorting an otherwise uniform Li abundance, the latter work finding the previously reported trends to become insignificant when a more recent temperature calibration based on the infra-red flux method (IRFM) was applied. Although the IRFM scale might be expected to provide better systematics, the large individual uncertainties attached to each temperature determination by this method limit the scale’s ability to distinguish between effects at the level of those claimed for Li. The existence of some large errors even in the metallicity estimates for program stars also hampers the efforts. In particular, the literature data utilised by Bonifacio & Molaro (1997) includes several poorly determined values which, in hindsight, frustrated their analysis by smearing out the data (Ryan, Norris, & Beers 1999, §7.3.3).
## 2. The Intrinsic Spread of <sup>7</sup>Li
In an effort to avoid the impact of undesirable errors, Ryan, Norris, & Beers (1999) set out to obtain a highly homogeneous data set on a sample occupying only a narrow range of $`T_{\mathrm{eff}}`$, \[Fe/H\], and evolutionary type. Restricting their sample to 6000 K $`{}_{}{}^{<}T_{\mathrm{eff}}^{}_{}^<`$ 6400 K and $`3.5_{}^<`$ \[Fe/H\] $`{}_{}{}^{<}2.5`$, applying double-blind data analysis techniques, obtaining multiple high-resolution, high-S/N observations of the targets, and using multiple temperatures indicators to minimise random errors, they achieved a formal abundance error as low as $`\sigma _{\mathrm{err}}`$ = 0.033 dex per star. These results are at considerably higher precision than most previous Li measurements (typically having $`\sigma _{\mathrm{err}}`$ $``$ 0.06 – 0.08 dex).
The sample was known to contain one previously known ultra-Li-deficient star, G186-26, which was excluded from the analysis. Remaining objects exhibited a total observed spread $`\sigma _{\mathrm{obs}}`$ = 0.053 dex, considerably less than that found by Thorburn (1994). However, this 0.053 dex was found to be dominated by an underlying metallicity dependence, and the spread of the Li abundances about this trend is a mere $`\sigma _{\mathrm{obs}}`$ = 0.031 dex (see Figure 1), and Gaussian in form. This corresponds to the spread in Li abundance at a given metallicity. Comparing this with the formal measurement errors of $`\sigma _{\mathrm{err}}`$ = 0.033 dex leads to the conclusion that the intrinsic spread in the stars must be negligible. We state a generous upper-limit on the intrinsic spread as being $`\sigma _{\mathrm{int}}<0.02`$ dex.
An important consequence of the very narrow spread of Li abundances is its ability to constrain the impact of possible extra-mixing in so far as extra-mixing models predict a spread in the final Li abundances of a population of stars. The rotationally-induced mixing models of Pinsonneault et al. (1993) suggested that Li depletion by as much as an order of magnitude could have occurred in halo turnoff dwarfs. The more recent work of Pinsonneault et al. (1999), in concert with the observational data of Thorburn (1994), revised downward the depletion level to $`0.2`$ – 0.4 dex. As Figure 2 shows, the data from Ryan et al. (1999) with their narrower spread (at a given metallicity) rule out rotationally-induced mixing models that exhibit even 0.1 dex median depletion. Considering the size of the observed sample and the absence of stars in the tail of the theoretical distribution, Poisson statistics provide only 10% chance that the observed and theoretical curves are compatible, or 90% probability that the median depletion is less than 0.1 dex. Other statistical tests discriminate the models even more significantly.
We seek now to explain previous results that yielded contrary conclusions. Thorburn’s (1994) data reduction process explicitly excluded sky background and scattered light subtractions, but this shortcut was not reflected in the formal error estimates. Incorporation of the errors introduced by this procedure are enough, on average, to inflate the error estimates to the size required by the observed scatter. That is, the scatter observed by Thorburn is almost certainly consistent with that resulting from data acquisition and analysis. Bonifacio & Molaro (1997) found no significant metallicity dependence in their analysis, but as discussed above, certain metallicities they adopted from the literature were found subsequently to be unreliable. This, and the large random errors inherent in the IRFM temperature scale, resulted in the weak Li evolution being washed out. (See Ryan et al. 1999, §7.3.3, for a detailed analysis.) Finally, we note that the small spread of abundances found by Deliyannis et al. (1993) is consistent with the observed spread in our sample if the underlying metallicity trend is overlooked. In fact, the Deliyannis et al. study pre-dated any claims of a metallicity dependence, and their result was probably driven by the large metallicity range in their sample.
## 3. The Underlying Li vs \[Fe/H\] Trend
Although Li GCE during the halo-forming era has often been ignored, we should not be surprised that it exists. If recent detections of <sup>6</sup>Li in halo stars (Smith, Lambert, & Nissen 1993,1998; Hobbs & Thorburn 1994,1997; Cayrel et al. 1999, Deliyannis & Ryan 2000) are correct, then we would be surprised not to see <sup>7</sup>Li GCE. With the measurement precision attainable using modern CCDs and large aperture telescopes, even small levels of <sup>7</sup>Li enrichment can be measured, and it is consistent with the measured <sup>6</sup>Li abundances.
To see whether the observed trend was compatible with GCE, Ryan et al. (2000a) examined the Fields & Olive (1999a,b) model. For halo stars, the $`\nu `$-process and GCR spallation are the most likely sources of <sup>7</sup>Li. The Fields & Olive model normalises the GCR contribution to meteoritic Be and <sup>6</sup>Li abundances, and normalises the $`\nu `$-process to the otherwise unaccounted for <sup>11</sup>B. The model does not include stellar <sup>7</sup>Li sources acting in the later stages of Galactic evolution, and hence does not model the Population I abundance. The models of Romano et al. (1999) incorporate many Population I sources (primarily the $`\nu `$-process, AGB stars, and novae). Figure 3 shows two variants of Romano et al’s models and a hybrid using the GCR predictions of Fields & Olive from Ryan et al. (2000a). The model reproduces not only the halo star Li evolution discussed above, but also fits new data around \[Fe/H\] $``$ $`1.5`$ (Ryan et al. 2000b; see below) and the lowest metallicity datum at \[Fe/H\] = $`3.7`$ (Norris, Beers, & Ryan 2000) which were added after the Fields & Olive model was produced.
Ryan et al. (1996) indicated that a selection bias existed in the Li data available in the literature, in that most studies centred on more-metal-poor objects. Few more-metal-rich halo stars had been examined, and those which had were on the whole cooler than the metal-poor ones. Given the difficulties with temperature scales for stars, the comparing of warmer metal-poor stars with cooler metal-rich stars was clearly undesirable. In an effort to address this bias, Ryan et al. (2000b) obtained data on 18 more-metal-rich halo stars, with $`2.0_{}^<`$ \[Fe/H\] $`{}_{}{}^{<}1.0`$, but in the warm temperature range $`T_{\mathrm{eff}}{}_{}{}^{>}6000`$ K as for the most metal-poor samples. The sample was found to contain four ultra-Li-deficient halo stars, which will be discussed separately in the next section. The remaining stars, shown as solid circles in Figure 3, were found to sit exactly where the Fields & Olive model predicts. We emphasise that the model was completed prior to the reduction and analysis of the metal-rich halo sample, so the agreement between the two is a genuine accomplishment, not something achieved artificially in the model. This is viewed as additional evidence that the trend of Li with \[Fe/H\] evidenced in Figure 1 is a result of natural GCE of the element during formation of the Galactic halo.
## 4. The Primordial Li Abundance and Uncertainties
We combine these measurements in Table 1 to present a new accounting for the primordial Li abundance. Beginning with the observed abundance at the mean metallicity of our sample, we apply corrections for the inferred GCE contribution (with uncertainties) and for stellar depletion. For the latter we take the value implied by classical models, but in the uncertainties allow for additional depletion up to the 0.1 dex limit of the rotationally-induced mixing models. Temperature scale uncertainties remain one of the largest sources of error, and in this analysis we apply an offset of 0.08 dex to the Li abundance, corresponding to a change from the temperature scale adopted in our original analysis (based on the Bell & Oke (1986) and Magain (1987) b-y scales) to the systematically hotter IRFM scale of Alonso, Arribas, & Martínez–Roger (1996). However, we associate an uncertainty of $`\pm `$0.08 dex with this process, in recognition of the remaining difficulties in the temperature scales for halo dwarfs. These and the other affects tabulated lead us to infer a primordial abundance $`A`$(Li)<sub>p</sub> = 2.09 $`{}_{0.13}{}^{}{}_{}{}^{+0.19}`$ dex, where the uncertainties resemble 2$`\sigma `$ limits (Ryan et al. 2000a).
## 5. The Ultra-Li-Deficient Halo Dwarfs: Blue Stragglers After All?
Boesgaard & Tripicco (1986a) showed that Hyades stars with 6400 K $`<T_{\mathrm{eff}}<`$ 6900 K exhibit extremely low surface Li abundances. These and similar stars in other Population I clusters became known as “Li-dip stars”. They showed that a second mechanism, besides convection on the pre-main sequence (and possibly main sequence) for lower mass stars, could greatly deplete surface Li abundances. Lambert, Heath, & Edvardsson (1991) showed that most of the strongly Li-depleted Population I field stars, shown for example in Figure 3, could be explained as either having evolved from the Li dip or being low mass convectively-depleted stars. However, they also noted that a small number of stars did not share these histories, and proposed that perhaps 10% of stars had experienced additional severe Li depletion through unknown causes. This work preceded the discovery of halo dwarfs whose temperatures and metallicities coincided with the Spite plateau but which were ultra-Li-deficient by more than an order of magnitude or so (Hobbs, Welty & Thorburn 1991; Thorburn 1992; Spite et al. 1993). The halo examples have been estimated at perhaps 3–5% of the Population, and may result from the same process as the Population I class proposed by Lambert et al. The nature of the process resulting in their Li-depletion has remained unclear, and our inability to explain them has been given as a reason to mistrust the entire Population II interpretation (e.g. Thorburn 1994). Whether or not such a view is held (cf. Ryan et al. 1999), they identify an embarrassing deficit in our knowledge of stellar processing of this important element. Efforts to identify common chemical signatures other than Li deficiency proved impossible; instead considerable diversity and heterogeneity was found amongst the complete sample (four) known at the end of 1998 (Norris et al. 1997; Ryan, Norris, & Beers 1998).
The study of 18 more-metal-rich halo stars by Ryan et al. (2000b) discussed above resulted in the discovery of four new ultra-Li-deficient halo stars; see Figure 4 (Ryan et al. 2000c). The discovery rate in that study, 22%, contrasts greatly with the previous Population estimate of 3–5%, and indicates that they are preferentially clustered in the stellar parameter range singled out in that investigation, namely warm, more-metal-rich halo stars. The stars are therefore seen to be grouped preferentially towards the main sequence turnoff of the halo, but not exclusively so. The clustering near the turnoff is reminiscent of blue stragglers, but the hypothesis that they were redward-evolving blue stragglers had already been ruled out for the previously known examples (Thorburn 1994; Norris et al. 1997). However, the discovery of four more such stars preferentially close to the main sequence turnoff resulted in the re-examination of the blue-straggler hypothesis, with the distinction that main-sequence blue-stragglers-to-be are now considered. Halo stars initially cooler than about 5700 K, corresponding to a mass of $`0.7M_{}`$, deplete their surface Li during their pre-main-sequence evolution. When two low-mass stars merge to become a single object higher up the main sequence near the halo turnoff with a total mass around 0.7-0.85 $`M_{}`$, they will in most cases form from stars which have already destroyed their Li. They will appear, then, as normal halo main-sequence stars except with respect to two parameters: (1) their Li abundances will be extremely low and hence appear abnormal, and (2) their main sequence lifetimes will be extended due to the delayed onset of nuclear burning at a rate expected of stars of their (combined) mass. The stars we now observe as ultra-Li-deficient halo stars, preferentially but not uniquely clustered towards the turnoff, may indeed be the progeny of such mergers and the progenitors of future blue-stragglers. Indeed, extreme Li deficiency may be the only common signal of future Pop II blue-stragglers-to-be.
If this hypothesis is correct, then we may finally remove such stars with confidence from discussions of the spread about the Spite plateau, and consider them as a truly distinct class of stars whose evolutionary history explains their abnormal Li abundances. Whether this mechanism can also explain the heterogeneity found for the abundances of their other elements remains to be seen.
## 6. Differences Between Halo Field and Globular Cluster Stars?
Although the halo field stars discussed above have minimal intrinsic spread about the Li Spite plateau (at a given abundance), data for globular cluster samples show a different picture. Figure 5 compares the very metal-poor field turnoff dwarf data for stars spanning a dex in \[Fe/H\] with observations of subgiants in M92 (Boesgaard et al. 1998). The two groups exhibit quite different Li characteristics! The considerable spread in the globular cluster sample prompted Boesgaard et al. to favour a mechanism in which a higher pre-stellar abundance has been depleted by varying degrees in the stars, possibly by the rotationally-induced mixing mechanism discussed earlier. Why this mechanism should differ for the globular cluster and field star samples is unclear. Differences in angular momentum evolution in the two environments may be responsible, but the details have yet to be proposed. Other examples of differences in the mixing of stellar envelopes in field star and globular cluster samples has been forthcoming in recent years (e.g. Hanson et al. 1998), adding to previous evidence of field-vs-cluster differences in CNO element ratios.
Until the cause of the difference is understood, one must choose whether to use the field star or the globular cluster data to interpret GCE. I would argue that the large Galactic volume sampled by field stars in contrast to the small total volume of globular clusters, and the greatly increased possibility of star-to-star interactions in the high stellar densities of the latter, would render field star samples more representative of the evolution of the Galaxy as a whole. Of course, this in no way reduces the importance of understanding the globular cluster element abundance patterns for what they may tell us about the evolution of the Galaxy and stellar processes in dense environments.
### Acknowledgments.
This work represents the outcomes of collaborations involving Prof. J. E. Norris (Australian National University), Dr T. C. Beers (Michigan State University), Dr K. A. Olive (University of Minnesota), Dr B. D. Fields (University of Illinois at Urbana-Champaign), Dr T. Kajino (National Astronomical Observatory of Japan), Ms. D. Romano (SISSA, Trieste), and Ms K. Rosolankova (The Open University, & St Hilda’s College, Oxford), all of whose contributions are gratefully acknowledged. The author is likewise grateful to the IAU for partial support to attend this meeting.
## References
Alonso, A., Arribas, S., & Martínez–Roger, C. 1996, A&AS, 117, 227
Bell, R. A. & Oke, J. B. 1986, ApJ, 307, 253
Boesgaard, A. M., Deliyannis, C. P., Stephens, A., & King, J. R. 1998, ApJ, 493, 206
Boesgaard, A. M. & Steigman, G. 1985, ARAA, 23, 319
Boesgaard, A. M. & Tripicco, M. J. 1986a, ApJ, 302, L49
Boesgaard, A. M. & Tripicco, M. J. 1986b, ApJ, 303, 724
Bonifacio, P. & Molaro, P. 1997, MNRAS, 285, 847
Cayrel, R., Spite, M., Spite, F., Vangioni-Flam, E., Cassé, M., & Audouze, J. 1999, A&A, 343, 923
Deliyannis, C. P., Demarque, P., & Kawaler, S. D. 1990, ApJS, 73, 21
Deliyannis, C. P., Pinsonneault, M. H., & Duncan, D. K. 1993, ApJ, 414, 740
Deliyannis, C. P. & Ryan, S. G. 2000, in prep
Fields, B. D. & Olive, K. A. 1999, New Astronomy, 4, 255
Fields, B. D. & Olive, K. A. 1999, ApJ, 516, 797
Hanson, R. B., Sneden, C., Kraft, R. P., & Fulbright, J. 1998, AJ, 116, 1286
Hobbs, L. M., & Thorburn, J. A. 1994, ApJ, 428, L25-L28
Hobbs, L. M., Welty, D. E., & Thorburn, J. A. 1991, ApJ, 373, L47
Magain, P. 1987, A&A, 181, 323
Hobbs, L. M. & Thorburn, J. A. 1997, ApJ, 491, 772
Lambert, D. L., Heath, J. E., & Edvardsson, B. 1991, MNRAS, 253, 610
Norris, J. E., Ryan, S. G., & Stringfellow, G. S. 1994, ApJ, 423, 386
Norris, J. E., Ryan, S. G., Beers, T. C., & Deliyannis, C. P. 1997, ApJ, 485, 370
Pinsonneault, M. H., Deliyannis, C. P., and Demarque, P. 1992, ApJS, 78, 179
Pinsonneault, M. H., Walker, T. P., Steigman, G., & Narayanan, V. K. 1999, ApJ, 527, 180
Rebolo, R., Molaro, P., & Beckman, J. E. 1988, A&A, 192, 192
Romano, D., Matteucci, F., Molaro, P., & Bonifacio, P. 1999, A&A, in press
Ryan, S. G., Beers, T. C., Deliyannis, C. P., & Thorburn, J. A. 1996, ApJ, 458, 543
Ryan, S. G., Beers, T. C., Olive, K. A., Fields, B. D., & Norris, J. E. 2000a, ApJL, in press
Ryan, S. G., Beers, T. C., Kajino, T., & Rosolankova, K. 2000c, in prep
Ryan, S. G., Kajino, T., Beers, T. C., Suzuki, T., Rosolankova, K., & Romano, D. 2000b, in prep
Ryan, S. G., Norris, J. E., & Beers, T. C. 1998, ApJ, 506, 892
Ryan, S. G., Norris, J. E., & Beers, T. C. 1999, ApJ, 523, 654
Smith, V. V., Lambert, D. L., & Nissen, P.-E. 1993, ApJ, 408, 262
Smith, V. V., Lambert, D. L., & Nissen, P. E. 1998, ApJ, 506, 405
Spite, F. & Spite, M. 1982, A&A, 115, 357
Spite, F. & Spite, M. 1993, A&A, 279, L9
Spite, M., Molaro, P., Francois, P., & Spite, F. 1993, A&A, 271, L1
Spite, M., & Spite, F. 1981, IAU Coll. 68, Astrophysical Parameters for Globular Clusters, A. G. David Phillip & D. S. Haynes eds (Kluwer, Dordrecht)
Thorburn, J. A. 1992, ApJ, 399, L83
Thorburn, J. A. 1994, ApJ, 421, 318 |
warning/0001/hep-ph0001229.html | ar5iv | text | # 1 Contours of the MSSM contribution to the muon 𝑔-2 in the unit of 10⁻¹⁰ (𝛿𝑎_𝜇) on the (𝜇,𝑀₂)-plane for tan𝛽=50. (a) 𝑚_𝐿̃ = 𝑚_𝐸̃ = 200 GeV. (b) 𝑚_𝐿̃ = 𝑚_𝐸̃ = 500 GeV. The region enclosed by the dashed line gives Δ𝜒²=𝜒²_MSSM-𝜒²_SM < 0 for the precision electroweak data, while the region enclosed by the bold line satisfies Δ𝜒² < 1. The darkly shaded zone is excluded from the current lower bound on the chargino mass. The lightly shaded region is excluded by the currently allowed value for muon 𝑔-2 in eq. (). I and II on the graphs indicate Region I in Eq. () and Region II in Eq. () respectively.
| KEK-TH-671 |
| --- |
| SNS-PH/00-02 |
| hep-ph/0001229 |
Muon $`g2`$ and precision electroweak physics
in the MSSM
Gi-Chol Cho<sup>1),2)</sup>, Kaoru Hagiwara<sup>1)</sup> and Masashi Hayakawa<sup>1)</sup>
<sup>1)</sup>Theory Division, KEK, Tsukuba, Ibaraki 305-0801, Japan
<sup>2)</sup>Scuola Normale Superiore, Piazza dei Cavalieri 7, I-56126 Pisa, Italy
Abstract
The minimal supersymmetric extension of standard model (MSSM) is examined by analyzing its quantum effects on the precision electroweak measurements and the muon $`g2`$. We examine carefully the effects of light charginos and neutralinos that are found to improve the fit to the electroweak data. We identify two distinct regions on the $`(\mu ,M_2)`$-plane that fit well to the electroweak data and give significant contribution to muon $`g2`$.
The minimal supersymmetric standard model (MSSM) has been one of the extensively investigated theories beyond the standard model since it was recognized that it has the potential to unify strong and electroweak gauge interactions. On the other hand, despite enormous efforts to find its signature, there has been no direct evidence for the existence of the supersymmetric particles. In this letter we study quantitatively the MSSM predictions for the muon $`g2`$ in the light of the latest precision electroweak data , and that of the present and future $`g2`$ experiments .
The importance of the muon $`g2`$, conventionally denoted as $`a_\mu =\frac{1}{2}(g_\mu 2)`$, has been widely discussed in the context of supersymmetric theories . The observation of the effect from the MSSM is found generally accessible with the target accuracy of the current experiment at Brookhaven National Laboratory (BNL)
$$\mathrm{\Delta }a_\mu (\mathrm{expt})=4.0\times 10^{10}.$$
(1)
This accuracy is about $`1/20`$ of the error (numerals in the parenthesis) of the current measurement
$$a_\mu (\text{expt})=11659235(73)\times 10^{10}.$$
(2)
The prediction of the standard model (SM) is
$$a_\mu (\mathrm{SM})=\mathrm{11659\; 168.75}(9.56)\times 10^{10},$$
(3)
where the details of specific contributions is found in Ref. . We shall come back to the theoretical uncertainties of the SM prediction not included in the above quoted error of $`9.56\times 10^{10}`$ in the summary part.
As noted above the experimental results on the precision measurements around the $`Z`$-pole have played an essential role in revealing the possibility of grand unification. This implies the existence of many supersymmetric particles below the TeV scale and hence we may expect to observe their quantum effects in the precision electroweak experiments as well as in the muon $`g2`$. Although no signal of supersymmetry has so far been identified, recent systematic study of the precision electroweak data found that the existence of the relatively light ($``$ 100 GeV) charginos and neutralinos can make the MSSM fit slightly better than that of the standard model . We therefore study carefully the MSSM effects on the muon $`g2`$ in the range $`\left|\mu \right|`$, $`M_2`$ $`<`$ 500 GeV. It is found that analysis performed here could give rise to a systematic criterion to identify the preferred range of the MSSM parameters.
As we will see, the additional contribution to the muon $`g2`$ in the MSSM can be greater than the target accuracy (1) of the present experiment. Thus it is convenient to define the additional new physics contribution to the muon $`g2`$ in the unit of $`10^{10}`$
$$\delta a_\mu 10^{10}\times \left\{a_\mu (\mathrm{MSSM})a_\mu (\mathrm{SM})\right\}.$$
(4)
The present experimental data (2) can then be expressed as
$$16\delta a_\mu 149,$$
(5)
at the 1-$`\sigma `$ level. Here we add the theoretical uncertainty of 10 in the unit of $`10^{10}`$ in eq. (3) linearly to the present experimental error of 73 in eq. (2).
In the MSSM $`a_\mu `$ receives essentially two new contributions. One comes from the chargino ($`\stackrel{~}{\chi }_j^{},j=1,2`$) and muon-sneutrino ($`\stackrel{~}{\nu }_\mu `$) propagation in the intermediate states, and the other from the neutralino ($`\stackrel{~}{\chi }_j^0,j=14`$) and smuon ($`\stackrel{~}{\mu }_i,i=1,2`$) intermediate states. The chargino-sneutrino contribution can be expressed as
$`a_\mu (\stackrel{~}{\chi }^{})`$ $`=`$ $`{\displaystyle \frac{1}{8\pi ^2}}{\displaystyle \frac{m_\mu }{m_{\stackrel{~}{\nu }_\mu }}}{\displaystyle \underset{j=1}{\overset{2}{}}}\{{\displaystyle \frac{m_\mu }{m_{\stackrel{~}{\nu }_\mu }}}G_1\left({\displaystyle \frac{m_{\stackrel{~}{\chi }_j^{}}^2}{m_{\stackrel{~}{\nu }_\mu }^2}}\right)\left(\right|g_L^{\stackrel{~}{\chi }_j^{}\mu \stackrel{~}{\nu }_\mu }|^2+\left|g_R^{\stackrel{~}{\chi }_j^{}\mu \stackrel{~}{\nu }_\mu }|^2\right)`$
$`+{\displaystyle \frac{m_{\stackrel{~}{\chi }_j^{}}}{m_{\stackrel{~}{\nu }_\mu }}}G_3\left({\displaystyle \frac{m_{\stackrel{~}{\chi }_j^{}}^2}{m_{\stackrel{~}{\nu }_\mu }^2}}\right)\mathrm{Re}\left[\left(g_R^{\stackrel{~}{\chi }_j^{}\mu \stackrel{~}{\nu }_\mu }\right)^{}g_L^{\stackrel{~}{\chi }_j^{}\mu \stackrel{~}{\nu }_\mu }\right]\},`$
$`G_1(x)`$ $`=`$ $`{\displaystyle \frac{1}{12(x1)^4}}\left[(x1)(x^25x2)+6x\mathrm{ln}x\right],`$ (6b)
$`G_3(x)`$ $`=`$ $`{\displaystyle \frac{1}{2(x1)^3}}\left[(x1)(x3)+2\mathrm{ln}x\right],`$ (6c)
while the neutralino-smuon contribution as
$`a_\mu (\stackrel{~}{\chi }^0)`$ $`=`$ $`{\displaystyle \frac{1}{8\pi ^2}}{\displaystyle \underset{i=1}{\overset{2}{}}}{\displaystyle \frac{m_\mu }{m_{\stackrel{~}{\mu }_i}}}{\displaystyle \underset{j=1}{\overset{4}{}}}\{{\displaystyle \frac{m_\mu }{m_{\stackrel{~}{\mu }_i}}}G_2\left({\displaystyle \frac{m_{\stackrel{~}{\chi }_j^0}^2}{m_{\stackrel{~}{\mu }_i}^2}}\right)\left(\right|g_L^{\stackrel{~}{\chi }_j^0\mu \stackrel{~}{\mu }_i}|^2+\left|g_R^{\stackrel{~}{\chi }_j^0\mu \stackrel{~}{\mu }_i}|^2\right)`$ (7a)
$`+{\displaystyle \frac{m_{\stackrel{~}{\chi }_j^0}}{m_{\stackrel{~}{\mu }_i}}}G_4\left({\displaystyle \frac{m_{\stackrel{~}{\chi }_j^0}^2}{m_{\stackrel{~}{\mu }_i}^2}}\right)\mathrm{Re}\left[\left(g_R^{\stackrel{~}{\chi }_j^0\mu \stackrel{~}{\mu }_i}\right)^{}g_L^{\stackrel{~}{\chi }_j^0\mu \stackrel{~}{\mu }_i}\right]\},`$
$`G_2(x)`$ $`=`$ $`{\displaystyle \frac{1}{12(x2)^4}}\left[(x1)(2x^2+5x1)6x^2\mathrm{ln}x\right],`$ (7b)
$`G_4(x)`$ $`=`$ $`{\displaystyle \frac{1}{2(x1)^3}}\left[(x1)(x+1)2x\mathrm{ln}x\right].`$ (7c)
Here we adopt the notation of Ref. for the coupling constants
$$=\underset{\alpha =L,R}{}g_\alpha ^{F_1F_2S}\overline{F}_1P_\alpha F_2S,$$
(8)
where $`F_1`$ and $`F_2`$ are four-component fermion fields, $`S`$ denotes a scalar field, and
$$P_L=\frac{1\gamma _5}{2},P_R=\frac{1+\gamma _5}{2}.$$
(9)
The charged Higgs boson contribution, which gives rise to a sizable effect in the case of $`bs\gamma `$ transition, is highly suppressed due to the small Yukawa couplings in the muon $`g2`$. The MSSM contribution to the muon $`g2`$ is most significant at large $`\mathrm{tan}\beta `$ . The chargino-sneutrino loop, eq. (S0.EGx1), then dominates over the other in almost all region of the $`(\mu ,M_2)`$ plane. However the neutralino-smuon loop contribution to $`\delta a_\mu `$ is also larger than the target accuracy (1) of the current experiment. Except for $`(\mu ,M_2,\mathrm{tan}\beta )`$ which determines the chargino and neutralino masses, the former depends on the left-handed SUSY breaking slepton mass, $`m_{\stackrel{~}{L}}`$, while the latter also depends on the right-handed one, $`m_{\stackrel{~}{E}}`$, in addition. We discuss this point in more detail when we study the dependence of $`\delta a_\mu `$ on $`m_{\stackrel{~}{E}}`$.
In our analysis we allow the MSSM parameters to vary freely restricted only by experimental constraints from direct and indirect searches. When restricted to specific models of supersymmetry (SUSY) breaking such as the minimal supergravity scenario, or the gauge mediated supersymmetry breaking scenario, we reproduce the known results . First we consider the case when all the squarks are so heavy that their effects on low-energy physics disappears. In our actual numerical calculation, we set all the squark masses and the extra Higgs boson masses at 1 TeV such that they do not make worse the fit to the electroweak data and the $`bs\gamma `$ rate .
The constraints on the MSSM parameter space from the electroweak experiments have been found in Ref. by using the 18 $`Z`$-pole data given by the LEP/SLC experiments and the $`W`$-boson mass $`m_W`$ given by the Tevatron and LEP2 experiments . The electroweak observables are affected by the supersymmetric particles through both the universal gauge boson propagator corrections and the process specific vertex/box corrections. Following the formalism introduced in Ref. , the universal part of the radiative corrections of the $`Z`$-pole observables can be represented by two oblique parameters $`S_Z`$ and $`T_Z`$, while the $`W`$-boson mass $`m_W`$ itself can be adopted as the third oblique parameter. The parameters $`S_Z`$ and $`T_Z`$ are expressed as the sum of the conventional $`S`$ and $`T`$ parameters and the $`R`$ parameter which measures the running effect of the $`Z`$-boson propagator correction between $`q^2=0`$ and $`q^2=m_Z^2`$. The current electroweak data favors new physics that gives negative contribution both to the $`S_Z`$ and $`T_Z`$ parameters, and hence they constrain strongly the additional positive contributions to these parameters in the MSSM. The squark contributions to the $`T_Z`$ parameter are generally found to be positive while they do not affect to the $`S_Z`$ parameter so much. The light charginos and neutralinos are found to give negative contributions to the $`R`$ parameter and hence they make both the $`S_Z`$ and $`T_Z`$ parameters negative, thereby improving the fit to the electroweak data. Light left-handed sleptons make $`S_Z`$ negative but keep $`T_Z`$ essentially unchanged. It has been shown that if the chargino mass is close to its lower mass bound from the LEP2 experiment, $`m_{\stackrel{~}{\chi }_1^{}}100`$ GeV, and the squarks and sleptons are sufficiently heavy, say 1 TeV, the total $`\chi ^2`$ in the MSSM is slightly better than that in the standard model :
$$\mathrm{\Delta }\chi ^2\chi _{\mathrm{MSSM}}^2\chi _{\mathrm{SM}}^21.$$
(10)
No other combinations of light supersymmetric particles are found to improve the fit to the electroweak data over the standard model.
Summing up, the precision electroweak data favors light charginos and neutralinos but disfavors light left-handed squarks and sleptons. The muon $`g2`$ is found to be sensitive to the left-handed slepton mass when there are light charginos and neutralinos. If the left-handed sleptons are relatively light, then the MSSM contribution to $`g2`$ grows but the electroweak fit worsens. Below we examine quantitatively the MSSM prediction for the muon $`g2`$ and the electroweak data, and look for the region of the MSSM parameter space which gives observable $`g2`$ effect without spoiling the good fit to the electroweak data.
Taking this circumstances in mind, we show in Fig. 1 the contours of fixed values of $`\delta a_\mu `$ defined in Eq. (4) on the ($`\mu `$, $`M_2`$)-plane for a relatively large value of $`\mathrm{tan}\beta =50`$. We set $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 200 GeV in Fig. 1(a) and $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 500 GeV in Fig. 1(b), respectively. For the gaugino masses, we adopt the “unification” condition, $`M_1/M_2`$ = $`\alpha _1(M_Z)/\alpha _2(M_Z)`$, for simplicity. However, the general aspect of our conclusion obtained here persists as long as the gaugino masses share a common order of magnitude. The shaded-region shows the region of the parameters already inconsistent with the mass bound for chargino
$$m_{\stackrel{~}{\chi }_1^{}}>93\mathrm{GeV}.$$
(11)
The physical slepton masses are obtained as the eigenvalues of the slepton mass-squared matrix for each flavor ($`l`$ = $`e`$, $`\mu `$ or $`\tau `$) :
$`M_{\stackrel{~}{l}}^2=\left(\begin{array}{cc}M_{LL}^2& M_{LR}^2\\ M_{RL}^2& M_{RR}^2\end{array}\right),`$ (12c)
$`M_{LL}^2=m_{\stackrel{~}{L}}^2+m_l^2M_Z^2\mathrm{cos}2\beta \left({\displaystyle \frac{1}{2}}\mathrm{sin}^2\theta _W\right),`$ (12d)
$`M_{RR}^2=m_{\stackrel{~}{E}}^2+m_l^2M_Z^2\mathrm{cos}2\beta \mathrm{sin}^2\theta _W,`$ (12e)
$`M_{LR}^2=(M_{RL}^2)^{}=m_l(A_l^{}\mu \mathrm{tan}\beta ),`$ (12f)
where $`m_l`$ denotes the charged-lepton mass. In Fig. 1 we show the contours of muon $`g2`$ when $`A_\mu `$ is set equal to zero. Effects of nonzero $`A_\mu `$ are discussed below.
Fig. 1(a) shows a generic feature that the MSSM contributions can be as large as $`(100300)\times 10^{10}`$ when $`\mathrm{tan}\beta `$ = 50 and $`m_{\stackrel{~}{L}}`$ = 200 GeV, which is much greater than the $`W`$\- and $`Z`$-contributions in the SM ,
$$a_\mu (\text{weak})=15.1(4)\times 10^{10}.$$
(13)
This is essentially because of the large Yukawa coupling for the $`\mu _R\stackrel{~}{h}^{}\stackrel{~}{\nu }_\mu `$ vertex at large $`\mathrm{tan}\beta `$ $``$ 1. Here $`\mu _R`$ is the right-handed muon component and $`\stackrel{~}{h}^{}`$ is the charged component of the higgsino which couples to charged leptons and down-type quarks. The magnitude of the MSSM predictions reduces to the level of the SM weak corrections (13) at around $`\mathrm{tan}\beta =3`$, as shown in Fig. 2. The MSSM contribution to the muon $`g2`$ scales roughly as $`r_{\stackrel{~}{L}}/m_{\stackrel{~}{L}}^2+r_{\stackrel{~}{E}}/m_{\stackrel{~}{E}}^2`$ in the large $`\mathrm{tan}\beta `$ case, where $`r_{\stackrel{~}{L}}`$ and $`r_{\stackrel{~}{E}}`$ are some constants. The dependence on $`m_{\stackrel{~}{L}}`$ and $`m_{\stackrel{~}{E}}`$ will be further evaluated below.
Also shown in the figures is the region where the fit to the electroweak data is better than that of the SM (the narrow region enclosed by the dashed lines). The region falls mostly into the region forbidden by the chargino mass bound (the darkly shaded regions). The region enclosed by the bold lines gives reasonably good fit to the electroweak data, where $`\mathrm{\Delta }\chi ^2`$ in Eq. (10) is less than one. We find that there are essentially two regions which respect the result from the precision electroweak measurements. One region is
$$\mathrm{Region}\mathrm{I}:120\mathrm{GeV}<\mu <c_1M_2,$$
(14)
where $`c_1`$ is a positive number greater than one, depending on the slepton masses. The other minor region appears only when $`m_{\stackrel{~}{L}}`$ is taken to be larger than about 500 GeV (see Fig. 1(b)):
$$\mathrm{Region}\mathrm{II}:\mathrm{\hspace{0.17em}100}\mathrm{GeV}<M_2<150\mathrm{GeV}.$$
(15)
The lightest chargino is generally higgsino-dominated in Region I while it is gaugino-dominated in Region II. Both domains correspond to the extreme regime in which the chargino mass becomes nearly equal to the experimental lower bound, around which the chargino-neutralino contributions to the oblique parameters pull both $`S_Z`$ and $`T_Z`$ back to negative directions. This negative contribution to $`\mathrm{\Delta }\chi ^2`$ is necessary to complement the positive contribution (which worsens the fit) due to relatively small $`m_{\stackrel{~}{L}}`$. Thus the MSSM with large $`\mathrm{tan}\beta `$ gives a sizable contribution to the muon $`g2`$ if $`\left|\mu \right|`$, $`M_2`$ and $`m_{\stackrel{~}{L}}`$ are all smaller than 500 GeV. Actually the present muon $`g2`$ data (5) already excludes the negative $`\mu `$ region of Fig. 1.
The muon $`g2`$ is a less powerful probe of the MSSM at smaller $`\mathrm{tan}\beta `$. Fig. 2 shows the similar graphs as Fig. 1, but for $`\mathrm{tan}\beta =3`$. As in Fig. 1(a) and (b), $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 200 GeV in Fig. 2(a), and $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 500 GeV in Fig. 2(b). Here we are interested in whether we can find some evidence of the MSSM with small $`\mathrm{tan}\beta `$ through the measurement of the muon $`g2`$ in view of the other precise low energy data. Fig. 2(a) shows that the region with positive $`\mu `$ and small $`M_2`$ for small slepton mass ($`m_{\stackrel{~}{L}}`$ = 200 GeV) is accessible by the present muon $`g2`$ experiment with its target precision (1), while respecting the result of the electroweak experiments. Fig. 2(b) shows that new two allowed regions appear on the negative $`\mu `$ side of the $`(\mu ,M_2)`$-plane for slepton masses larger than about 500 GeV. Those two regions are
$`\mathrm{Region}\mathrm{III}:c_2M_2<\mu <80\mathrm{GeV},`$ (16a)
$`\mathrm{Region}\mathrm{IV}:80\mathrm{GeV}<M_2<c_3\mu ,\mu <120\mathrm{GeV},`$ (16b)
where $`c_2`$ and $`c_3`$ are positive numbers depending on $`m_{\stackrel{~}{L}}`$ and $`m_{\stackrel{~}{E}}`$. The MSSM correction to the muon $`g2`$ is small compared to the target precision (1) in those regions in Fig. 2(b). However, when $`m_{\stackrel{~}{E}}`$ $`<`$ $`m_{\stackrel{~}{L}}`$, the observable enhancement might be expected in Region III and IV, as will be shown in Fig. 3(b).
Next we study the MSSM prediction to the muon $`g2`$ without assuming the universality between $`m_{\stackrel{~}{L}}`$ and $`m_{\stackrel{~}{E}}`$. In Fig. 3 we show the $`m_{\stackrel{~}{L}}`$-dependence of the maximally admissible $`\delta a_\mu `$ for (a) $`\mathrm{tan}\beta =50`$ and (b) $`\mathrm{tan}\beta =3`$ when $`A_\mu =0`$. In each figure, the solid, dotted and dashed lines represent $`m_{\stackrel{~}{E}}=100`$ GeV, 300 GeV and 500 GeV, respectively. We find that, for $`\mathrm{tan}\beta `$ = 50, the maximum of $`\delta a_\mu `$ is achieved when $`M_2/\mu `$ is nearly equal to $`\pm 1`$. For $`\mathrm{tan}\beta `$ = 3, this property also holds as long as $`m_{\stackrel{~}{L}}`$ is small enough (less than about 500 GeV).
It should be remarked that the precision measurements around $`Z`$-pole favor large left-handed slepton mass but is rather blind to the right-handed slepton mass. Fig. 3 tells us that the MSSM prediction of the muon $`g2`$ has sizable $`m_{\stackrel{~}{E}}`$-dependence when the target accuracy (1) of the current $`g2`$ experiment is taken into account. Let us recall that the chargino-sneutrino loop contribution (S0.EGx1) depends on $`m_{\stackrel{~}{L}}`$ while the neutralino-smuon loop contribution (S0.EGx2) depends on both $`m_{\stackrel{~}{L}}`$ and $`m_{\stackrel{~}{E}}`$. The numerical study shows that the chargino-sneutrino loop contribution decreases faster than the neutralino-smuon loop contribution if $`m_{\stackrel{~}{L}}`$ is increased while $`m_{\stackrel{~}{E}}`$ is kept fixed. Thus $`\delta a_\mu `$ is more sensitive to relatively small $`m_{\stackrel{~}{E}}`$ for larger $`m_{\stackrel{~}{L}}`$. This is because only the neutralino-smuon loop contribution depends on $`m_{\stackrel{~}{E}}`$ and the lighter smuon is almost the right-handed slepton component for $`m_{\stackrel{~}{E}}`$ $``$ $`m_{\stackrel{~}{L}}`$. The neutralino-smuon loop gives nonzero contribution to $`\delta a_\mu `$ even for $`m_{\stackrel{~}{L}}\mathrm{}`$. This explains why $`\delta a_\mu `$ does not decouple in Fig. 3 at large $`m_{\stackrel{~}{L}}`$. Indeed, for $`\mathrm{tan}\beta `$ = 50, $`m_{\stackrel{~}{E}}`$ = 100 GeV and $`m_{\stackrel{~}{L}}`$ $``$ $`\mathrm{}`$, the maximally possible $`\delta a_\mu `$ is 65, which is observable in the light of the precision (1) for the muon $`g2`$ measurement. Fig. 3 shows that, on account of the constraint on $`\delta a_\mu `$ (5), there is a possibility that the MSSM with small $`\mathrm{tan}\beta `$ for negative $`\mu `$ and $`m_{\stackrel{~}{E}}`$ $`<`$ $`m_{\stackrel{~}{L}}`$ can be probed by the current muon $`g2`$ experiment with the target precision (1).
So far, all our results have been shown for $`A_\mu `$ = 0. We discuss the dependence of these results on $`A_\mu `$ here. We first recall that $`A_\mu `$ enters only in the left-right mixing of the smuon mass-squared matrix (S0.EGx3) for all the electroweak precision observables and the muon $`g2`$ at one-loop order. Thus $`A_\mu `$-dependence comes from the neutralino-smuon loop contribution. The numerical analysis shows that $`\delta a_\mu `$ varies at most 5 when $`A_\mu `$ varies from $``$1 to 1 TeV for the various choices of $`m_{\stackrel{~}{L}}`$, $`m_{\stackrel{~}{E}}`$, $`M_2`$ and $`\mu `$ in the allowed region at both the large and small $`\mathrm{tan}\beta `$. Thus our findings in Fig. 1 and Fig. 3(a) remain valid up to the target precision (2) of the current muon $`g2`$ experiment even with nonzero $`A_\mu `$. In particular the slepton mass dependence does not change quantitatively. For small $`\mathrm{tan}\beta `$, the impact of nonzero $`A_\mu `$ is comparable to the magnitude of $`\delta a_\mu `$ in $`A_\mu =0`$ case. Then, our discussion on the small $`\mathrm{tan}\beta `$ case with $`A_\mu =0`$ might be modified depending on the magnitude and the sign of $`A_\mu `$. Note that, if $`\mathrm{tan}\beta `$ is small, $`A_\mu `$ could affect $`\delta a_\mu `$ when the relative sign between $`A_\mu `$ and $`\mu `$ is opposite. This is because $`A_\mu `$ always appears as a linear combination with $`\mu \mathrm{tan}\beta `$ in the slepton mass-squared matrix element (S0.EGx3). We find that $`\delta a_\mu `$ for positive $`\mu `$ could be at most 15 if $`A_\mu `$ is negative and large, say $`1`$ TeV.
To summarize the paper we performed the quantitative analysis on the MSSM contribution to the muon $`g2`$ on account of the constraint from the precision electroweak data. The analysis on the precision electroweak measurements tells us that the fit of the MSSM might be better than that of the SM when the chargino mass is close to the LEP2 bound ($``$ 100 GeV) and the left-handed sfermions are heavy enough, say a few hundred GeV. Our study has been done by taking account of this point and the target precision $`4\times 10^{10}`$ in (2) of the current muon $`g2`$ experiment.
At first, assuming the universality between the left- and the right-handed SUSY breaking mass parameters, $`m_{\stackrel{~}{L}}=m_{\stackrel{~}{E}}`$, we found four regions on the $`(\mu ,M_2)`$-plane where the fit to the electroweak data is not worse as compared to the SM and the predicted MSSM contribution to the muon $`g2`$ is allowed from the current experimental result. We found that the muon $`g2`$ is significantly enhanced (of the order 100 $`\times 10^{10}`$) in Region I (120 GeV $`<`$ $`\mu `$ $`<`$ $`c_1M_2`$, with $`c_1`$ $`>`$ 1) for $`\mathrm{tan}\beta `$ = 50 and $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 200 GeV (see Fig. 1(a)). For the heavy slepton mass, $`m_{\stackrel{~}{L}}`$ = $`m_{\stackrel{~}{E}}`$ = 500 GeV, and $`\mathrm{tan}\beta `$ = 50, Region II (100 GeV $`<`$ $`M_2`$ $`<`$ 150 GeV) appears in addition, where the predicted value of the muon $`g2`$ is allowed in the current experiment but the MSSM correction is at most about 60 $`\times `$ $`10^{10}`$ in both Region I and II. (See Fig. 1(b).)
We also examined the dependence on $`m_{\stackrel{~}{L}}`$ and $`m_{\stackrel{~}{E}}`$ of those results without universality assumption. We found that $`m_{\stackrel{~}{E}}`$-dependence of the muon $`g2`$ is larger than its target precision of the present experiment in the large $`\mathrm{tan}\beta `$ case. Thus the muon $`g2`$ complements the precision electroweak measurement which is rather blind to $`m_{\stackrel{~}{E}}`$. It might be a unique possible indirect probe to give constraints on the right-handed slepton. When $`m_{\stackrel{~}{E}}`$ $`<`$ $`m_{\stackrel{~}{L}}`$, the MSSM with small $`\mathrm{tan}\beta `$ is also accessible by the current $`g2`$ experiment (See Fig. 3(b)). The model falls in Region III (16a) or in Region IV (16b) on the $`(\mu ,M_2)`$-plane for $`\mu <0`$. It is also found that the MSSM contribution to the muon $`g2`$ is always affected by at most $`5\times 10^{10}`$ from the $`A_\mu `$-term. There is no essential change in the above results in the large $`\mathrm{tan}\beta `$ case. For negative $`A_\mu `$ of the order 1 TeV, the MSSM with the small $`\mathrm{tan}\beta `$ in Region I and II, rather than Region III and IV, can be probed by the muon $`g2`$ measurement. The detectable limit of the chargino at Tevatron Run-II is expected to be 250 GeV . If the chargino is found at Tevatron, the muon $`g2`$ will enable us to select the parameters of the MSSM .
Here we remind the reader that the allowed range (5) for $`\delta a_\mu `$ is derived from (2) and (3). If the result in Ref. , which utilizes the accurate data for hadronic decay of $`\tau `$ (See Ref. and the references therein) further with the help of quark-hadron duality in evaluation of dispersion integral, is applied instead of the result in Ref. , the leading order QCD correction involved in the above quantity receives $`(89)\times 10^{10}`$ modification, resulting in
$$a_\mu (\mathrm{SM})=\mathrm{11659\; 160.05}(6.44)\times 10^{10}.$$
(17)
This could then change the allowed range of $`\delta a_\mu `$ from (5) to
$$4<\delta a_\mu <154.$$
(18)
In so far as we illustrate the tendency of the excluded region at large $`\mathrm{tan}\beta `$ with the current experimental value (2), the difference between (5) and (18) does not matter. But such a difference will become crucial to find appropriate sets of ($`\mu `$, $`M_2`$) when the experimental uncertainty is reduced to the level (1). This aspect is quite interesting from the view point of searching the MSSM at large $`\mathrm{tan}\beta `$. Thus the reduction of the error residing in QCD corrections to the muon $`g2`$ is a necessary task to be put forward promptly.
Acknowledgment
The authors thank S. Eidelman for the various usuful comments on the manuscript. |
warning/0001/cond-mat0001325.html | ar5iv | text | # Mobile Bipolaron
\[
## Abstract
We explore the properties of the bipolaron in a 1D Holstein-Hubbard model with dynamical quantum phonons. Using a recently developed variational method combined with analytical strong coupling calculations, we compute correlation functions, effective mass, bipolaron isotope effect and the phase diagram. The two site bipolaron has a significantly reduced mass and isotope effect compared to the on-site bipolaron, and is bound in the strong coupling regime up to twice the Hubbard $`U`$ naively expected. The model can be described in this regime as an effective t-J-V model with nearest neighbor repulsion. These are the most accurate bipolaron calculations to date.
\]
While there is a generally accepted belief that in high $`T_c`$ superconductors a dominantly electronic interaction is responsible for the unusually high transition temperatures, the interplay between the electron-phonon interaction and the strong electron-electron interaction nevertheless plays a significant role in determining the physical properties of these strongly correlated systems . Although the study of lattice effects in strongly correlated materials is steadily growing , the understanding of the influence of the Hubbard interaction $`U`$ on bipolaron formation is still incomplete. In particular, it is known that in the strong coupling regime bipolarons have an extremely large effective mass, which represents one of the main objections against the theory of small bipolaron superconductivity . Recent calculations in the adiabatic (static phonon) limit show that a first-order phase transition exists between the on-site (S0) and lighter neighboring-site (S1) bipolaron . The properties of the S1 bipolaron were also investigated by variational and exact diagonalization methods .
In this letter we present accurate numerical solutions of the Holstein-Hubbard model bipolaron on a 1D infinite lattice. Our results are exact to within the line-widths on the figures in the intermediate and near the strong coupling regimes. The results are compared to analytical calculations in the strong coupling regime.
We consider the Holstein-Hubbard Hamiltonian
$`H`$ $`=t{\displaystyle \underset{js}{}}(c_{j+1,s}^{}c_{j,s}+H.c.)`$ (2)
$`g\omega {\displaystyle \underset{js}{}}c_{j,s}^{}c_{j,s}(a_j+a_j^{})+\omega {\displaystyle \underset{j}{}}a_j^{}a_j+U{\displaystyle \underset{j}{}}n_jn_j,`$
where $`c_{j,s}^{}`$ creates an electron of spin $`s`$ and $`a_j^{}`$ creates a phonon on site $`j`$. The last term in Eq. (2) represents the on-site Coulomb repulsion. We consider the case where two electrons with opposite spins couple to dispersionless optical phonons.
Basis states for the many-body Hilbert space can be written $`|M=|j_1,j_2;\mathrm{},n_m,n_{m+1},\mathrm{},`$, where the up and down electrons are on sites $`j1`$ and $`j2`$, and there are $`n_m`$ phonons on site $`m`$. A variational subspace is constructed beginning with an initial state where both electrons are on the same site with no phonons, and operating repeatedly ($`N_h`$-times) with the off-diagonal pieces ($`t`$ and $`g`$) of the Hamiltonian, Eq. (2). All translations of these states are included on an infinite lattice. This method was used previously for the polaron (one electron) . The method is very efficient in the intermediate coupling regime, where it provides results that are variational in the thermodynamic limit and bipolaron energies that are accurate to 7 digits for the case $`N_h=18`$ and size of the Hilbert space $`N=2.2\times 10^6`$ phonon and down electron configurations for a given up electron position.
Before presenting the numerical calculations, we show that many interesting properties of the bipolaron can be found in second-order strong coupling perturbation theory. Following Lang and Firsov we use the canonical transformation $`\stackrel{~}{H}=e^SHe^S`$, where $`S=g_{js}n_{js}(a_ja_j^{})`$. The transformed Hamiltonian takes the form
$`\stackrel{~}{H}`$ $`=`$ $`H_0+T`$ (3)
$`H_0`$ $`=`$ $`\omega {\displaystyle \underset{j}{}}a_j^{}a_j\omega g^2{\displaystyle \underset{j}{}}n_j+\left(U2\omega g^2\right){\displaystyle \underset{j}{}}n_jn_j`$ (4)
$`T`$ $`=`$ $`te^{g^2}{\displaystyle \underset{js}{}}(c_{j+1,s}^{}c_{j,s}e^{g\left(a_{j+1}^{}a_j^{}\right)}e^{g\left(a_{j+1}a_j\right)}+\mathrm{H}.\mathrm{c}.),`$ (5)
where $`n_j=n_j+n_j`$. The first term in $`H_0`$ is the energy of the phonon excitations, and the second is the energy gained by displacing the oscillator in the force of the electron. The exponential factors in the $`T`$ hopping term arise because the Lang-Firsov transformation redefines the origin of the harmonic oscillator when the number of electrons on the site changes. In the limit $`g0`$ and $`\omega \mathrm{}`$ with $`\omega g^2`$ constant, the phonon interaction is instantaneous and the Holstein-Hubbard model maps onto a Hubbard model with an effective Hubbard interaction $`\stackrel{~}{U}=U2\omega g^2`$. In 1D a bound state exists for $`\stackrel{~}{U}<0`$.
In the strong coupling or antiadiabatic limit, $`T`$ in Eq. (3) is considered a perturbation. It represents the hopping of electrons, including possible creation and destruction of phonon excitations. The S0 state $`\varphi _0=c_0^{}c_0^{}|0`$ has the lowest energy to zeroth order in $`T`$ when $`\stackrel{~}{U}<0`$. In perturbation theory to second order, the energy of the S0 bipolaron is
$`E_{bi}^{S0}(k)`$ $`=`$ $`U4\omega g^2`$ (6)
$``$ $`{\displaystyle \frac{4t^2e^{2g^2}}{\omega }}{\displaystyle \underset{n=0}{}}{\displaystyle \frac{(2g^2)^n}{n!}}{\displaystyle \frac{(1+(1)^n\mathrm{cos}(k))}{\stackrel{~}{U}/\omega +n}}.`$ (7)
The effective mass $`m_{S0}^1=^2E_{bi}(k)/k^2`$ can be obtained from Eq. (7) by calculating the infinite sums
$$m_{S0}^{}{}_{}{}^{1}=\frac{4t^2}{\omega }e^{\left(2g^2\frac{\stackrel{~}{U}}{\omega }\mathrm{ln}2g^2\right)}\left[\mathrm{\Gamma }\left(\frac{\stackrel{~}{U}}{\omega }\right)\mathrm{\Gamma }(\frac{\stackrel{~}{U}}{\omega },2g^2)\right],$$
(8)
where $`\mathrm{\Gamma }(x)`$ and $`\mathrm{\Gamma }(a,x)`$ are the gamma and incomplete gamma functions respectively. Taking the large $`g`$ limit in Eq. (8), one finds (see also Ref. ) $`m_{SO}^14\sqrt{\pi }t^2\mathrm{exp}(4g^2)/(\omega g)`$ for $`U=0`$. At large $`g`$, $`m_{S0}`$ is roughly a factor $`\mathrm{exp}(3g^2)`$ larger than the polaron mass, $`m_{po}^12t\mathrm{exp}(g^2)`$.
One would naively expect that within the strong coupling approximation, a bipolaron unbinds when $`\stackrel{~}{U}0`$. This is false: a bound bipolaron may exist even for $`\stackrel{~}{U}0`$. In this regime, a set of degenerate states $`\varphi _i=c_0^{}c_i^{}|0`$ for $`i0`$, written in a translationally invariant form, represents states with minimum energy of $`H_0`$. The energy of a S1 bipolaron is obtained by solving the secular equation $`|T_{ij}\delta _{ij}E_{bi}^{S1}|=0`$, where matrix elements $`T_{ij}=\varphi _i|T|\varphi _j`$ are calculated to second order in $`T`$.
The main source of binding arises because the second order matrix element $`T_{11}`$ for two neighboring electrons $``$ to exchange sites is exponentially larger than all other off-diagonal matrix elements. While in the limit $`g\mathrm{}`$, $`T_{11}t^2/U`$, the second largest (first order) matrix elements $`T_{i,i+1}t\mathrm{exp}(g^2)`$ for $`i\{0,1\}`$. The exponential suppression arises from phonon rearrangement whenever the initial and final charge distributions differ. In the singlet configuration, $`\varphi _1^{S=0}=(\varphi _1+\varphi _1)/\sqrt{2}`$, the diagonal correction to the energy is given by $`T_{11}+T_{11}`$. There is also a contribution to $`T_{11}`$ that resembles a retardation effect, in which one electron hops creating one or more phonon quanta on its initial site, and the second electron follows absorbing the phonons. This process, however, decreases exponentially with $`g`$ as $`t^2\mathrm{exp}(g^2)/(\omega g^2)`$, and is not strong enough to bind two polarons in the triplet configuration.
To obtain the S1 bipolaron stability criterion, we note that the secular equation for fixed center of mass momentum $`k`$ can be mapped onto a tight-binding model for a linear chain. Since all the off-diagonal matrix elements except $`T_{11}`$ scale as $`\mathrm{exp}(g^2)`$, the condition for a bound state is given by $`T_{11}+T_{11}<T_{ii}`$. Keeping only terms of order $`t^2/(\omega g^2)`$ and setting $`k=0`$, the matrix elements can be expressed as $`T_{11}=T_a+T_b`$, $`T_{11}=T_b`$, and $`T_{ii}=2T_a;i>1`$, with
$`T_a`$ $`=`$ $`{\displaystyle \frac{2t^2e^{2g^2}}{\omega }}{\displaystyle \underset{n=1}{}}{\displaystyle \frac{(2g^2)^n}{n!n}}{\displaystyle \frac{t^2}{\omega g^2}},`$ (9)
$`T_b`$ $`=`$ $`{\displaystyle \frac{2t^2e^{2g^2}}{\omega }}{\displaystyle \underset{n=0}{}}{\displaystyle \frac{(2g^2)^n}{n!}}{\displaystyle \frac{1}{\frac{\stackrel{~}{U}}{\omega }+n}}{\displaystyle \frac{2t^2}{U}},`$ (10)
where the final terms refer to the large $`g`$ limit. In the large $`g`$ limit, binding occurs for $`U<4\omega g^2`$.
The effective S1 bipolaron mass is computed by approximating the S1 bipolaron wavefunction with only $`\varphi _1^{S=0}`$ (omitting the exponential tail),
$`m_{S1}^1`$ $``$ $`{\displaystyle \frac{2t^2e^{2g^2}}{\omega }}\left[{\displaystyle \frac{2\omega }{\stackrel{~}{U}}}+{\displaystyle \underset{n=1}{}}{\displaystyle \frac{g^{2n}}{n!}}\left({\displaystyle \frac{2}{\stackrel{~}{U}/\omega +n}}+{\displaystyle \frac{1}{n}}\right)\right]`$ (11)
$``$ $`2t^2e^{g^2}\left[{\displaystyle \frac{2}{U\omega g^2}}+{\displaystyle \frac{1}{\omega g^2}}\right].`$ (12)
There are three distinct processes contributing to $`m_{S1}`$: an S1 pair can move by one lattice site through an intermediate doubly occupied state with $`n`$ phonons (terms that contain $`U`$ in Eq. (12)), or through an intermediate state with only phonon degrees of freedom (terms without $`U`$).
The bipolaron isotope effect is a measure of how the bipolaron mass varies with the ion mass $`M`$, $`\alpha d\mathrm{ln}m_{bi}/d\mathrm{ln}M`$. This can also be written $`\alpha =\frac{1}{2}d\mathrm{ln}m_{bi}/d\mathrm{ln}\omega `$, where the derivative is taken with $`\omega g^2`$ held constant. The bipolaron isotope effect $`\alpha `$ is equal to the superconductivity isotope effect $`\alpha _{SC}d\mathrm{ln}T_c/d\mathrm{ln}M`$ only in the low density limit, when superconductivity occurs as a weakly interacting gas of bipolarons Bose condenses. Then in mean field theory (ignoring fluctuations that are important in low dimensions), the transition temperature $`T_c`$ at fixed density is proportional to $`m_{bi}^1`$, and $`\alpha _{SC}=\alpha `$. The bipolaron isotope effect is expected to change substantially between the S0 and S1 regimes. Indeed, using Eqs. (8, 12) in the strong coupling limit, we obtain $`\alpha _{S0}2g^2\frac{1}{4}`$ for $`U=0`$ and $`\alpha _{S0}`$ decreasing with increasing $`U`$. In the S1 regime, $`\alpha _{S1}g^2/2`$ and is only weakly dependent on $`U`$. The bipolaron isotope effect has also been calculated by Alexandrov ; the superconductivity isotope effect has been measured by many groups .
We now present numerical variational results, using units where hopping $`t=1`$. Fig. (1) shows the ground state electron-electron density correlation function $`C(ij)=\psi _o|n_in_j|\psi _o`$, where $`n_i=n_i+n_i`$. At $`g=1`$, the bipolaron widens with increasing $`U`$ and transforms into two unbound polarons (which can only move a finite distance apart in the variational space). The value $`U=1.5`$ is below the transition to the unbound state at $`U_c=2.17`$, calculated by comparing the polaron and bipolaron energies. At $`U=1.5`$, $`C(i)`$ falls off exponentially, while for $`U>U_c`$ the typical distance between electrons is the order of the maximum allowed separation $`N_h`$ . A state of separated polarons is clearly seen for $`U=20`$.
Two distinct regimes are seen at $`g=2`$ within the bipolaronic region. At $`U=7<U_02\omega g^2=8`$, the correlation function represents the S0 bipolaron, while at $`U=9>U_0`$ we find the largest probability for two electrons to be on neighboring sites, which is characteristic of the S1 bipolaron. In contrast to previous calculations where phonons were treated classically , we find a crossover rather than a phase transition between the two regimes. The precision of presented correlation functions in the bipolaron regime is within the size of the plot symbols in the thermodynamic limit.
Figure (2a) plots the bipolaron mass ratio $`R_m=m_{bi}/2m_{po}`$ vs. $`U`$ for different values of $`\omega `$ and $`g`$. In all cases presented in Fig. (2), $`R_m`$ approaches 1 as $`U`$ approaches $`U=U_c`$ in agreement with a state of two free polarons. At fixed $`\omega =1`$ the bipolaron mass ratio increases by several orders of magnitude with increasing $`g`$ at $`U=0`$. The increase can be understood within strong coupling theory. Increasing $`U`$ sharply decreases $`R_m`$ in the S0 regime. Note that the scale in Fig. (2) is logarithmic. Near the strong coupling regime $`(g=2)`$ and for small $`U`$, good agreement is found between the numerical and the strong coupling result obtained from Eq. (8). The difference between these results increases as $`U`$ approaches $`U_0=8`$, where the perturbation theory based on the S0 bipolaron breaks down. In the S1 regime $`U>U_0`$, $`R_m`$ is small, as predicted by the strong coupling result.
The bipolaron isotope effect, shown in Fig. (2b), is large in the strong coupling ($`\omega =1,g=2`$) and small $`U`$ regime, where its value is somewhat below the large $`g`$ strong coupling prediction $`\alpha _{S0}2g^2\frac{1}{4}=7.75`$. With increasing $`U`$, $`\alpha `$ decreases and in the S1 regime approaches $`\alpha _{S1}=g^2/2=2`$. A kink is observed in the crossover regime.
We conclude with the phase diagram $`U_c(g)`$ shown in Fig. (3) at fixed $`\omega =1`$. Numerical results, shown as circles, indicate the phase boundary between two dissociated polarons each having energy $`E_{po}`$ and a bipolaron bound state with energy $`E_{bi}`$. In the inset of Fig. (3) we show the bipolaron binding energy defined as $`\mathrm{\Delta }=E_{bi}2E_{po}`$. The phase diagram is obtained from $`\mathrm{\Delta }=0`$. The dashed line, given by $`U_0=2\omega g^2`$, is a reasonable estimate for the phase boundary at small $`g`$. At large $`g`$ the dashed line roughly represents the crossover between a massive S0 and lighter S1 bipolaron. The dot-dashed line is the phase boundary between S1 and the unbound polaron phase, as obtained by degenerate strong coupling perturbation theory. Numerical results approach this line at larger $`g`$. The dot-dashed line asymptotically approaches $`U_c=4\omega g^2`$.
In the S1 regime, the strong coupling expansion can be used to rewrite the Holstein-Hubbard Hamiltonian Eqs. (2,3) as an effective t-J-V model that applies to an arbitrary number of particles,
$`H_{eff}`$ $`=`$ $`t_1{\displaystyle \underset{i,s}{}}\stackrel{~}{c}_{i,s}^{}\stackrel{~}{c}_{i+1,s}+h.c.+{\displaystyle \underset{i}{}}J𝐒_i𝐒_{i+1}+Vn_in_{i+1},`$ (13)
where $`t_1=t\mathrm{exp}(g^2)`$, $`J=2T_b>0`$, $`V=T_a+T_b/2>0`$ and $`\stackrel{~}{c}_{i,s}=c_{i,s}(1n_{i,s})`$ . The $`V`$ term is a repulsion between nearest neighbors. For simplicity and in keeping with the approximations used to derive the standard $`tJ`$ model, we have omitted next-nearest neighbor hopping terms that are of order $`t^2\mathrm{exp}(g^2)/(\omega g^2)`$, and a constant term. While the standard $`tJ`$ model can be derived from the Hubbard model only for $`J<t`$, the parameters of $`H_{eff}`$ are quite different, with $`J`$ typically much larger than $`t_1`$. In the static limit, an S1 bipolaron is stable in the interval $`J/4<V<3J/4`$. In the dilute limit, only singlet bipolarons exist with binding energy (in the static limit) $`E_{S=0}=3J/4+V4t^2/U+t^2/(\omega g^2)`$, while triplet fluctuations are almost completely frozen out due to the high energy $`E_{S=1}=J/4+V`$. Such a state is very similar to a negative-$`U`$ Hubbard model, except that singlets occupy two sites, like in the RVB model.
In conclusion, we have demonstrated that near the strong coupling limit a mobile S1 bipolaron exists with an effective mass of the order of a polaron mass, and an isotope effect in the range $`0<\alpha <g^2/2`$. The wavefunction of the S1 bipolaron is a spin singlet with extended $`s`$wave spatial symmetry. Taking into account the asymptotic stability criterion $`U_c=4\omega g^2`$, it is clear that a triplet S1 bipolaron that corresponds to the $`U\mathrm{}`$ solution is not bound. In the static limit, it can be shown that bound states of three or more polarons are not stable in the S1 regime, thus ruling out phase separation. An effective t-J-V model captures the physics of many S1 bipolarons in the strong coupling regime of the Holstein-Hubbard model, where antiferromagnetism is a consequence of the original Hubbard interaction $`U`$, while the attractive interaction is mediated by phonons. Taking into account the similarity between a system of S1 bipolarons and the negative-U Hubbard model, one should not rule out the possibility that S1 bipolarons form a superconducting state of either $`s`$wave or $`d`$wave symmetry in two spatial dimensions. In such a state, strong electron-electron interactions and electron-phonon coupling should be treated on an equal footing.
We acknowledge stimulating discussions with V.V. Kabanov and F. Marsiglio. J.B. greatfully acknowledges the support of Los Alamos National Laboratory where part of this work has been performed, and the financial support by the Slovene Ministry of Science. This work was supported in part by the US DOE. |
warning/0001/hep-ex0001028.html | ar5iv | text | # Laser cooling of electron beams at linear colliders.Invited talk at the International Symposium on New Visions in Laser-Beam Interactions, October 11-15, 1999, Tokyo, Metropolitan University Tokyo, Japan. To be published in Nucl. Instr. and Meth. B.
## 1 Introduction, one pass laser cooling
To explore the energy region beyond LEP-II, linear colliders (LC) with center–of–mass energy 0.5–2 TeV are developed now in the main accelerator centers ,,. Besides e<sup>+</sup>e<sup>-</sup> collisions, at linear colliders one can “convert” electrons to high energy photons using the Compton backscattering of laser light, thus obtaining $`\gamma \gamma `$ and $`\gamma e`$ collisions with energies and luminosities close to those in e<sup>+</sup>e<sup>-</sup> collisions -.
To attain high luminosity, beams in linear colliders should be very tiny. At the interaction point (IP) in the current LC designs, beams with transverse sizes as low as $`\sigma _x`$/$`\sigma _y`$ $``$ 300/3 nm are planned. Beams for e<sup>+</sup>e<sup>-</sup> collisions should be flat in order to reduce beamstrahlung energy loss. For $`\gamma \gamma `$ collision, the beamstrahlung radiation is absent also there are no beam instabilities therefore beams with smaller $`\sigma _x`$ ($`\sigma _x/\sigma _y10/1`$ nm) can be used ,- to obtain higher luminosity.
The transverse beam sizes are determined by the emittances $`ϵ_x`$, and $`ϵ_y`$. The beam sizes at the interaction point (IP) are $`\sigma _i=\sqrt{ϵ_i\beta _i}`$, where $`\beta _i`$ is the beta function at the IP. With the increase of the beam energy the emittance of the bunch decreases: $`ϵ_i=ϵ_{ni}/\gamma `$, where $`\gamma =E/mc^2,`$ $`ϵ_{ni}`$ is the normalized emittance.
The beams with a small $`ϵ_{ni}`$ are usually prepared in damping rings which naturally produce bunches with $`ϵ_{ny}ϵ_{nx}`$ . Laser RF photoguns can also produce beams with low emittances . However, for linear colliders it is desirable to have smaller emittances.
Recently, a new method of electron beam cooling was proposed which allows further reduction of the transverse emittances after damping rings or guns by 1–3 orders of magnitude ,.
The idea of laser cooling of electron beams is very simple, <sup>1</sup><sup>1</sup>1This idea was mentioned in the talk given by B.Palmer at the Berkeley Workshop on Gamma–Gamma colliders , first analyses of this method was done in ref. see fig. 1.
During head-on collisions with optical laser photons (with an electromagnetic wave in the case of a very strong field) the transverse distribution of electrons ($`\sigma _i`$) remains almost unchanged. Also the angular spread ($`\sigma _i^{}`$) is almost constant, because for photon energies (a few eV) much lower than the electron beam energy (several GeV) the scattered photons follow the initial electron trajectory with a small additional spread. So, the emittance $`ϵ_i=\sigma _i\sigma _i^{}`$ remains almost unchanged. At the same time, the electron energy decreases from $`E_0`$ down to $`E`$. This means that the transverse normalized emittances have decreased: $`ϵ_n=\gamma ϵ=ϵ_{n}^{}{}_{0}{}^{}(E/E_0)`$. One can reaccelerate the electron beam up to the initial energy and repeat the procedure. Then after N stages of cooling $`ϵ_n/ϵ_{n}^{}{}_{0}{}^{}=(E/E_0)^N`$ (if $`ϵ_n`$ is far from its limit).
The ultimate emittance can be estimated in the following way. In the electron beam reference system the counter moving laser photons have an energy $`\omega \gamma \omega _0`$ ($`\gamma =E/mc^2`$ and $`\omega _0`$ is the energy of laser photons) and scatter almost isotropically (roughly) without change of the energy. As a result, after multiple Compton scattering, the transverse energy of the electrons is equal to the transverse energy of laser photons
$$p_t^2/m\gamma \omega _0.$$
(1)
Here we assumed that $`\gamma \omega _0mc^2`$. On the other hand the r.m.s. angular spread of the electrons in the laboratory system by definition (see above)
$$\frac{p_t}{\gamma mc}\sqrt{\frac{ϵ_n}{\gamma \beta }},$$
(2)
where $`ϵ_n`$ is the normalized emittance and $`\beta `$ is the beta-function of the electron beam in the cooling region. Substituting eq.1 to eq.2 we obtain
$$ϵ_n\frac{\lambda _C}{\lambda }\beta ,$$
(3)
where $`\lambda _C=\mathrm{}/mc`$ is the Compton wave length, $`\lambda `$ is the laser wave length.
The physics of the cooling process is almost the same as radiative cooling of electrons in damping rings. However, here the process takes only 1 ps and the ultimate emittance is much lower than that in the damping rings. This is because in the “linear” laser cooling there are no bends (as in damping rings) which cause a growth of the horizontal emittance. Also the intra-beam scattering is not important due to a short “damping” time and following fast acceleration.
There are several question to this method
1. requirements for laser parameters (these parameters should be attainable);
2. an energy spread of the beam after cooling (at the final energy of a linear collider it is necessary to have $`\sigma _E/E0.1\%`$; also with a large energy spread it is difficult to “match” the cooled beam with the accelerator due to the chromaticity of focusing systems and also to focus the beams for the next stage of cooling.)
3. the limit on the final normalized emittances (it is desirable to have this limit lower than that obtained with storage rings and photoguns);
4. depolarization of electron beams (polarization is very important for linear colliders).
5. radiation damage of mirrors by the photons scattered at the large angles (these are X-ray photons).
Below we will see that this method is perfectly suited for linear colliders, however the main problem here obtaining of a very powerful laser pulses with high repetition rate. On my opinion, the most promising approach to this problem is a pulse stacking of laser pulses in an optical cavity ,. All these problems and possible solutions are discussed below.
## 2 Flash energy.
In the cooling region, a laser photon with energy $`\omega _0`$ (wave length $`\lambda `$) collides almost head–on with an electron of energy $`E`$. The kinematics is determined by two parameters $`x`$ and $`\xi `$ . The first one
$$x=\frac{4E\omega _0}{m^2c^4}=0.019\left[\frac{E}{\text{GeV}}\right]\left[\frac{\mu \text{m}}{\lambda }\right]$$
(4)
determines the maximum energy of the scattered photons:
$$\omega _m=Ex/(x+1)4\gamma ^2\omega _0(x1).$$
(5)
If the electron beam is cooled at the initial energy $`E_0=5\text{GeV}`$ (after damping ring and bunch compression) and $`\lambda =0.5\mu \text{m}`$ (Nd:glass laser) then $`x_00.2`$ (we will provide $`E`$ and $`x`$ with the index 0 for designation of their values at the begining of a cooling region).
For our further consideration we will need also the following formulae for the Compton scattering in the case $`x<<1`$. The energy spectrum of scattered photons(normalized per one scattering)
$$dp=(3/2)[12\omega /\omega _m+2(\omega /\omega _m)^2]d\omega /\omega _m.$$
(6)
The angle of the electron after scattering
$$\theta _1^2=(\omega _m\omega \omega ^2)/(\gamma ^2E^2).$$
(7)
The second parameter characterizes a strength of an electromagnatic wave <sup>2</sup><sup>2</sup>2Usually $`\xi ^2`$ is defined with $`\overline{B^2}`$ instead of $`B_0^2`$, that is 2 time smaller than in my “definition” given in ref., which I use here also.
$$\xi ^2=\left(\frac{eB_0\mathrm{}}{m\omega _0c}\right)^2,$$
(8)
where $`B_0`$ is the magnetic (or electric) field strength in the laser wave. At $`\xi ^21`$ an electron interacts with one photon from the field (Compton scattering, undulator radiation), while at $`\xi ^21`$ an electron scatters on many laser photons simultaneously (synchrotron radiation (SR), wiggler). We will see that in the considered method $`\xi ^2`$ may be “small” and “large”.
In the cooling region near the laser focus the r.m.s radius of the laser beam depends on the distance $`z`$ to the focus (along the beam) in the following way : $`r_\gamma =a_\gamma \sqrt{1+z^2/Z_R^2}`$, where $`Z_R=2\pi a_\gamma ^2/\lambda `$ is the Rayleigh length (an effective depth of laser focus), $`a_\gamma `$ is the r.m.s. focal spot radius. The density of laser photons is $`n_\gamma =(A/\pi r_\gamma ^2\omega _0)\mathrm{exp}(r^2/r_\gamma ^2)F_\gamma (z+ct)`$, where $`A`$ is the laser flash energy and $`F_\gamma (z)𝑑z=1.`$
In the case of strong field ($`\xi ^21`$) it is more appropriate to speak in terms of strength of the electromagnetic field which is $`\overline{B^2}/4\pi =n_\gamma \omega _0,B=B_0cos(\omega _0t/\mathrm{}kz)`$. Assuming $`F_\gamma =1/l_\gamma `$ and $`Z_Rl_\gamma l_e`$ and using the classical formula for radiation loss ($`dE/dx=(2/3)r_e^2\gamma ^2B^2,r_e=e^2/mc^2`$) we obtain the ratio of emittances before and after the laser target
$$\frac{ϵ_{n}^{}{}_{0}{}^{}}{ϵ_n}\frac{E_0}{E}=1+\frac{r_e^2}{3m^2c^4}B_0^2𝑑z=1+\frac{64\pi ^2r_e^2\gamma _0}{3mc^2\lambda ł_e}A$$
(9)
$$A[J]=\frac{25\lambda [\mu \text{m}]l_e[\text{mm}]}{E_0[\text{GeV}]}\left(\frac{E_0}{E}1\right).$$
(10)
These equations are correct at $`x1`$ for any value of $`\xi ^2`$. For example: at $`\lambda =0.5\mu \text{m},l_e=0.2\text{mm},E_0=5\text{GeV},E_0/E=10`$ the required laser flash energy $`A=4.5J.`$ To reduce the laser flash energy in the case of long electron bunches, one can compress the bunch (length) before cooling as much as possible and stretch it after cooling up to the required value.
The eqs.9,10 were obtained for $`Z_Rl_\gamma l_e`$ and give the minimum flash energy for a certain ratio $`E_0/E`$. To further estimate the photon density at the laser focus we will assume $`Z_R0.25l_e`$. In this case, the required flash energy is still close to its minimum, but the field strength is not so high as for very small $`Z_R`$. From the previous equation for $`Z_R=0.25l_e`$ it follows $`B_0^2/(8\pi )=\omega _0n_\gamma =A/(\pi a_\gamma ^2l_e)=8A/(\lambda l_e^2)`$. Substituting $`B_0`$ into eq.8 we get
$$\xi ^2=\frac{16r_e\lambda A}{\pi l_e^2mc^2}=\frac{3\lambda ^2}{4\pi ^3r_el_e\gamma _0}(\frac{E_0}{E}1)==4.3\frac{\lambda ^2[\mu \text{m}]}{l_e[\text{mm}]E_0[\text{GeV}]}(\frac{E_0}{E}1).$$
(11)
Example: for $`\lambda `$ = 0.5 $`\mu `$m, $`E_0`$ = 5 GeV , $`E_0/E`$ = 10, $`l_e`$ = 0.2 mm (the NLC project) $`\xi ^2`$ = 9.7. For larger bunch lengths and shorter wave lengths, $`\xi ^2`$ may be smaller. So, both ”undulator” and ”wiggler” cases are possible.
Later we will see that in order to have lower limit on emittance and smaller depolarization it is necessary to have a low $`\xi ^2`$. With a usual optics one can reduce $`\xi ^2`$ only by increasing $`l_\gamma `$ (and $`Z_R`$) with a simultaneous increase of the laser flash energy. From (10) and (11) we get
$$A\frac{\lambda ^3}{\gamma _0^2\xi ^2}\left(\frac{E_0}{E}1\right)^2.$$
(12)
Is it possible to reduce $`\xi ^2`$ keeping all other parameters (including flash energy) constant? Yes, providing a way to stretch the focus depth without changing the radius of this area is found. In this case, the collision probability (or $`B^2𝑑z`$) remains the same but the maximum value of $`\xi ^2`$ will be smaller. A solution of this problem was given in . It is based on the non-monochromaticity of the laser light and the chirped pulse technique. In this scheme, the cooling region consists of many laser focal points (continuously) and light comes to each point exactly at the moment when the electron bunch is there. One can consider that a short electron bunch collides on its way sequentially with $`n_f`$ (“number of focuses”) short light pulses of length $`l_\gamma l_e`$ and focused with $`2Z_Rl_e`$. There is one restriction on $`n_f`$: along the cooling length $`Ln_fl_e`$ the transverse size of an electron beam should be smaller than the laser spot size $`a_\gamma \sqrt{\lambda Z_R/2\pi }\sqrt{\lambda l_e/4\pi }`$. In further examples we will use $`n_f10`$ for stretching the cooling region from 100 $`\mu `$m to 1 mm.
Other method of “stretching” is using of several lasers focused to different, though this require larger flash energy (see sect.10).
## 3 Energy spread
The electron energy spread arises from the quantum-statistical nature of radiation. After energy loss $`\mathrm{\Delta }E`$, the increase of the energy spread $`\mathrm{\Delta }(\sigma _E^2)=\epsilon ^2\dot{n}(\omega )𝑑\omega 𝑑t=aE^2\mathrm{\Delta }E`$, where $`\dot{n}(\omega )`$ is the spectral density of photons emitted per unit time, $`a=14\omega _0/5m^2c^4=7x_0/10E_0`$ for the Compton case and $`a=55\mathrm{}eB_0/(8\pi \sqrt{3}m^3c^5)`$ for the “wiggler” case.
There is the second effect which leads to decreasing the energy spread. It is due to the fact that $`dE/dxE^2`$ and an electron with higher (lower) energy than the average loses more (less) than on average. This results in the damping: $`d(\sigma _E^2)/\sigma _E^2=4dE/E`$ (here $`dE`$ has negative sign). The full equation for the energy spread is $`d\sigma _E^2=aE^2dE+4(dE/E)\sigma _E^2`$, with solution $`\sigma _E^2/E^2=\sigma _{E_0}^2E^2/E_0^4+aE_0(E/E_0)(1E/E_0)`$. In our case
$$\frac{\sigma _E^2}{E^2}\frac{\sigma _{E_0}^2E^2}{E_0^4}+\frac{7}{10}x_0(1+\frac{275\sqrt{3}}{336\pi }\xi )\frac{E}{E_0}\left(1\frac{E}{E_0}\right),$$
(13)
here the results for the Compton scattering and SR are joined together. Example: at $`\lambda =0.5\mu \text{m},E_0=5\text{GeV}(x_0=0.19)`$ and $`E_0/E=10`$, the Compton term alone gives $`\sigma _E/E0.11`$ and with the “wiggler” term ($`\xi ^2=9.7`$, see the example above) $`\sigma _E/E0.17`$. What $`\sigma _E/E`$ is acceptable? In the last example $`\sigma _E/E0.17`$ at E = 0.5 GeV (after cooling). This means that at the collider energy E = 250 GeV we will have $`\sigma _E/E0.034\%`$, that is better than necessary (about 0.1 %).
In a two stage cooling system, after reacceleration to the initial energy $`E_0`$ = 5 GeV the energy spread is $`\sigma _E/E_01.7\%`$. For this value there may be a problem with focusing of electrons which can be solved using a focusing scheme with correction of chromatic aberrations. It is even more difficult to “match” the electron beam after the laser cooling with the accelerator (see sect.9.
## 4 Minimum normalize emittance.
It is determined by the quantum nature of the radiation. Let us start with the case of pure Compton scattering at $`\xi ^21`$ and $`x_01`$. In this case, the scattered photons have the energy distribution given by eq.6 The angle of the electron after scattering is connected with the energy by eq.7. After averaging over the energy spectrum we get the average $`\theta _1^2`$ in one collision: $`\theta _1^2=12\omega _0^2/(5m^2c^4)`$. After many Compton collisions ($`N_{coll}`$) the r.m.s. angular spread in i=x,y projection $`\mathrm{\Delta }\theta _i^2=0.5\mathrm{\Delta }\theta ^2=0.5N_{coll}\theta _1^2=0.5(\mathrm{\Delta }E/\overline{\omega })\theta _1^2=3\omega _0\mathrm{\Delta }E/5E^2`$.
The normalized emittance $`ϵ_{ni}{}_{}{}^{2}=(E^2/m^2c^4)r_i^2\theta _i^2`$ does not change when $`\mathrm{\Delta }\theta _i^2/\theta _i^2`$ = $`2\mathrm{\Delta }E/E.`$ Taking into account that $`\theta _i^2ϵ_{ni}/\gamma \beta _i`$ we get the equilibrium emittance due to the Compton scattering
$$ϵ_{ni}^{}{}_{,min}{}^{}0.5\gamma E\beta _i\mathrm{\Delta }\theta _i^2/\mathrm{\Delta }E=\frac{3\pi }{5}\frac{\lambda _c}{\lambda }\beta _i=\frac{7.2\times 10^{10}\beta _i[mm]}{\lambda [\mu \text{m}]}\text{m}\text{rad},$$
(14)
where $`\lambda _c=\mathrm{}/mc`$. For example: $`\lambda =0.5\mu \text{m},\beta =l_e/2=0.1\text{mm}(`$NLC$`)ϵ_n_{,min}=1.4\times 10^{10}\text{m}`$rad. For comparison in the NLC project the damping rings have $`ϵ_{nx}=3\times 10^6\text{m}`$rad, $`ϵ_{ny}=3\times 10^8\text{m}`$rad.
If $`\xi ^21`$, the electron moves as in a wiggler.Assume that the “laser wiggler” is planar and deflects the electron in the horizontal plane. If an electron with energy E emits a photon with energy $`\omega `$ along its trajectory the emittance changes as follows : $`\delta ϵ_x=(\omega ^2/2E^2)H(s);H(s)=\beta _x\eta _x^2+2\alpha _x\eta _x\eta _x^{}+\gamma _x\eta _x^2`$; where $`\alpha _x=\beta _x^{}/2,\gamma _x=(1+\alpha _x^2)/\beta _x,\beta _x`$ is the horizontal beta-function, $`\eta _x`$ is the dispersion function, $`s`$ is the coordinate along the trajectory. For $`\beta _x=const`$ the second term in H is equal to zero, the second term in a wiggler with $`\lambda _w\beta `$ is small, so that H(s) $`\beta \eta ^2`$. In a sinusoidal wiggler field $`B(z)=B_wcosk_wz,k_w=2\pi /\lambda _w`$, $`\eta ^{\prime \prime }=1/\rho `$ ($`\rho `$ is the radius of curvature) one finds that $`\eta ^{}=(eB_w/k_wE)sink_wz`$. The increase of $`ϵ_x`$ on a distance $`dz`$ is
$$\mathrm{\Delta }ϵ_x=\frac{H}{2}\left(\frac{\omega }{E}\right)^2\dot{n}(\omega )𝑑\omega 𝑑t=\frac{55}{48\sqrt{3}}\frac{r_e\mathrm{}c}{(mc^2)^6}E^5\frac{H}{\rho ^3}dz,$$
(15)
where $`H/\rho ^3_w=8\beta _x\lambda _w^2(eB_w)^5/(140E^5\pi ^3)`$ for the wiggler and $`\dot{n}(\omega )`$ is the spectral density of photons emitted per unit time. The energy loss averaged over the wiggler period is $`\mathrm{\Delta }E=r_e^2B_w^2E^2dz/(3m^2c^4)`$. The normalized emittance $`ϵ_n=\gamma ϵ`$ is not changed when $`Edϵ+ϵdE=0`$. Using this and replacing $`B_w`$ by 2$`B_0`$, $`\lambda _w`$ by $`\lambda /2`$ we obtain the equilibrium normalized emittance in the linear polarized electromagnatic wave for $`\xi ^21`$
$$ϵ_{nx}=\frac{11e^3\mathrm{}c\lambda ^2B_0^3\beta _x}{24\sqrt{3}\pi ^3(mc^2)^4}=\frac{11}{3\sqrt{3}}\frac{\lambda _C}{\lambda }\beta _x\xi ^3\frac{810^{10}\beta _x[\text{mm}]\xi ^3}{\lambda [\mu \text{m}]}\text{m}\text{rad}.$$
(16)
Using eq.11 we can get a scaling of the minimum $`ϵ_{nx}`$ for a multistage cooling system with a cooling factor $`E_0/E`$ in one stage: $`ϵ_{nx}\beta _x\lambda ^2(E_0/E)^{3/2}/(l_e\gamma _0)^{3/2}`$ when $`l_\gamma l_e`$(minimum A) and $`ϵ_{nx}\beta _x\lambda ^{7/2}(E_0/E)^3/(\gamma _0^3A^{3/2})`$ for free A and $`ł_\gamma >l_e`$ (for $`\beta _x=const`$). Stretching the laser focus depth by a factor $`n_f`$ , one can further reduce the horizontal normalized emittance: $`ϵ_{nx}1/n_f^{1/2}`$(if $`\beta _xn_f`$). For our previous example we have $`\xi ^2=9.7`$ and $`ϵ_{nx}=510^9\text{m}`$rad (in the NLC $`ϵ_{nx}`$ = 3$`10^6`$ m$``$rad). Stretching the cooling region with $`n_f`$=10, further decreases the horizontal emittance by a factor 3.2.
Comparing with the Compton case (14) we see that in the strong field the horizontal emittance is larger by a factor $`\xi ^3`$. The origin of this factor is clear: $`ϵ_{nx}\eta _x^2\omega _{crit.}`$, where $`\eta _x^{}\xi \theta _{compt.}`$ and $`\omega _{crit.}\xi \omega _{compt.}`$.
Let us roughly estimate the minimum vertical normalized emittance at $`\xi 1`$. Assuming that all photons are emitted at an angle $`\theta _y=1/(\sqrt{2}\gamma )`$ with the $`\omega =\omega _c`$ similarly to the Compton case, one gets $`\mathrm{\Delta }\theta _y^2=(\omega _c\mathrm{\Delta }E)/(2\gamma ^2E^2)=(3e\mathrm{}\overline{B_w}\mathrm{\Delta }E)/(4E^2mc)`$. Using the first part of eq.14 we get
$$ϵ_{ny}^{}{}_{min}{}^{}\frac{3}{8}\frac{\mathrm{}e\overline{B_w}}{m^2c^3}\beta _y=3\left(\frac{\lambda _C}{\lambda }\right)\beta _y\xi \frac{1.210^9\beta _y[\text{mm}]\xi }{\lambda [\mu \text{m}]}\text{m}\text{rad}.$$
(17)
For the previous example (NLC beams), eq.17 gives $`ϵ_{ny}^{}{}_{min}{}^{}7.510^{10}`$ m$``$rad (for comparison in the NLC project $`ϵ_{ny}=310^8`$ m$``$rad). The scaling: $`ϵ_{ny}\beta _y(E_0/E)^{1/2}/(l_e\gamma _0)^{1/2}`$ when $`l_\gamma l_e`$(minimum A) and $`ϵ_{ny}\beta _y\lambda ^{1/2}(E_0/E)/(\gamma _0A^{1/2})`$ for free A and $`ł_\gamma >l_e.`$
For arbitrary $`\xi `$ the minimum emittances can be estimated as the sum of (14) and (16) for $`ϵ_{nx}`$ and sum of (14) and (17) for $`ϵ_{ny}`$
$$ϵ_{nx}\frac{3\pi }{5}\frac{\lambda _C}{\lambda }\beta _x(1+1.1\xi ^3);ϵ_{ny}\frac{3\pi }{5}\frac{\lambda _C}{\lambda }\beta _y(1+1.6\xi ).$$
(18)
## 5 Depolarization
Finally, let us consider the problem of the depolarization. For the Compton scattering the probability of spin flip in one collision is $`w=(3/40)x^2`$ for $`x1`$ (it follows from formulae of ref.). The average energy losses in one collision are $`\overline{\omega }=0.5xE`$. The decrease of polarization degree after many collisions is $`dp=2wdE/\overline{\omega }=(3/10)x(dE/E)=(3/10)x_0(dE/E_0)`$. After integration, we obtain the relative decrease of the longitudinal polarization $`\zeta `$ during one stage of the cooling (at $`E_0/E1`$)
$$\mathrm{\Delta }\zeta /\zeta =0.3x_0E_0/\lambda ,$$
(19)
For $`\lambda =0.5\mu \text{m}`$ and $`E_0=5\text{GeV}`$, we have $`x_0=0.19`$ and $`\mathrm{\Delta }\zeta /\zeta =5.7\%`$. This is valid only for $`\xi ^21.`$
In the case of strong field ($`\xi ^21`$) the spin flip probability per unit time is the same as in the uniform magnetic field $`w=(35\sqrt{3}r_e^3\gamma ^2ce\overline{B^3})/(144\alpha (mc^2)^2)`$, where for the wiggler $`\overline{B^3}=(4/3\pi )B_w^3`$. Using the relation between $`dE`$ and $`dt`$ in the wiggler we get
$$\frac{\mathrm{\Delta }\zeta }{\zeta }=\frac{35\sqrt{3}er_eB_0}{9\pi \alpha (mc^2)^2}𝑑E\frac{35\sqrt{3}}{36\pi }x_0\xi .$$
(20)
For the general case, the depolarization can be estimated as the sum of equations (19) and (20)
$$\mathrm{\Delta }\zeta /\zeta 0.3x_0(1+1.8\xi ).$$
(21)
For the previous example with $`\xi ^2=9.7`$ and $`x_0=0.19`$ we get $`\mathrm{\Delta }\zeta /\zeta =0.057+0.32=0.38`$, that is not acceptable. This example shows that the depolarization effect imposes very demanding requirements on the parameters of the cooling system. The main contribution to depolarization gives the second term. Stretching the cooling region by a factor of ten we can get $`\mathrm{\Delta }\zeta /\zeta =0.057+0.115\%`$.
## 6 Ponderomotive forces.
It is well known that in a non-uniform oscillating field an avarage force acting on a particle is non-zero, this is so called ponderomotive force. This force leads to the repulsion of the electrons from the laser focus. This effect was not described in my first paper on laser cooling , but it was checked that it is not essential for the considered examples. Nevertheless, it is important, especially for low beam energies, let us consider this effect in more details.
The total force acting on an electron colliding head-on with a laser wave <sup>3</sup><sup>3</sup>3In this section $`\omega _0`$ is the frequency, in other sections it is the energy of the laser photon.
$$F=2eE_0(x)\mathrm{sin}2\omega _0t=2e(E_0+\frac{E_0}{x}x)\mathrm{sin}2\omega _0t.$$
(22)
Substituting $`x(eE_0/2\gamma m\omega _0^2)\mathrm{sin}2\omega _0t`$ we get
$$F=2e(E_0\frac{e}{2\gamma m\omega _0^2}\frac{E_0}{x}E_0\mathrm{sin}2\omega _0t)\mathrm{sin}2\omega _0t.$$
(23)
After averaging over time we get the ponderomotive force
$$\overline{F}=\frac{e^2}{4\gamma m\omega _0^2}\frac{E_0^2}{x}=\frac{U_{eff}}{x},$$
(24)
where the effective potential
$$U_{eff}=\frac{mc^2}{4\gamma }\xi ^2.$$
(25)
Particle move in the direction with the minimum ponential, in our case the electrons are repulsing from the laser target. Note, that in all considerations we assume a linearly polarized laser light with $`\stackrel{}{E}`$ laying in the horizontal plane. In this case there are forces only in the horizontal direction.
Let us assume that the laser spot size is about a factor of 2 larger than the horizontal size of the electron beam in the laser focus, then $`\xi ^2/x\xi ^2/2\sigma _{x,L}`$. The deflection angle on the cooling length $`l_c`$
$$\mathrm{\Delta }\vartheta \frac{p_t}{p}\frac{F(l_c/c)}{\gamma mc}\frac{\xi ^2l_c}{8\gamma ^2\sigma _{x,L}}.$$
(26)
The ponderomotive forces are not important if this angle is smaller than the angular spread of the electrons in the cooling region: $`\mathrm{\Delta }\vartheta <\sqrt{ϵ_{nx}/\gamma \beta }`$. This gives the minimum normalized horizontal emittance when ponderomotive forces are still not important
$$ϵ_{nx}^{}{}_{,min}{}^{}\frac{\xi ^4l_c^2}{64\gamma ^3\sigma _{x,L}^2}\beta .$$
(27)
We have seen before that the cooling by a factor of ten can be done at $`\xi ^21`$ on the length $`l_c`$ 1 mm. The minimum laser spot size $`\sigma _{x,L}\sqrt{\lambda Z_R/4\pi }2.5`$ $`\mu `$m at $`\lambda =0.5`$ $`\mu `$m and $`Z_R\sigma _{z,e}0.1`$ mm. Note, that for the first stages of cooling after the damping ring the minimum $`\sigma _{x,L}`$ is determined by the electron beam size and should be larger by a factor of 3 than this estimate. Now we investigate the minimum emittance therefore let us take $`\sigma _{x,L}=2.5`$ $`\mu `$m. Substituting this number to eq.27 and we get for $`E=500`$ GeV (the minimum energy in the cooling process) the estimate of the minimum normalized emittance when ponderomotive forces are still not important
$$ϵ_{nx}^{}{}_{min}{}^{}2.5\times 10^9\beta [\text{mm}]\text{m}\text{rad}$$
(28)
This is exectly equal to the limit on the emittance in the laser cooling obtained in the previous sections. So, at the chosen beam energies the ponderomotive forces still do not limit the minimum emittance in the laser cooling, but they will be important at the lower beam energies.
## 7 Some ”intermediate” conclusions.
Before considering “technical” aspects in the laser cooling we can summarize the results of the previous section as a possible ”optimistic” set of parameters for the laser cooling: $`E_0=4.5`$ GeV, $`l_e=0.2`$ mm, $`\lambda =0.5`$ $`\mu `$m, flash energy $`A10`$ J, focusing system with stretching factor $`n_f`$=10. The final electron bunch will have an energy of 0.45 GeV with an energy spread $`\sigma _E/E13\%`$, the normalized emittances $`ϵ_{nx}`$,$`ϵ_{ny}`$ are reduced by a factor 10, the limit on the final emittance is $`ϵ_{nx}ϵ_{ny}2\times 10^9`$ m$``$rad at $`\beta _i=1\text{mm}`$, depolarization $`\mathrm{\Delta }\zeta /\zeta 15\%`$. The two stage system with the same parameters gives 100 times reduction of emittances (with the same limits).
For the cooling of the electron bunch train one laser pulse can be used many times. According to (10) $`\mathrm{\Delta }E/E=\mathrm{\Delta }A/A`$ and even 25% attenuation of laser power leads only to small additional energy spread.
## 8 Laser systems
We have seen that the “very” minimum flash energy required for the one stage of the laser cooling is about 10 J for visible light and $`E5`$ GeV beam energy. If $`\lambda 1`$ $`\mu `$m (most powerful solid state lasers) then $`A20`$ J. If the electron beam is somewhat longer than in the considered examples, say $`\sigma _z=`$ 300-400 $`\mu `$m as in the TESLA project, then the required flash is already $``$ 40–50 J. Beside, as we will see in the next section, for decreasing the radiation damage of the mirrors one has to put the mirrors far enough from the electron beam. That leads already to more than 100 J flash energies. Moreover, the repetition rate should be equal to the rep.rate of linear colliders which is of the order of 10 kHz. So, the average power of the laser system should be of the order of one MegaWatt! At present, the best short pulse laser systems can produce several Joule pulses with the repetition rate several Hz and only there are hopes that next year a commercial 100 W laser with short (ps) pulses will be built ,. However, the situation is not so pessimistic.
To overcome the “repetition rate” problem it is quite natural to consider a laser system where one laser bunch is used for e$`\gamma `$ conversion many times. Indeed, one Joule laser flash contains about $`10^{19}`$ laser photons and only $`10^{12}`$ photons are knocked out in the collision with one electron bunch ($`100`$ Compron scattering per one electron).
The simplest solution is to trap the laser pulse to some optical loop and use it many times. In such a system the laser pulse enters via the film polarizer and then is trapped using Pockels cells and polarization rotating plates. Unfortunately, such a system will not work with Terawatt laser pulses due to a self-focusing effect.
Fortunately, there is one way to “create” a powerful laser pulse in the optical “trap” without any “nonlinear” material inside (only very thin dielectric coating of mirrors). This very promising technique is discussed below.
### 8.1 Laser pulse stacking in an “external” optical cavity.
Shortly, the method is the following. Using the train of low energy laser pulses one can create in the external passive cavity (with one mirror having some small transparency) an optical pulse of the same duration but with much higher energy (pulse stacking). This pulse circulates many times in the cavity each time colliding with electron bunches passing the center of the cavity.
The idea of pulse stacking is simple but not trivial and not well known. This method is used now in several experiments on detection of gravitation waves. In my opinion, pulse stacking is very natural for photon colliders and allows not only to build a relatively cheap laser system for $`e\gamma `$ conversion but gives us the practical way for realization of laser cooling, i.e. opens up the way to ultimate luminosities of photon colliders.
As this method is very important (may be crucial) for the laser cooling, let me explain it in more detail ,. The principle of pulse stacking is shown in Fig.2.
The secret consists in the following. There is a well known optical theorem: at any surface, the reflection coefficients for light coming from one and the other sides have opposite signs. In our case, this means that light from the laser entering through semi-transparent mirror into the cavity interferes with reflected light inside the cavity constructively, while the light leaking from the cavity interferes with the reflected laser light destructively. Namely, this fact produces asymmetry between cavity and space outside the cavity!
Let R be the reflection coefficient, T the transparency coefficient and $`\delta `$ the passive losses in the right mirror. From the energy conservation $`R+T+\delta =1`$. Let $`E_1`$ and $`E_0`$ be the amplitudes of the laser field and the field inside the cavity. In equilibrium, $`E_0=E_{0,R}+E_{1,T}`$. Taking into account that $`E_{0,R}=E_0\sqrt{R}`$, $`E_{1,T}=E_1\sqrt{T}`$ and $`\sqrt{R}1T/2\delta /2`$ for $`R1`$ we obtain $`E_0^2/E_1^2=4T/(T+\delta )^2.`$ The maximum ratio of intensities is obtained at $`T=\delta `$, then $`I_0/I_1=1/\delta Q`$, where $`Q`$ is the quality factor of the optical cavity. Even with two metal mirrors inside the cavity, one can hope to get a gain factor of about 50–100; with multi-layer mirrors it can reach $`10^5`$. ILC(TESLA) colliders have 120(2800) electron bunches in the train, so the factor 100(1000) would be perfect for our goal, but even the factor of ten means a drastic reduction of the cost.
Obtaining of high gains requires a very good stabilization of cavity size: $`\delta L\lambda /4\pi Q`$, laser wave length: $`\delta \lambda /\lambda \lambda /4\pi QL`$ and distance between the laser and the cavity: $`\delta s\lambda /4\pi `$. Otherwise, the condition of constructive interference will not be fulfilled. Besides, the frequency spectrum of the laser should coincide with the cavity modes, that is automatically fulfilled when the ratio of the cavity length and that of the laser oscillator is equal to an integer number 1, 2, 3… .
For $`\lambda =1\mu m`$ and $`Q=100`$, the stability of the cavity length should be about $`10^7`$ cm. In the LIGO experiment on detection of gravitational waves which uses similar techniques with $`L4`$ km and $`Q10^5`$ the expected sensitivity is about $`10^{16}`$ cm. In comparison with this project our goal seems to be very realistic.
In HEP literature I have found only one reference on pulse stacking of short pulses ($`1`$ ps) generated by FEL with the wave length of 5 $`\mu `$m. They observed pulses in the cavity with 70 times the energy of the incident FEL pulses, though no long term stabilization was done.
## 9 Radiation damage of mirrors and other “technical” aspects
The use of pulse stacking in the optical cavity makes the idea of laser cooling quit realistic.
Considering a practical scheme for laser cooling we should take into account many important practical aspects:
$``$ Radiation damage of the mirrors. X-ray radiation due to the Compton scattering here is many orders larger than the radiation level at the same angles in the $`e\gamma `$ conversion point. It is so because a) the electron energies are lower and b) each electron undergoes about one hundred Compton scattering. At $`\vartheta 1/\gamma `$ and $`x1`$ ($`x`$ is defined in sect.2) the energy of the Compton scattered photons $`\omega =4\omega _0/\vartheta ^2`$ and does not depend on the electron energy. However, at the lower beam energies the spectrum is softer ($`\omega _{max}=4\omega _0\gamma ^2)`$ and more photons (per one Compton scattering) have large angles. Simple calculations show that the number of photons/per electron emitted on the angle $`\vartheta `$ during the cooling of electrons from some large energy to the energy $`E_{min}`$ is
$$dn/d\mathrm{\Omega }=mc^2/4\pi \omega _0\gamma _{min}^3\vartheta ^4.$$
The total energy hitting the mirrors/cm<sup>2</sup>/sec is
$$dP/dS=mc^2N\nu /\pi \gamma _{min}^3\vartheta ^6L^2,$$
where $`L`$ is the distance between the collision (cooling) point (CP) and the focusing mirrors, N and $`\nu `$ are the number of electrons in the bunch and the collision rate. One can see a strong dependence of X-ray background on $`\gamma _{min}`$ and $`\vartheta `$. During the cooling the electron beam loses almost all its energy to photons. For $`E_0=5`$ GeV, $`N=2\times 10^{10}`$, $`\nu =15`$ kHz the total energy losses are about 200 kW, fortunately the flux decreases rapidly with increasing the angle. At $`\vartheta `$ = 30 mrad and $`L=5`$ m the power density $`dP/dS10^5`$ W/cm<sup>2</sup> and X-ray photons have an energies of about 4 keV (for 1 $`\mu `$m laser wave length). My estimations shows that rescattering of photons on the quads can give a comparable background.
I have describing this item in detail because for laser cooling the required flash energy is very high and to reach the goal we need very high reflectivity of the mirrors in the optical cavity. For TESLA with 3000 bunches in a train it would be nice to have mirrors with $`R>0.999`$. Such values of R are not a problem for dielectric mirrors, however the radiation damage may cause problems, better to avoid this problem.
$``$ Laser spot size should be several times larger than that of the focused electron beam to avoid an additional energy spread of the cooled electrons.
$``$ The cooled electron beam at the energy E=500–1000 GeV has an energy spread of $`\sigma _E/E15`$ % at the point where the $`\beta `$\- function is small ($`15`$ mm). Matching this beam with the accelerator is not a simple problem and requires special insertions for chromaticity correction. A similar problem exists for the final focus at linear colliders, it has been solved and tested at the FFTB at SLAC. Here the factor $`(F/\beta )\sigma _E/E`$ characterizing the chromaticity problem is smaller and the beam energy is 500 times smaller, so one can hope that it will be no problem.
$``$ The parameter $`\xi ^2`$ (defined above) should be small enough ($``$ 1) to keep the minimum attainable emittance, depolarization and the energy spread small enough. This is impossible with one laser (with required flash energy) without additional ”stretching” of the cooling region along the beam line. The simplest way to do this is to focus several lasers at different points along the beam axis.
## 10 Possible variant of the laser cooling system
The possible optical scheme for the TESLA project is shown in fig.3 (only the final focusing
mirrors are shown). The system consist of 8 independent identical optical cavities focusing the laser beams to the points distibuted along the beam direction on the length $`\mathrm{\Delta }z2`$ mm. The length of the cavity (the distance between the “left” mirror and an entrance semi-transparent mirror (not shown)) is equal to half the distance between the electron bunches in the train, 50 m for TESLA). The large enough angle between the edges of the mirrors and the beam axis (30 mrad) makes X-ray flux rather small (see the estimation above). Also this clear angle allows the final quads to be placed at a distance about 50 cm (from the side of the cooled beam), much closer than the focusing mirrors. Smaller focal distance makes the problem of chromaticity correction easier.
The maximum distance from the CP to the mirrors is determined only by the mirror size, the diameter of 20 cm seems reasonable, which gives $`L=5`$ m. The laser spot size at the CP is 7.5 $`\mu `$m, at least 3 times larger than the horizontal electron beam size with $`\beta _x<`$ 5 mm. The circulating flash energy in each cavity is 25 J and 200 J in the whole system, not small. The average power circulating inside the system is $`200\times 15`$ kHz = 3 MW! However, if the Q factor of the cavities is about 1000–3000 (3000 bunches in the electron train at TESLA), the required laser power is only 1–3 kW, or 0.15–0.4 kW/per each laser, that is already reasonable.
What about damage to the mirrors by such powerful laser light? The maximum laser flash energy/cm<sup>2</sup> on the mirrors is 0.13 J/cm<sup>2</sup> (0.7-2 has been achieved for 1 ps pulses ), the average power/cm<sup>2</sup> is 2 kW/cm<sup>2</sup> (there are systems with $`>5`$ kW/cm<sup>2</sup> working long time ). The average power inside one train ($`\mathrm{\Delta }t=1`$ msec) is 200 times higher (400 J/cm<sup>2</sup>/1 msec), but from the same ref. is known that 100 J/cm<sup>2</sup> for a time of 100 ns is OK, and extrapolating as $`\sqrt{t}`$ (thermoconductivity) one can expect the limit of about 10 kJ for 1 msec, much larger that expected in our case. One of potential problems is the vatiation of the laser amplifire temperatute inside one beam train, that is not simple to correct by adaptive optics. Note, here we are speaking about circulating, not absorbed energy. So, all power densities are below the known limits, this all depends, of course, on specific choice of mirrors.
At last, the main numbers. After one stage of such a cooling system the normalized emittance is decreased by a factor of 6. The ultimate normalized emittance (after several cooling sections) is proportional to the $`\beta `$-function at the CP, at $`\beta _{x,y}=1`$ mm it is about $`2\times 10^9`$ m rad, smaller than can be produced by the TESLA damping ring by a factor of 5000(15) in x(y) directions. From this point of view such a small $`\beta _x`$ is not necessary, but it should be small enough ($`<5`$ mm to have a small electron spot size in the cooling region. The first stage of cooling will be the most efficient because the beam is cooled in both horizontal and vertical directions (far from the limits). Besides, after decreasing the horizontal emittance the $`\beta `$\- function at the LC final focus can be made as small as possible, $`\sigma _z.`$ All together this can give a factor of ten in the luminosity.
## 11 Conclusion
The laser cooling of electron beams allows to reach a very luminosity at the high energy gamma-gamma colliders, 1–2 orders high than it is possible without such cooling. The method is quit straighforward, but the task is quit chelenging due to very high required laser power (peak and average). There are hopes that this problem can be solved using pulse stacking of laser pulses in an “external” optical cavity. This requires intensive R&D. |
warning/0001/hep-th0001102.html | ar5iv | text | # Untitled Document
POSSIBLE INSTABILITY OF THE VACUUM
IN A STRONG MAGNETIC FIELD
Giorgio CALUCCI<sup>*</sup><sup>*</sup>E-mail: giorgio@ts.infn.it
Dipartimento di Fisica Teorica dell’Università di Trieste, Trieste, I 34014 Italy
INFN, Sezione di Trieste, Italy
Abstract
The possibility that a static magnetic field may decay through production of electron positron pairs is studied. The conclusion is that this decay cannot happen through production of single pairs, as in the electric case, but only through the production of a many-body state, since the mutual magnetic interactions of the created pairs play a relevant role.
The investigation is made in view of the proposed existence of huge magnetic field strengths around some kind of neutron stars.
1.Introduction
The perturbative calculations in QED have been pushed to an exceptional degree of refinement as far as their comparison with the high precision experiments. The study of the situations which require a genuine non-perturbative treatment is less systematic, also in this field, however, there are problems that can be considered classical and have gained a high degree of clarification. One of these problems concerns the effects of intense external field: the decay of the vacuum subject to a strong electric field through emission of $`e^{}e^+`$ pairs is well settled through the analysis given by Schwinger, starting from the Euler-Heisenberg effective Lagrangian. If one considers strong magnetic fields there is a strict analogy if we admit the existence of monopoles, in that case the problem is only quantitative since the monopoles are expected to be very heavy, but for ordinary particles the parallel with the electric case is less straightforward.
The difference is that we associate a potential energy to the charge in an electric field, but we cannot do the same for an electric charge in a static magnetic field. The possible instability of the vacuum in the magnetic field, on the other side, seems interesting to be analyzed in view of the guess that extremely high magnetic field, of macroscopic extension, are realized in nature, around some neutron stars. If the magnetic field is not completely static there is certainly a pair production, it can be seen $`e.g.`$ treating the problem with the formalism of the adiabatic approximations, but in this case the rate of production depends on the square of the time derivative of the field. Here a qualitative and partially quantitative analysis of the possibility that the some kinds of vacuum instability leads to pair production even in static conditions is presented.
The main idea is stated in this way: any number of pairs in a given magnetic field cannot give rise to instabilities unless we take to some extent into account also their mutual interaction: in fact in this last case every pair shields partially the magnetic field in which the other particles lie, the overall effect could be a decreasing of the field intensity. This effect, when the original field is very strong, may decrease the density of magnetic energy and compensate the cost in energy for the creation of the pairs of charged particles. If this is true the effect is largely collective and cannot be studied particle by particle. A simple energy balance of this collective production is presented in section 2 and the reason for believing in the instability are given, since at this point we are in presence of an $`e^{}e^+`$ multiparticle state the superimposed Coulomb effects are studied in section 3, some conclusions are presented in the last section.
2.Main features of the model
The physical model is described in these terms: there is a classical magnetic field $`\stackrel{}{B}`$ constant in time and in it there is a second quantized electron field $`\mathrm{\Psi }`$. The total energy of the system is
$$_T=\frac{1}{2}B^2d^3r+\mathrm{with}=\{\mathrm{\Psi }^{}[\stackrel{}{\alpha }(\stackrel{}{p}+e\stackrel{}{A})+\beta m]\mathrm{\Psi }\}d^3r.$$
$`(1)`$
The usual variational principle, for the stationary case, $`\delta _T=0`$ yields the equation for the magnetic field. This equation is then brought to numerical form by putting it among eigenstates of $``$. Two known relations are used :
$$<\sigma |_s|\sigma >=_s<\sigma ||\sigma >=_sE_\sigma (s)$$
$`(2a)`$
when $``$ depends on the parameter $`s`$, $`|\sigma >`$is one of its eigenstates and $`E_\sigma `$ is the corresponding eigenvalue. Since $`E`$ will be expressed in terms of $`\stackrel{}{B}=curl\stackrel{}{A}`$, one must remember that:
$$\frac{\delta }{\delta A_i(\stackrel{}{r})}=ϵ_{ikl}_k\frac{\delta }{\delta B_l(\stackrel{}{r})}.$$
$`(2b)`$
The standard relation $`curl\stackrel{}{B}=\stackrel{}{J}`$, yields then
$$ϵ_{ikl}_kB_l=ϵ_{ikl}_k\frac{\delta }{\delta B_l}Ei.e.B_l=\frac{\delta }{\delta B_l}E+_lU.$$
$`(3)`$
The functional derivative is zero in the limit $`e0`$, so $`_lU`$ represents the field in absence of vacuum polarization, it will be denoted by $`B_l^{(o)}`$<sup>*</sup><sup>*</sup>A clear although artificial way of dealing with this kind of boundary condition may be found in the formalism of space-dependent coupling constants as proposed by Bogoliubov and Shirkov..
Now we specialize to a definite family of field configurations: let the magnetic field be uniform $`\stackrel{}{B}=B\stackrel{}{n}`$ and the volume of quantization of the system be a prism of height $`L`$, parallel to $`\stackrel{}{n}`$, and square section of side $`R`$. The multiparticle state is obtained in the most trivial way $`i.e.`$ by filling the one-particle states until some Fermi level and leaving completely empty the higher one-particle states. It turns out natural to consider two independent parameters in order to fix the highest populated level one for the longitudinal and one for the transverse degrees of freedom; so it results (see $`e.g.`$ ):
$$\underset{\mathrm{levels}}{}=\frac{L}{2\pi }_K^K𝑑keB\frac{R^2}{2\pi }\underset{o}{\overset{N}{}}\underset{s=\pm 1}{}$$
Looking for a uniform solution $`\stackrel{}{B}^{(o)}`$ must be a constant vector and $`\delta E/\delta B_l`$ is also a constant playing the role of a total magnetization $`M_l`$.
The energy coming from $``$ has a constant density $`ϵ=E/V,V=R^2L`$. For notational simplicity all the vectors are projected onto the direction $`\stackrel{}{n}`$ and only these components are now used. The total energy density is
$$ϵ_T=\frac{1}{2}B^2+ϵ=\frac{1}{2}(B^{(o)}+M)^2+ϵ,$$
$`(4)`$
and if it can become smaller that the energy density $`\frac{1}{2}B_{}^{(o)}{}_{}{}^{2}`$, we may argue that the vacuum polarization leads to an instability of the magnetic vacuum.
The single-particle energy levels are the standard relativistic Landau levels
$$w_{n,s}(k)=\sqrt{m^2+k^2+eB(2n+1+s)}$$
so the eigenvalue of $``$ is
$$E(K,N)=2\frac{1}{4\pi ^2}LR^2eB_K^K𝑑k\left[2\underset{n=1}{\overset{N}{}}\sqrt{m^2+k^2+2eBn}+\sqrt{m^2+k^2}\right].$$
$`(5)`$
The factor 2 in front of the whole expression arises when the contribution of the positron states is added to the contribution of the electron states. In order to perform the discrete sum the approximation used is
$$\underset{n=1}{\overset{N}{}}\sqrt{F+Gn}\frac{2}{3G}[F+(N+\frac{1}{2})G]^{3/2},$$
which is valid up to terms constant in $`N`$.
In the actual situation $`F=m^2+k^2,G=2eB`$. The integration over $`k`$ may be carried out completely in a straightforward way; the result is complicated and not very transparent. Since we are looking for a possible instability we are free to choose the trial state and so the range of the parameters $`N`$ and $`K`$, a choice that seems promising is
$$NeB>>K^2>>eB,K>>m,$$
because it allows an expansion in decreasing powers of $`\sqrt{NeB}.`$
$$ϵ=\frac{1}{3\pi ^2}[2K(2NeB)^{3/2}+K^3\sqrt{2NeB}]+\mathrm{}=aB^{3/2}+c\sqrt{B}+d+\mathrm{}.$$
$`(6)`$
The coefficients $`a,c`$ are defined in the expression, $`d`$ denotes the addendum which does not contains the factor $`NeB`$.
This expression is now used to calculate $`M`$:
$$M=\frac{3}{2}a\sqrt{B}\frac{1}{2}c/\sqrt{B}+\mathrm{}.$$
$`(7)`$
Then it is inserted into the expression for the total energy difference
$$\mathrm{\Delta }=\frac{1}{2}B^2+ϵ\frac{1}{2}[B^{(o)}]^2=ϵ+MB\frac{1}{2}M^2,$$
with the result:
$$\mathrm{\Delta }=\frac{1}{2}aB^{3/2}\frac{9}{8}a^2B+\frac{1}{2}c\sqrt{B}(\frac{3}{4}acd).$$
$`(8)`$
The coefficient of the leading power in $`B`$ is negative and we may therefore conclude that there must exist configurations in which the creation of a collective state of electrons and positrons has the effect of lowering the total energy notwithstanding the cost in energy of the mass and kinetic terms for the charged particles.
3. Effect of the Coulomb interaction
Since we are now considering the creation of a plasma of $`e^+,e^{}`$ we are also led to consider the possible effect of the Coulomb interaction. It may be estimated perturbatively by inserting the corresponding two-body operator:
$$𝒱_c=\frac{1}{2}\mathrm{\Psi }^{}(\stackrel{}{r})\mathrm{\Psi }(\stackrel{}{r})\frac{\alpha }{|\stackrel{}{r}\stackrel{}{s}|}\mathrm{\Psi }^{}(\stackrel{}{s})\mathrm{\Psi }(\stackrel{}{s})d^3rd^3s$$
$`(9)`$
between the original states. This interaction is expected to give a negative contribution to the energy and since it does not contain $`\stackrel{}{A}`$ it does not modify the magnetization. The procedure to deal with this problem can be found in standard textbooks, the actual calculations become, however, very laborious if we want to use the correct wave functions of the electron, which are essentially harmonic-oscillator functions. A simplified investigation is here presented: the particle states are simply represented by plane waves, the sums are cut at the same levels as in the previous case, so a maximum longitudinal momentum $`K`$ is used and a maximum transverse momentum $`P`$ is introduced, with the later identification $`P^2=NeB`$. In this way it is possible to get a simple estimate of the Coulomb effect, in considering the pair interaction we obtain the overall cancellation of the diagonal terms $`(e^{},e^{}),(e^+,e^+)`$ with $`(e^{},e^+)`$, which in this case is particularly evident, whereas the exchange terms survive and the contributions of the negative and positive charge add up; the result it obviously definite negative and may be expressed as:
$$E_c=4\pi V\frac{1}{(2\pi )^6}\frac{\alpha }{(\stackrel{}{p}_1\stackrel{}{p}_2)^2}d^3p_1d^3p_22,$$
$`(10)`$
the last factor $`2`$ comes here also from the sum over the two charge states.
The integrand depends only on the difference of the momenta, so three integrations are easily performed; defining $`\rho =K^2/P^2`$ the result is
$$I=\frac{1}{(\stackrel{}{p}_1\stackrel{}{p}_2)^2}d^3p_1d^3p_2=128K^2P^2_o^1\frac{(1u)(1u)(1w)}{u^2+v^2+\rho w^2}𝑑u𝑑v𝑑w$$
In the limit $`\rho 0i.e.P>>K`$ the integral develops a logarithmic singularity which gives the leading term:
$$I16\pi K^2P^2[\mathrm{ln}P^2/K^2+const]$$
With the identification $`P^2=NeB`$ the leading term of the Coulomb energy is:
$$ϵ_c=\frac{\alpha }{8\pi ^5}I=\frac{2\alpha }{\pi ^4}NeBK^2\mathrm{ln}(NeB/K^2)$$
$`(11)`$
Comparing it with the results of eq. (6) we see that the Coulomb energy is definitively sub-leading. This results gives us confidence that higher order perturbative corrections, which may also affect the magnetization, will not destroy the main result of the previous chapter.
4.Conclusions
The main conclusion we may draw from this particularly simple model is that the indications of an instability of the magnetic vacuum are confirmed. The source of this instability is to be found in the dependence of the energy density of the electron field on the magnetic field. The leading term grows as $`ϵ_oB^\sigma `$ and is obviously positive, the diamagnetic term has a leading addendum $`B(ϵ_o/B)`$, it carries a minus sign, and since $`1<\sigma <2`$ this term over-compensates the positive amount of energy required for the creation of pairs, the applied magnetic field $`B^{(o)}`$ is reduced by the pair creations, it results in fact, for the actual case $`\sigma =\frac{3}{2}`$ $`BB^{(o)}\frac{3}{2}a\sqrt{B^{(o)}}+\frac{9}{8}a^2`$; the effects of the Coulomb interaction are seen only in the sub-leading terms. The requirement of very large magnetic field is essential in order that the classification in leading and sub-leading terms be meaningful, in fact looking blindly at eq (8) one would get the impression that the total energy continuously decreases with increasing $`N`$, this is clearly impossible, for a given $`B^{(o)}`$ at a certain value of $`N`$, the resulting field $`B`$ becomes too small for the whole treatment to be correct. The determination of the actual values of $`B^{(o)}`$ for which the process of spontaneous pair creation can happen is not possible within the present treatment; the main problem in performing a quantitative estimate of the process lies in the fact that it proceeds necessarily via a tunnel effect, because the creation of a small number of pairs is not enough to lower the magnetic energy.
In fact the possible existence of the instability of the magnetic vacuum is strictly related to the mutual interaction of the created electron pairs, expressed by the fact that they are located in the field modified by the existence of the other pairs. The existence of an instability when only non interacting pairs are considered is excluded, in a static field, because under these conditions, the effective Euler-Heisenberg Lagrangian, which describes the effect of virtual charged spinors in a given external field, never develops an imaginary part, contrary to what happens for the electrostatic case. The situation is also different from what expected in the spin-one case, where for very intense magnetic fields instabilities are expected already at the level of single pair creation, but the known charged particles with spin one are much heavier than the electron so there should not be sizable interference between the two processes at the foreseen field strengths of the order of $`10^{10}`$T.
It must be, finally, noted the all the virtual effects have been ignored, they are likely to be important for very large field strengths, because they renormalize the electron charge and may also destroy the spherical symmetry of the Coulomb interaction, also the photon degrees of freedom coupled with the electron field may have a role, here only static interactions were analyzed, but all these and possibly other dynamical features are superimposed complications, they do not give, however, any indication of being able to destroy the main conclusion that emerges from the simple analysis presented in section 2.
Acknowledgments
This work has been partially supported by the Italian Ministry of the University and of Scientific and Technological Research by means of the Fondi per la Ricerca scientifica - Università di Trieste .
References
1. J.Schwinger, Phys. Rev. 82, 664 (1951) and Phys. Rev. 93, 615 (1954).
2. V.I Ritus: The Lagrangian function of an intense electromagnetic field
A.I.Nikishov: The S-matrix of QED with pair creating external field in Issues in intense field QED; ed. V.L. Ginzburg - Nova scientia pub. N.Y. 1987.
Qiong-gui Lin J.Phys. G25,17 (1999)
3. C.Thompson, R.C.Duncan, Astrophys. J. 408, 194 (1993);
H. Hanami, Astrophys. J. 491, 687 (1997);
K. Hurley et al. , Nature 397, 41 (1999);
M. Feroci $`et.al.`$, Astrophys. J. 515, L9 (1999).
4. G. Calucci Modern Phys. L. A 14, 1183 (1999).
5. N.N. Bogoliubov, D.V.Shirkov: Introduction to the theory of quantized fields - Ch.VI Interscience Pub. New York 1959.
6. L.D.Landau, E.M.Lifshits: Quantum mechanics - Ch. XV ¶112 Pergamon Press Oxford 1977.
7. L.D.Landau, E.M.Lifshits: Statistical Physics - Ch. VII ¶80 Pergamon Press Oxford 1980.
8. J.S.Schwinger Particles, Sources and Fields, vol.2, Ch.4 Addison Wesley 1973.
9. N.K.Nielsen and P.Olesen N.Phys. B144 (1978), 376;
J. Ambjorn and P.Olesen Nucl Phys B315 (1989), 606.
10. R. Ragazzon, J. Phys. A: Math. Gen 25, 2997 (1992);
G. Calucci, R.Ragazzon J. Phys. A: Math. Gen 27, 2161 (1994). |
warning/0001/astro-ph0001432.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The exploration of anisotropy and polarization of the Cosmic Microwave Background (CMB) is fundamental in developing our knowledge about the Universe. During the last ten years several CMB measurement projects are being carried out <sup>1</sup>. After the successful COBE mission the attention has been focused on obtaining higher angular resolution of observational data in the range of the so-called Doppler peak and subsequent peaks. The level of the CMB anisotropy at multipoles $`l>30`$ up to $`l10^3210^3`$ is a gold mine of cosmological information about the early Universe and its most important parameters. However, statistical analysis of modern observational data is extremely complicated and time consuming. This complexity leads to development of different approximate methods of the data reduction such as Wiener-filtration <sup>2,3</sup>, radical compression method <sup>4</sup>, likelihood method, band-power method and others <sup>5,6,7</sup>. The existence of a large number of processing methods of observational data for different types of experiments is easily understood. Firstly, there are many difficulties and peculiarities of the measurements connected with the scan strategy, specific beam chopping, map making, etc. Secondly, the reconstruction of the parameters of the CMB signal is a sort of “inverse problem”. These difficulties are not specific for CMB measurements only. They are well known also in optical astronomy and in image reconstruction methods from satellite or airplane photography <sup>8</sup>.
Which one of the data reduction methods mentioned above is the most preferable for any type of experiments? The answer of this question depends on the our “intuitive” hopes. We can use the following criterion which was suggested by Tegmark <sup>2</sup>: “one method is better than another if it retains more of the cosmological information which operationally means that it will lead to smaller error bars on the parameter estimates”. However, before CMB image reconstruction from the observational scans is done we have no information about the structure of the signal (systematic errors, Gaussianity, different types of foreground sources and noise, etc). So, roughly speaking, the Tegmark criterion plays an important role a posteriori, when the reconstruction of the CMB by different methods is performed. We have to remember that for different strategies of observations and for different concrete experiments different methods of data processing can be considered as the most preferable. In addition different methods are preferable for different goals for using obtained data. In general, we may conclude that different methods complement each other.
The purpose of this paper is to propose and investigate a new method with the following aim: how can we estimate the probability of the presence of the CMB signal in an observational map and determine its characteristics if we have a hypothesis about its statistical properties (for example that it is Gaussian), and have a hypothesis about its power spectrum.
We suggest a new type of linear filtration of one-dimensional scans of CMB maps that is sensitive to non-Gaussian noise of any origin. This filtration does not change the primordial CMB power spectrum in contrast to the well known Wiener filter method. We use geometrical and topological characteristics such as the Minkowski functionals <sup>9-11</sup> and statistics of peak distribution above and below some definite threshold <sup>12-15</sup>. We show, that these characteristics are very sensitive criteria of Gaussianity of the restored signal. The filtration of the two-dimensional maps of the CMB with the filter which preserves the statistical properties of the underlying signal was described in <sup>7</sup>.
The full advantage of the new method can be demonstrated even on very small data sets. We apply our method to a “toy” model of a mixture of CMB signal and foreground point sources to model balloon- and ground-based measurements at the frequency band $`\nu 10200`$GHz. The application of the method to real observational data will be considered in a separate paper.
The rest of this paper is organized as follows. In Section 2 we describe the model of the real experiments and describe a new so called “power filter” for reconstructing the CMB signal. Section 3 is devoted to peak statistics and cluster analysis of the CMB signal in one-dimensional records. In Section 4 we present the new reconstruction technique of the CMB spectrum from records with point sources and demonstrate the strength of our method. Section 5 contains our conclusions.
## 2 The model of real experiments and reconstruction of the CMB signal
The statistically isotropic distribution of the CMB temperature fluctuations on the sky is usually described by the spherical harmonics expansion $`Y_{lm}(\stackrel{}{q})`$:
$`T(q)={\displaystyle \underset{l=1}{\overset{\mathrm{}}{}}}{\displaystyle \underset{m=l}{\overset{l}{}}}a_{lm}B_{lm}Y_{lm}(\stackrel{}{q}),`$ (1)
where $`a_{lm}`$ are zero-mean Gaussian deviates, $`a_{lm}a_{l^{}m^{}}^{}=C_l\delta _{ll^{}}\delta _{mm^{}}`$, with $`C_l`$ given by the CMB power spectrum, $`B_{lm}`$ the antenna beam, and $`\stackrel{}{q}(\theta ,\phi )`$ the unit vector defining the position of the observational direction on the sky. Eq.(1) can be written for all frequency channels $`\nu _1\mathrm{}\nu _{\stackrel{~}{m}}`$, where $`\stackrel{~}{m}`$ is the number of bands. After the COBE mission, different CMB experiments such as Saskatoon, QMAP, CAT and others have observed rather small parts of the sky with higher angular resolution. They estimated statistical characteristics of the CMB signal from the maps using some specific pixelization strategy. Two future space missions, MAP and PLANCK, will cover a significantly greater part of the sky and will provide data with higher precision.
The traditional view is that most information of a Gaussian random field is contained in the multipole power spectrum $`C_l`$. As a result a great deal of efforts has gone into obtaining the best estimate of the power spectrum $`C_l`$ from the experimental data. The power spectrum $`C_l`$ describes properties of a two-dimensional map. However, there exist many balloon-borne experiments and ground-based projects which performed observations of the CMB anisotropy along one-dimensional sky patches. Besides that the future MAP and PLANCK missions will collect data from a large number of intersecting circles which will then be reconstructed into a two-dimensional sky map. Briefly speaking, because of the simplicity of data analysis in the one-dimensional case, the collection of CMB observational data in this format is an interesting option for modern anisotropy experiments.
Usually (see ref. <sup>2</sup> for a recent review) one has a time-ordered set of data and can, knowing the adopted scan strategy, obtain a map with pixel values denoting the sum of a pure CMB signal $`𝐱`$ and different types $`𝐧`$ of noise, including foregrounds, atmospheric emission, radiometer noise and so on.
In the following we consider final maps consisting of a pure CMB signal $`𝐱`$ and noise $`𝐧=_i𝐧_i`$ and use therefore for a one-dimensional single record
$`𝐘(𝐭)=𝐱(t)+𝐧(t)`$ (2)
where $`t`$ is an independent variable along the record.
For this simple model of the measured signal Eq.(2) one can approximately reconstruct the pure CMB signal using a linear reconstruction technique (see <sup>8</sup> for a general description of the method). Note that in Eq.(2) we do not assume the random fields $`𝐱`$ and $`𝐧`$ to be Gaussian.
The general task of linear methods of reconstruction of a signal is to obtain an approximate estimation $`\overline{𝐱}^{}(t)`$ of a signal $`𝐱(t)`$ from the measured signal $`𝐘(𝐭)`$ using linear operations. Following <sup>8</sup> the estimation $`\overline{𝐱}^{}(t)`$ of the CMB signal $`𝐱(t)`$ from Eq.(2) can be written as
$`\overline{𝐱}^{}(t)={\displaystyle G(tt^{})𝐘(t^{})𝑑t^{}}`$ (3)
where $`G(tt^{})`$ is the filter of the reconstruction.
Before describing the new method which we propose for the reconstruction of the CMB signal, let us briefly discuss the standard linear filtering methods which were reviewed by Tegmark <sup>2</sup>. Firstly, following <sup>2</sup>, we point out that linear methods of filtration of maps do not destroy information about phases which is contained in the initial maps (or one-dimensional records). Secondly, the non-linear extraction of the signal, such as band-power method <sup>4,5</sup>, needs some special assumptions about the statistical properties of CMB signal and noise (for example Gaussian statistics of noise). Besides that, all non-linear methods destroy information contained in the maps <sup>2</sup>.
Let us return to Eq.(2) and Eq.(3) and consider one example of CMB signal reconstruction which uses a simple Wiener filtering method <sup>2</sup>. Following <sup>2</sup>, the estimated signal from Eq.(2) is
$`\overline{𝐱}^{\prime \prime }(t)={\displaystyle W(tt^{})𝐘(t)𝑑t},`$ (4)
where $`W(tt^{})`$ is the so called Wiener filter which minimizes the reconstruction error $`\mathrm{\Delta }^2(\overline{𝐱}^{\prime \prime }𝐱)^2`$. Tegmark and Efstathiou <sup>3</sup> and Tegmark <sup>2</sup> have shown that the W filter has a particularly simple form in Fourier space under assumptions in Eq.(2). For this case $`W`$ has the following form:
$`W(k)={\displaystyle \frac{P_{CMB}(k)}{P_{tot}(k)}}=\left(1+{\displaystyle \frac{P_{noise}(k)}{P_{CMB}(k)}}\right)^1,`$ (5)
where $`P_{tot}(k)=P_{CMB}(k)+P_{noise}(k)`$ is the power spectrum of the measured signal $`𝐘(t)`$, $`P_{CMB}(k)`$ the CMB power spectrum folded with the antenna beam and $`P_{noise}(k)`$ the noise power spectrum.
Let us return to Eq.(4). It is straightforward to prove that the power spectrum of the reconstructed signal $`P_{\overline{x}^{\prime \prime }}(k)`$ is equal to
$`P_{\overline{x}^{\prime \prime }}(k)=W^2(k)P_{tot}(k)=P_{CMB}(k)W(k).`$ (6)
As one can see from Eq.(5), $`W<1`$ which, together with Eq.(6), shows that the Wiener filter decreases the CMB power spectrum of the reconstructed signal.
For a linear reconstruction of the CMB signal in one-dimensional scans we propose to use the so called “power filter” $`G_p`$ <sup>8</sup>. As we told in Section 1 the corresponding consideration for two-dimensional CMB maps was done in <sup>7</sup>. This filter minimizes the difference between the power spectrum $`P_{\overline{x}^{}}(k)`$ and the CMB power spectrum. It is clear that this condition corresponds to
$`P_{\overline{x}^{}}(k)=P_{CMB}.`$ (7)
We will demonstrate that for our case $`G_p=W^{1/2}`$. For the proof let us rewrite Eq.(3) for our filter $`G_p`$:
$`\overline{𝐱}^{}(t)={\displaystyle G_p(tt^{})𝐘(t^{})𝑑t^{}}.`$ (8)
Now note that from Eq.(8) one can conclude that
$`P_x^{}(k)=G_p^2(k)P_{tot}(k).`$ (9)
Using Eq.(5), Eq.(7) and Eq.(9) we conclude that
$`G_p^2(k)={\displaystyle \frac{P_{CMB}(k)}{P_{tot}(k)}}=W(k)`$ (10)
which proves our statement: $`G_p=W^{1/2}`$.
## 3 Peak statistics and cluster analysis of the reconstructed CMB signal
Using power filtration by a $`G_p`$ filter we transform the initial observational map to a reconstructed record $`\{\overline{𝐱^{}}(t)\}`$. Like any other linear filter the $`G_p`$ filter transforms the power spectrum of the initial signal preserving phases of the signal. This property of the linear filtration is extremely important and we will consider it in more details. Let us consider a one-dimensional record $`Y(t)`$ and its Fourier expansion:
$`Y(t)={\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}\left[a_n\mathrm{cos}\left({\displaystyle \frac{2\pi nt}{\stackrel{~}{T}}}\right)+b_n\mathrm{sin}\left({\displaystyle \frac{2\pi nt}{\stackrel{~}{T}}}\right)\right],`$ (11)
where $`\stackrel{~}{T}`$ is the total length of a record. After a power filtration we get
$`Y^{}(t)={\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}\left[a_nG_n\mathrm{cos}\left({\displaystyle \frac{2\pi nt}{\stackrel{~}{T}}}\right)+b_nG_n\mathrm{sin}\left({\displaystyle \frac{2\pi nt}{\stackrel{~}{T}}}\right)\right].`$ (12)
The initial power spectrum of the record (Eq.(11)) is given by
$`P_n^{(in)}={\displaystyle \frac{1}{2}}(a_n^2+b_n^2)`$ (13)
and of the reconstructed power spectrum (Eq.(12)) by
$`P_n^{(rec)}={\displaystyle \frac{1}{2}}G_n^2(a_n^2+b_n^2).`$ (14)
However our $`G_p`$ filter does not destroy the spectrum of phases since
$`\mathrm{\Phi }_n^{(in)}=\mathrm{arctan}\left({\displaystyle \frac{b_n}{a_n}}\right)=\mathrm{\Phi }_n^{(rec)}.`$ (15)
So, roughly speaking, we transform the power spectrum of the initial signal by $`G_p`$ filter but preserve the spectrum of phases. Thus Eq.(15) demonstrates that such a filtration preserves a Gaussian or non-Gaussian property of the modes which are not suppressed by the filter. This property is valid for all linear filters. The specific property of the $`G_p`$ filtration is that in addition to preserve the phases it also preserves the spectrum of the CMB signal. $`G_p`$ is a unique filter with this property. Cluster analysis and peak statistics <sup>11-14</sup> are very sensitive to both phases and the power spectrum. Thus, if we believe that after the $`G_p`$ filtration the reconstructed signal really represents the approximate distribution of $`\mathrm{\Delta }T`$ of CMB signal, it should have the same phase characteristics and the same peak statistics and properties of clusterisation which are predicted for the initial CMB signal (under an accepted hypothesis about the cosmological model). Confrontation of peak statistics and clusterisation of the reconstructed signal with the theoretical predictions is a good test of the “quality” of the reconstruction.
We will investigate how this method works in the case of one-dimensional records. In the previous section we have shown that the reconstructed record $`𝐱^{}(t)`$ has the power spectrum $`P_{CMB}`$. We would like to point out that all non-Gaussian features of the initial signal are transformed to the reconstructed record $`𝐱^{}(t)`$ because of the linear character of filtration. We also note that in the case of application of the Wiener filter, peak statistics of the reconstructed signal is different from the hypothetical CMB signal because of the change of the power spectrum.
Assuming the primordial cosmological signal to be Gaussian, we perform the test of the filtered signal with regard to Gaussianity. In fact, we calculate the probability that the filtered signal has a Gaussian nature.
The $`G_p`$ filtering obviously makes the power spectrum of the observed signal equal to the expected CMB spectrum. The rest of the information to be investigated is about the phases only. In order to perform this test, we use a geometrical approach.
We normalize the $`𝐱^{}(t)`$ record to its variance: $`\nu (t)\mathrm{\Delta }T/\sqrt{\mathrm{\Delta }T^2}`$ and consider the following characteristics of the $`\nu (t)`$ statistics:
A) Normalized distribution functions of maxima and minima above the threshold $`\nu _t`$:
$`f_{max}(\nu _t)={\displaystyle \frac{N_{max}(\nu _t)}{\overline{N_{max}}}}f_{min}(\nu _t)={\displaystyle \frac{N_{min}(\nu _t)}{\overline{N_{min}}}},`$ (16)
where $`N_{max}(\nu _t)`$ and $`N_{min}(\nu _t)`$ are the number of maxima and minima, respectively, and $`\overline{N_{max}}=N_{max}(\nu _t\mathrm{})`$, $`\overline{N_{min}}=N_{min}(\nu _t\mathrm{})`$ are the total number of maxima and minima.
B) The global Minkowski functionals $`M_1`$ and $`M_2`$ for a one-dimensional record, which characterize the cumulative distribution function $`(M_1)`$ and the Euler characteristic (genus) of the random signal $`(M_2)`$. The last quantity was first introduced in cosmology by Doroshkevich <sup>16</sup>. For the N dimensional case there are N+1 Minkowski functionals.
C) Finally we consider the mean length of the clusters above and below a constant threshold $`\nu _t`$. A one-dimensional cluster of maxima is the continuous part of the curve with $`\nu (t)>\nu _t`$ inside the interval $`t(t_1,t_2)`$. Here $`t_1`$ and $`t_2`$ are roots of the equation $`\nu (t_{1,2})=\nu _t`$ and $`t_2>t_1`$. The length $`k(\nu _t)`$ of the cluster is defined as the number of peaks with height $`\nu _{peak}>\nu _t`$ in the cluster. If the value of $`\nu _t`$ is high ($`\nu _t1`$) then only high maxima (separated from each other by regions with $`\nu (t)<\nu _t`$) are present above the threshold and the typical length of a cluster is $`k(\nu _t)=1`$. The reduction of the threshold level $`\nu _t`$ down to $`\nu _t0`$ or $`\nu _t<0`$ leads to the appearance of big clusters when maxima of smaller clusters begin to connect together and generate a new cluster. This process is characterized by the number of clusters $`N_k(\nu _t)`$ of the length $`k`$ and the total number of clusters $`N(\nu _t)`$ which are present in the anisotropy records for an appropriate threshold $`\nu _t`$:
$`N(\nu _t)=\mathrm{\Sigma }_{k=1}^{\mathrm{}}N_k(\nu _t).`$ (17)
Thus for different statistics of initial records the mean length of a cluster at threshold level $`\nu _t`$ is
$`k(\nu _t)={\displaystyle \frac{\mathrm{\Sigma }_{k=1}^{\mathrm{}}kN_k(\nu _t)}{\mathrm{\Sigma }_{k=1}^{\mathrm{}}N_k(\nu _t)}}.`$ (18)
It is necessary to note, that only the first characteristic considered under A and the first Minkowski functional $`M_1`$ are independent. The rest of them can be represented in terms of the distribution of peaks (A). Thus, we use them only for visual clarification and representation of our results.
The defining Eq.(16) - Eq.(18) do not depend on the nature of the initial signal or its composition.
Below we use the standard definition of the spectral parameters:
$`\begin{array}{c}\sigma _n^2=k^{2n}P_{CMB}(k)𝑑k,n=0,1,2\hfill \\ \gamma =\frac{\sigma _1^2}{\sigma _0\sigma _2}\hfill \end{array}`$ (21)
where $`P_{CMB}(k)`$ is the one-dimensional CMB power spectrum.
The characteristics A, B and C have a rather simple form in case of a random Gaussian field <sup>10-12,14</sup>:
The number of maxima and minima above a threshold $`\nu _t`$ are:
$`\begin{array}{c}N_{max}(\nu _t)=\frac{1}{4\pi }\frac{\sigma _2}{\sigma _1}\left(1\mathrm{\Phi }\left(\frac{\nu _t}{\sqrt{2(1\gamma ^2)}}\right)+\gamma e^{\frac{\nu _t^2}{2}}\left(1+\mathrm{\Phi }\left(\frac{\gamma \nu _t}{\sqrt{2(1\gamma ^2)}}\right)\right)\right)\hfill \\ N_{min}(\nu _t)=\frac{1}{4\pi }\frac{\sigma _2}{\sigma _1}\left(1\mathrm{\Phi }\left(\frac{\nu _t}{\sqrt{2(1\gamma ^2)}}\right)\gamma e^{\frac{\nu _t^2}{2}}\left(1\mathrm{\Phi }\left(\frac{\gamma \nu _t}{\sqrt{2(1\gamma ^2)}}\right)\right)\right)\hfill \\ \overline{N_{max}}=\overline{N_{min}}=\frac{1}{2\pi }\frac{\sigma _2}{\sigma _1}\hfill \end{array}`$ (25)
The Global Minkowski functionals for a one-dimensional Gaussian field are given by:
$`\begin{array}{c}M_1(\nu _t)=\frac{1}{2}\left(1\mathrm{\Phi }(\frac{\nu _t)}{\sqrt{2}})\right)\hfill \\ M_2(\nu _t))=f_{max}(\nu _t)f_{min}(\nu _t)=\gamma e^{\frac{\nu _t^2}{2}},\hfill \end{array}`$ (28)
where $`\mathrm{\Phi }(x)=\frac{2}{\sqrt{\pi }}_0^xe^{t^2}𝑑t`$.
As it has been already mentioned above, the last characteristic ($`M_2`$) can be written in terms of the peak distributions (A) <sup>14</sup>:
$`k(\nu _t)={\displaystyle \frac{N_{max}(\nu _t)}{N_{max}(\nu _t)N_{min}(\nu _t)}}`$ (29)
It is not difficult to demonstrate that for a Gaussian distribution all these characteristics are functions of only two variables: threshold $`\nu _t`$ and spectral parameter $`\gamma `$. This fact makes our analysis transparent.
To calculate the probability that the restored (filtered) signal is Gaussian, we use only the characteristic as mentioned under A. The distributions of maxima and minima, $`f_{max}(\nu _t,\gamma )`$ and $`f_{min}(\nu _t,\gamma )`$, above a threshold $`\nu _t`$ are monotonic functions of $`\nu _t`$ with
$`\begin{array}{c}f_{max}(\mathrm{},\gamma )=f_{min}(\mathrm{},\gamma )=1\hfill \\ f_{max}(+\mathrm{},\gamma )=f_{min}(+\mathrm{},\gamma )=0.\hfill \end{array}`$ (32)
For a limited number of peaks these distributions look like a step functions with number of steps equal to the total number of maxima and minima. If the observed signal has a cosmological nature, then this function should be consistent with the smooth analytical curve (see Eq. (16) and (20)) for $`\gamma `$ of the CMB spectrum. In case of real observations in a limited record we deal with bad statistics (small number of peaks). For this kind of statistics we have to perform an appropriate test, e.g. a Kolmogorov-Smirnov test <sup>17</sup>. Before applying such a test and to demonstrate the advantage of our method, we should do the following:
1. As representations of the spectrum of the filtered signal (being identical to the spectrum of the hypothetical CMB signal) we simulate a large number $`n_r`$ of different realizations of a random Gaussian field in the considered record using a FFT algorithm.
2. For each realization r=1,..,$`n_r`$ we calculate $`\gamma _r`$ using the definition (19) as well as $`f_{max}^r`$ and $`f_{min}^r`$ using Eq.(16). Furthermore we calculate the quantities:
$`\begin{array}{c}I_{max,real}(r,\gamma _r)=_5^{+5}f_{max}^r(\nu _t,\gamma _r)𝑑\nu _t\hfill \\ I_{max,anal}(r,\gamma _r)=_5^{+5}f_{max}(\nu _t,\gamma _r)𝑑\nu _t\hfill \\ I_{min,real}(r,\gamma _r)=_5^{+5}f_{min}^r(\nu _t,\gamma _r)𝑑\nu _t\hfill \\ I_{min,anal}(r,\gamma _r)=_5^{+5}f_{min}(\nu _t,\gamma _r)𝑑\nu _t,\hfill \end{array}`$ (37)
and
$`\begin{array}{c}\mathrm{\Delta }_{max}=I_{max,real}I_{max,anal}\hfill \\ \mathrm{\Delta }_{min}=I_{min,real}I_{min,anal},\hfill \end{array}`$ (40)
where $`f_{max}(\nu _t,\gamma _r)`$, $`f_{min}(\nu _t,\gamma _r)`$ have the analytical form (20) and $`f_{max}^r(\nu _t,\gamma _r)`$, $`f_{min}^r(\nu _t,\gamma _r)`$ are distributions of maxima and minima in a given realization. In (24) we performed integrations with the limits (-5,+5) where $`f_{max}`$ and $`f_{min}`$ in practice have reached their asymptotic values given by Eq.(23). Obviously, $`\gamma `$ will differ slightly from one realization to another because of the cosmic variance. This difference becomes smaller for longer records. Now we derive the formal probability that the distribution of peaks corresponds to a Gaussian distribution. There are different methods to do this. One of them is the following:
3. We count the fraction of realizations, where $`|\mathrm{\Delta }_{max}(r)|`$ and $`|\mathrm{\Delta }_{min}(r)|`$ are greater than some positive value $`z`$ and calculate $`P(z)`$
$`\begin{array}{c}P_{max}(z)=\frac{N(|\mathrm{\Delta }_{max}(r)|>z)}{n_r}\hfill \\ P_{min}(z)=\frac{N(|\mathrm{\Delta }_{min}(r)|>z)}{n_r}\hfill \end{array}`$ (43)
Here $`P(z)`$ represents a numerical estimate of how well the distribution of maxima and minima corresponds to a Gaussian signal.
4. For a given filtered record we can now estimate to what extent the signal has a Gaussian nature. We derive $`\mathrm{\Delta }_{max}`$ and $`\mathrm{\Delta }_{min}`$ to determine the probability $`P(z)`$. It is clear that if $`P(z)`$ is larger than $`3`$% and therefore $`z`$ is within the $`3\sigma `$ level of the distribution, there is a good chance that we observe a Gaussian signal. Alternatively one can calculate the differential functions $`n(\mathrm{\Delta })`$ of the distributions of $`\mathrm{\Delta }_{max}`$ and $`\mathrm{\Delta }_{min}`$ and consider the actual value $`n(\mathrm{\Delta })`$ for the filtered signal as a measure of closeness to a Gaussian signal. If this $`n(\mathrm{\Delta })`$ is greater than a few percent the filtered signal is Gaussian within the ”$`3\sigma `$ level” using the analogy with a standard Gaussian distribution.
One could also use a Kolmogorov-Smirnov test, some of its generalizations or other methods to be discussed in a separate paper.
In the next section we consider the application of this technique to a concrete toy model of an observed record.
## 4 Application of the power filtration to a toy model of an observational record
Now we demonstrate the application of the $`G_p`$ filtration and estimate to what extent we can recover the Gaussian nature of the signal. Furthermore we compare with the result from a Wiener filtration.
We consider a CMB spectrum of a standard Cold Dark Matter cosmological model with parameters: $`h_{100}=0.75`$, $`\mathrm{\Omega }=1`$, $`\mathrm{\Omega }_b=0.0125`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$. For this spectrum we simulate $`n_r`$ = 1000 different realizations of the random Gaussian field (using a FFT algorithm) for a record with length $`360^o`$ and with FWHM of the antenna beam equal to $`0.5^o`$. Using the definitions in the previous section we calculate the values of $`\gamma _r`$, $`I_{max}`$, $`I_{min}`$ for each realization. These values are slightly different from one record to another because of the cosmic variance. In Fig. 1 we show the dependence of $`I_{max}`$ and $`I_{min}`$ on $`\gamma _r`$ for the realizations. In the same Fig. 1 we also show the corresponding Gaussian dependences.
Now we consider one specific realization (in analogy to a realization on the sky) which is marked by a star on Fig. 1 ($`\gamma `$= 0.3738, $`I_{max}`$= 2.751 and $`I_{min}`$= 2.271). Next we compute for this realization $`f(\nu _t)`$ (see Fig. 2a), $`M_1(\nu _t)`$ (see Fig. 2b), rate of clusterisation ($`N_{max}/(N_{max}N_{min})`$ and $`N_{min}/(N_{min}N_{max})`$ as a function of $`\nu _t`$ (see Fig. 2c), and $`M_2(\nu _t)`$ (see Fig. 2d). On Fig. 2(a-d) we also show the corresponding analytical dependence (see Eq.(25) and Eq.(28)). The graphs for the realization on Fig. 2(a-d) are only slightly different from the corresponding analytical curves given by dashed lines.
Then we add to the CMB signal a signal from 100 identical foreground point-like sources with a random distribution over the record and with an amplitude such that the rms of the maps increases by a factor $`\sqrt{2}`$ relative to the pure CMB signal. We consider the signal from the point-like sources as “noise”. The signal/noise ratio in this example is thus equal to 1. Of course this is not a representation of a realistic noise but only used to demonstrate the efficiency of the power filter $`G_p`$. Discussion of more realistic models will be given in a separate paper. The power spectrum of the CMB signal plus noise is computed for the same antenna beam with FWHM $`0.5^o`$. For this realization we determine $`\gamma _r^{CMB+noise}`$ and $`I_{max}`$ and $`I_{min}`$. The corresponding points are marked by crosses on Fig. 1 for the parameters $`\gamma `$= 0.4833, $`I_{max}`$= 2.871 and $`I_{min}`$= 2.248.
It is seen that the crosses are far away from the clouds of realizations of the CMB and far from the Gaussian points (marked by dots with the same $`\gamma _r^{CMB+noise}`$ on Fig. 1). For this realization we calculated $`f_{max}`$, $`f_{min}`$, $`M_1`$, the rate of clusterisation, and $`M_2`$ and the corresponding analytical Gaussian functions for the given $`\gamma _r^{CMB+noise}`$. The corresponding plots are presented on Fig. 3. One can see that the distributions of these values on Fig. 3(a-d) differ significantly from the corresponding Gaussian functions. This means that the CMB+noise signal is far from being Gaussian. The same is seen from Fig.1. Now we perform filtration of this CMB+noise signal using the power filter $`G_p`$. As a result we expect to obtain a reconstructed signal with the power spectrum nearly identical to the CMB power spectrum of the considered realization. The corresponding points for $`I_{max}`$ and $`I_{min}`$ are indicated in Fig. 1 by circles with the parameters $`\gamma `$= 0.3930, $`I_{max}`$= 2.782 and $`I_{min}`$= 2.272.
It is seen that the circles in fact are within the corresponding clouds of numerical realizations of the CMB signal and close to the corresponding Gaussian points with the same $`\gamma `$.
In Fig. 4 we have plotted $`f_{max}`$, $`f_{min}`$, $`M_1`$, the rate of clusterisation, and $`M_2`$ for the reconstructed signal as well as the corresponding Gaussian analytical functions. One sees that the reconstructed realization is considerably closer to Gaussianity than the CMB+noise signal, demonstrating how effective the power filter works. It is particularly important that the filtered signal has practically the same $`\gamma `$ as the input CMB signal. Furthermore we may get a numerical estimate of how good the power filtration really is by using the method described in Section 3. The values of $`n(\mathrm{\Delta })`$ for the filtered signal are $`n(\mathrm{\Delta }_{max})`$= 0.12 and $`n(\mathrm{\Delta }_{min})`$= 0.51.
For comparison we have performed a Wiener filtration of the same realization of CMB+noise signal and computed $`\gamma ^W`$, $`I_{max}^W`$ and $`I_{min}^W`$. The values ($`\gamma `$= 0.6232, $`I_{max}`$= 2.930 and $`I_{min}`$= 2.067) are plotted as triangles on Fig. 1 together with the Gaussian values for the same $`\gamma ^W`$ as dots.
One can see that the Wiener filter displaces the corresponding points far away from the clouds of realizations of the CMB signal and that the power spectrum of the reconstructed signal is very far to be spectrum of the CMB signal.
## 5 Conclusions
In this paper we propose to use a power filter <sup>8</sup> for linear reconstruction of the CMB signal in one-dimensional scans. The specific property of the $`G_p`$ filter is that it preserves the power spectrum of the CMB signal. Under conditions described in Section 2 we get $`G_p=W^{1/2}`$ where $`W`$ is the Wiener filter. Unlike the power filter $`G_p`$ the Wiener filter $`W`$ diminishes the power spectrum of the reconstructed signal compared to the CMB power spectrum. In section 4 we have demonstrated that peak statistics and cluster analysis can be used to estimate the efficiency of the power filtration and to estimate the probability that a CMB signal is present in an observational record. In Section 4 we also demonstrated how the power filter works using a toy model of an observational record consisting of the CMB signal and noise in the form of point like sources. We note that in the example of Section 4 we used a signal/noise ratio equal to unity as described in section 4, but tests with different toy models have demonstrated that this method works well also in the case with this ratio significantly less than unity. Discussions of more realistic models of the noise as well as application of the described method to real observations will be done in subsequent papers.
Acknowledgments
P. Naselsky and D. Novikov are grateful to the staff of TAC and the Astronomical Observatory of the Copenhagen University for providing excellent working conditions during their visit to these institutions. This investigation was supported in part by a grant ISF MEZ 300, by a grant INTAS 97-1192, by the Danish Natural Science Research Council through grant No 9701841, by Danmarks Grundforskningsfond through its support for establishment of the Theoretical Astrophysics Center, by Alexander von Humboldt-Stiftung and by the Board of the Glasstone Benefaction.
References
1. Hu., W. home page:www.sns.ias.edu/ whu/physics/physics.html
2. Tegmark., M., ApJ. Lett 480, L87, 1997.
3. Tegmark., M. and G. Efstathiou, MNRAS 281 1297, 1996.
4. Bond., J.R., A.N. Jaffe and L. Knox, Phys. Rev. D. 57 2117, 1998.
5. Knox, L., astro-ph /9902046 v 2.28 Sep. 1999.
6. Taylor., A., A. Heavens, B. Ballinger and M. Tegmark: in Proceedings of the Particle Physics and Early Universe Conference (PPPEUC), University of Cambridge, 7-11 April 1997.
7. Gorski, K.M., Proceedings of the XXXI-st Recontres de Marion Astrophysics Meeting, p.77, Editions Frontieres, (1997). astro-ph/9701191.
8. Andrews, H.C. and B.R. Hunt. Digital image restoration. New Jersey: Pentice-Hall, 1977.
9. Minkowski., H, Mathematische Annalen, 57, 443, 1903.
10. Schmalzing., J. and K. Gorski, MNRAS 297, 335, 1998.
11. Naselsky., P. and D. Novikov, ApJ., 507, 31, 1998.
12. Novikov, D., H. Feldman, S. Shandarin, Int. Journal of Modern Physics D., 8 N3, 291, 1999.
13. Naselsky., P. and D. Novikov, ApJ. Lett., 444, L1, 1996.
14. Novikov, D.I. and H.E. Jørgensen, Int. Journal of Modern Physics D., 5 N4, 319, 1996.
15. Delabrouilee, J., K. Gorski and E. Hivon, MNRAS 298, 445, 1998.
16. Doroshkevich A.G., Astrophysics 6, 320, 1970.
17. Press, W.H., S.A. Teukolsky, W.T. Vetteling and B.P. Flannery, Numerical Recipes. Cambridge University Press. 1992. |
warning/0001/nucl-th0001040.html | ar5iv | text | # I Introduction
## I Introduction
Single-gluon exchange between two quarks is attractive in the color anti-triplet channel. Therefore, sufficiently cold and dense quark matter is a color superconductor .
In some aspects, color superconductivity is similar to ordinary (BCS) superconductivity . For instance, like electrons in a BCS superconductor, quarks form Cooper pairs. At zero temperature, $`T=0`$, the ground state of the system is no longer a Fermi sea of quarks (and a Dirac sea of antiquarks), but a Bose condensate of quark Cooper pairs. In the normal phase the excitation of a particle–hole pair at the Fermi surface costs no energy. In the superconducting phase, however, exciting a quasiparticle–quasiparticle-hole pair costs at least an energy $`2\varphi _0`$, where $`\varphi _0`$ is the zero-temperature gap. Another similarity between color and BCS superconductivity is that, in weak coupling, the critical temperature $`T_c`$ for “melting” the Cooper pair condensate is $`T_c0.57\varphi _0`$ .
There are, however, also fundamental differences between color and BCS superconductivity. First of all, a BCS superconductor requires the presence of an atomic lattice with phonons that cause electrons to form Cooper pairs. On the other hand, in QCD gluons themselves cause quarks to condense. Another difference is that in BCS theory the zero-temperature gap depends on the BCS coupling constant $`G`$ as $`\varphi _0\mu \mathrm{exp}(c_{\mathrm{BCS}}/G^2)`$ , where $`\mu `$ is the chemical potential, and $`c_{\mathrm{BCS}}=`$const., while in a color superconductor, $`\varphi _0\mu \mathrm{exp}(c_{\mathrm{QCD}}/g)`$ , where $`g`$ is the QCD coupling constant, and $`c_{\mathrm{BCS}}c_{\mathrm{QCD}}=`$ const.
The physical reason for the change in the parametric dependence on the coupling constant is that, because gluons are massless, gluon-mediated interactions are long-range, in contrast to BCS theory, where phonon exchange is typically assumed to be a point-like interaction . The long-range nature of gluon exchange manifests itself in the infrared singular behavior of the gluon propagator. This enhances the contribution of very soft, collinear gluons in the gap equations , and causes the $`1/g`$ in the exponent, instead of a $`1/g^2`$ which would appear if gluons were massive , or gluon exchange a point-like interaction as assumed in Nambu–Jona-Lasinio-type approaches to color superconductivity .
Some care has to be taken in determining the coefficient $`c_{\mathrm{QCD}}`$. This constant differs when one uses the free gluon propagator in the solution of the gap equations instead of a propagator which takes into account the presence of the cold and dense quark medium. By now, several authors have confirmed Son’s original result $`c_{\mathrm{QCD}}=3\pi ^2/\sqrt{2}`$ , obtained by using the gluon propagator in the so-called “hard-dense-loop” (HDL) limit . The gluon propagator in the HDL limit is obtained by resummation of the gluon self-energy, computed to one-loop order for gluon energies $`p_0`$ and momenta $`p`$ that are much smaller than the quark chemical potential $`\mu `$.
In weak coupling, the temperatures where quark matter is color-superconducting are much smaller than the quark chemical potential, $`T\varphi _0\mu \mathrm{exp}(c_{\mathrm{QCD}}/g)\mu `$. Therefore, to leading order the contributions of gluon and ghost loops to the one-loop gluon self-energy can be neglected, and the main contribution comes from the quark loop. This is very similar to ordinary superconductivity, where the one-loop photon self-energy is determined by an electron loop.
In the standard HDL approximation, however, the quark excitations in the loop are considered to be those of the normal and not of the superconducting phase. This is in principle inconsistent. The aim of the present work is to amend this shortcoming and to compute the gluon self-energy in the color-superconducting phase.
For the sake of definiteness, I consider a color superconductor with $`N_f=2`$ flavors of massless quarks, and assume that quarks condense in a channel with total spin $`J=0`$ and even parity. In this case, the quark-quark condensate breaks $`SU(3)_c`$ to $`SU(2)_c`$. Consequently, one expects that the three gluons of the unbroken $`SU(2)_c`$ subgroup remain massless, while the other five gluons of the original $`SU(3)_c`$ obtain masses through the Anderson–Higgs mechanism. It is therefore necessary to consider different gluon colors separately.
I derive a general expression for the quark contribution to the gluon self-energy, and study the limit where the gluon energy $`p_0=0`$ and the gluon momentum $`p0`$. For electric gluons, this limit gives the Debye mass, while for magnetic gluons, it gives the Meissner mass. I also consider the limit where $`p_0=0`$, but $`p\varphi _0`$. In this case, the gluon momentum is large enough to resolve individual quarks in a Cooper pair; consequently, the Debye masses approach their values in the normal phase and the Meissner effect vanishes.
Debye screening of static color-electric fields and the Meissner effect for static color-magnetic fields are in principle quite analogous to Debye screening and the Meissner effect for electromagnetic fields in ordinary superconductors . However, the somewhat more complicated color and flavor structure of a quark-quark condensate in comparison to an electron-electron condensate gives rise to an additional degree of complexity. While studying these effects in a color superconductor is interesting in itself, they might have, however, far greater implications for color superconductivity than the corresponding effects in ordinary superconductors: unlike photons, gluons themselves are responsible for condensation of quark pairs. The modification of the gluon self-energy in the superconducting phase directly enters the gap equation through the gluon propagator, and so might change the value for the gap. On the other hand, the influence of the photon self-energy on electron condensation is at best a higher order effect.
Although effects from quark condensation in the gluon propagator vanish for large gluon energies and momenta, one can a priori not exclude that they will not change the solution of the gap equations. For instance, to assess the importance of the Meissner effect, note that, in the HDL approximation, the main contribution to the gap equations comes from color-magnetic fields with momenta $`p(m_g^2\varphi _0)^{1/3}\varphi _0`$, where $`m_g`$ is the gluon mass . As will be seen below, the Meissner effect is small, but not absent, at the same momentum scale. This means that the Meissner effect can indeed influence the solution of the gap equation. A first estimate of this effect (neglecting the color-flavor structure of the condensate and considering only the dominant contribution to the gluon self-energy) was given in , and a reduction of the zero-temperature gap was found.
This paper is organized as follows. In Sec. II a compact derivation of the quark contribution to the gluon self-energy is presented, mainly to introduce the notation and the concept of Nambu–Gor’kov spinors , which considerably simplify calculations at nonzero chemical potential. In Sec. III the quark contribution to the gluon self-energy is explicitly computed in the normal phase. The HDL limit is derived to show that the Nambu–Gor’kov method indeed gives the correct answer. Section IV generalizes the previous results to the superconducting phase. In Sec. V, the zero-energy, zero-momentum limit of the gluon self-energy is studied, which yields the Debye as well as the Meissner masses in the superconducting phase. Section VI discusses how, for nonzero gluon momenta $`p\varphi _0`$, the Debye masses approach their values in the normal phase, and the Meissner effect vanishes. Readers not interested in technical details should skip Secs. II to VI and move on to Sec. VII, where the main results of this work are summarized, conclusions are drawn, and an outlook for future studies is given.
I use natural units, $`\mathrm{}=c=k_B=1`$, and work in Euclidean space-time $`𝐑^4V/T`$, where $`V`$ is the volume and $`T`$ the temperature of the system. Nevertheless, I find it convenient to retain the Minkowski notation for 4-vectors, with a metric tensor $`g^{\mu \nu }=\mathrm{diag}(+,,,)`$. For instance, the space-time coordinate vector is $`x^\mu (t,𝐱)`$, $`ti\tau `$, where $`\tau `$ is Euclidean time. 4-momenta are denoted as $`K^\mu (k_0,𝐤)`$, $`k_0i\omega _n`$, where $`\omega _n`$ is the Matsubara frequency, $`\omega _n2n\pi T`$ for bosons and $`\omega _n(2n+1)\pi T`$ for fermions, $`n=0,\pm 1,\pm 2,\mathrm{}`$. The absolute value of the 3-momentum $`𝐤`$ is denoted as $`k|𝐤|`$, and its direction as $`\widehat{𝐤}𝐤/k`$.
## II The generating functional at nonzero chemical potential
Consider QCD with $`N_f`$ quark flavors, at nonzero chemical potential. The $`N_f\times N_f`$ matrix of quark masses $`m_f`$ will be denoted as $`m\mathrm{diag}(m_1,m_2,\mathrm{},m_{N_f})`$. Let us consider a color neutral system, i.e., there is no chemical potential for color, however, there can be in general a chemical potential $`\mu _f`$ for each quark flavor $`f`$. Let us denote the $`N_f\times N_f`$ chemical potential matrix as $`\mu \mathrm{diag}(\mu _1,\mu _2,\mathrm{},\mu _{N_f})`$. Then, the generating functional for the $`N`$–point functions of the theory reads (normalization factors are suppressed)
$`𝒵[J,\overline{\eta },\eta ]`$ $`=`$ $`{\displaystyle 𝒟U[A]\mathrm{exp}\left[_x\left(_A+J_\mu ^aA_a^\mu \right)\right]𝒵[A,\overline{\eta },\eta ]},`$ (2)
$`𝒵[A,\overline{\eta },\eta ]`$ $`=`$ $`{\displaystyle 𝒟\overline{\psi }𝒟\psi \mathrm{exp}\left\{_x\left[\overline{\psi }\left(i\gamma ^\mu _\mu +\mu \gamma _0m+g\gamma ^\mu A_\mu ^aT_a\right)\psi +\overline{\eta }\psi +\overline{\psi }\eta \right]\right\}}.`$ (3)
Here, $`𝒟U[A]`$ is the gauge invariant measure for the integration over the gauge fields $`A_a^\mu `$. The space-time integration is defined as $`_x_0^{1/T}𝑑\tau _Vd^3𝐱`$. $`g`$ is the QCD coupling constant, $`\gamma ^\mu `$ are the Dirac matrices, and $`T_a=\lambda _a/2`$ the generators of $`SU(N_c)`$; for QCD, $`N_c=3`$, and $`\lambda _a`$ are the Gell-Mann matrices. The quark fields $`\psi `$ (as well as the external fields $`\eta `$) are $`4N_cN_f`$-component spinors, i.e., they carry Dirac indices $`\alpha =1,\mathrm{},4`$, fundamental color indices $`i=1,\mathrm{},N_c`$, and flavor indices $`f=1,\mathrm{},N_f`$. The Lagrangian for the gauge fields consists in general of three parts,
$$_A=_F+_{\mathrm{gf}}+_{FPG},$$
(4)
where
$$_F=\frac{1}{4}F_a^{\mu \nu }F_{\mu \nu }^a$$
(5)
is the gauge field part, $`F_{\mu \nu }^a=_\mu A_\nu ^a_\nu A_\mu ^a+gf^{abc}A_\mu ^bA_\nu ^c`$ is the field strength tensor. The parts corresponding to gauge fixing, $`_{\mathrm{gf}}`$, and to Fadeev–Popov ghosts, $`_{FPG}`$, need not be specified: it will be seen that they are inconsequential for the following.
In the vacuum, the ground state of the system consists of the Dirac sea, i.e., all negative energy (antiquark) states are occupied, while all positive energy (quark) states are empty. At zero temperature and nonzero chemical potential, $`\mu _f>0`$, however, the ground state consists of the Dirac sea and the Fermi sea, i.e., positive energy states which are occupied up to the Fermi energy $`\mu _f`$. Formally, this is expressed by the term $`\overline{\psi }\mu \gamma _0\psi `$ in the generating functional (3), which ensures that the energy of excited states of flavor $`f`$ is measured with respect to the Fermi energy $`\mu _f`$, and not with respect to the vacuum at zero density.
This shift of the energy scale introduces an apparent asymmetry. One can restore the symmetry by the following trick. Introduce $`M`$ identical copies (“replicas”) of the original quark fields. All copies are supposed to interact with the gluon field in the same way. At the end, after having computed $`N`$-point functions for this extended system, $`M`$ will be set equal to 1. The generating functional (3) for the quark part is replaced by
$$𝒵[A,\overline{\eta },\eta ]𝒵_M[A,\overline{\eta },\eta ]\left(𝒵[A,\overline{\eta },\eta ]\right)^M.$$
(6)
Now define the charge conjugate spinors $`\psi _C,\overline{\psi }_C`$ through
$$\psi C\overline{\psi }_C^T,\overline{\psi }\psi _C^TC,$$
(7)
where $`Ci\gamma ^2\gamma _0`$ is the charge conjugation matrix; $`C=C^1=C^T=C^{}`$. In half of the $`M`$ copies in Eq. (6), replace $`\overline{\psi }`$, $`\psi `$ by the charge conjugate spinors $`\overline{\psi }_C`$, $`\psi _C`$. Using $`C\gamma _\mu C^1=\gamma _\mu ^T`$, and the anticommutation property of the (Grassmann-valued) quark spinors, one obtains after an integration by parts (and disregarding the overall normalization)
$`𝒵_M[A,\overline{\eta },\eta ,\overline{\eta }_C,\eta _C]`$ $`=`$ $`({\displaystyle }𝒟\overline{\psi }𝒟\psi 𝒟\overline{\psi }_C𝒟\psi _C\mathrm{exp}\{{\displaystyle _x}[\overline{\psi }(i\gamma ^\mu _\mu +\mu \gamma _0m+gA_\mu ^a\mathrm{\Gamma }_a^\mu )\psi `$ (9)
$`+\overline{\psi }_C(i\gamma ^\mu _\mu \mu \gamma _0m+gA_\mu ^a\overline{\mathrm{\Gamma }}_a^\mu )\psi _C+\overline{\eta }\psi +\overline{\psi }\eta +\overline{\eta }_C\psi _C+\overline{\psi }_C\eta _C]{\displaystyle \frac{}{}}\})^{M/2}.`$
Here,
$$\mathrm{\Gamma }_a^\mu \gamma ^\mu T_a,\overline{\mathrm{\Gamma }}_a^\mu C(\gamma ^\mu )^TC^1T_a^T\gamma ^\mu T_a^T,$$
(10)
and charge conjugate external fields $`\overline{\eta }_C`$ and $`\eta _C`$ were defined analogous to Eq. (7). Let us now introduce the $`8N_cN_f`$-component (Nambu–Gor’kov) spinors
$$\mathrm{\Psi }\left(\begin{array}{c}\psi \\ \psi _C\end{array}\right),\overline{\mathrm{\Psi }}(\overline{\psi },\overline{\psi }_C),H\left(\begin{array}{c}\eta \\ \eta _C\end{array}\right),\overline{H}(\overline{\eta },\overline{\eta }_C),$$
(11)
and the $`8N_cN_f\times 8N_cN_f`$-dimensional inverse propagator
$$𝒮_0^1(x,y)\left(\begin{array}{cc}[G_0^+]^1(x,y)& 0\\ 0& [G_0^{}]^1(x,y)\end{array}\right),$$
(12)
where
$$[G_0^\pm ]^1(x,y)i\left(i\gamma _\mu _x^\mu \pm \mu \gamma _0m\right)\delta ^{(4)}(xy)$$
(13)
is the inverse propagator for non-interacting quarks (upper sign) or charge conjugate quarks (lower sign), respectively. Furthermore, denote
$$\widehat{\mathrm{\Gamma }}_a^\mu \left(\begin{array}{cc}\mathrm{\Gamma }_a^\mu & 0\\ 0& \overline{\mathrm{\Gamma }}_a^\mu \end{array}\right).$$
(14)
Then, the generating functional (9) can be written in the compact form
$`𝒵_M[A,\overline{H},H]`$ (15)
$`=`$ $`{\displaystyle \underset{r=1}{\overset{M/2}{}}𝒟\overline{\mathrm{\Psi }}_r𝒟\mathrm{\Psi }_r\mathrm{exp}\left\{\underset{r=1}{\overset{M/2}{}}\left[_{x,y}\overline{\mathrm{\Psi }}_r(x)𝒮_0^1(x,y)\mathrm{\Psi }_r(y)+_x\left(g\overline{\mathrm{\Psi }}_rA_\mu ^a\widehat{\mathrm{\Gamma }}_a^\mu \mathrm{\Psi }_r+\overline{H}_r\mathrm{\Psi }_r+\overline{\mathrm{\Psi }}_rH_r\right)\right]\right\}}.`$ (16)
In this form, all reference to the chemical potentials $`\mu _f`$ has been absorbed in the inverse propagator (12). Therefore, the generating functional for QCD, Eq. (2) with (16), is formally identical to that at zero chemical potential. The apparent asymmetry introduced by a nonzero chemical potential $`\mu _f`$ has been restored by the introduction of charge conjugate fields; the associated charge conjugate propagator $`G_0^{}`$ appears on equal footing with the ordinary propagator $`G_0^+`$.
## III The gluon self-energy in the normal phase
The gluon self-energy is defined as
$$\mathrm{\Pi }\mathrm{\Delta }^1\mathrm{\Delta }_0^1,$$
(17)
where $`\mathrm{\Delta }^1`$ is the resummed and $`\mathrm{\Delta }_0^1`$ the free inverse gluon propagator; for instance, in momentum space and in covariant gauge,
$$[\mathrm{\Delta }_0^1]_{ab}^{\mu \nu }(P)=\delta _{ab}\left(P^2g^{\mu \nu }+\frac{1\alpha }{\alpha }P^\mu P^\nu \right).$$
(18)
To one-loop order, the gluon self-energy receives contributions from gluon loops (through the 3-gluon and 4-gluon vertices), ghost loops (through the ghost-gluon vertex), and quark loops (through the quark-gluon vertex),
$$\mathrm{\Pi }=\mathrm{\Pi }_g+\mathrm{\Pi }_{FPG}+\mathrm{\Pi }_q+O(g^3).$$
(19)
$`\mathrm{\Pi }_g`$ and $`\mathrm{\Pi }_{FPG}`$ are independent of $`\mu `$, effects from nonzero chemical potential enter only through $`\mathrm{\Pi }_q`$. For dimensional reasons,
$$\mathrm{\Pi }_g,\mathrm{\Pi }_{FPG}g^2T^2,\mathrm{\Pi }_qg^2(\mu ^2+aT^2),$$
(20)
with some constant $`a`$.
The superconducting condensate melts when the temperature $`T`$ exceeds the critical temperature $`T_c0.57\varphi _0`$ , where $`\varphi _0`$ is the magnitude of the superconducting gap at $`T=0`$. In weak coupling QCD, $`\varphi _0\mu \mathrm{exp}(c_{\mathrm{QCD}}/g)\mu `$ , and temperature effects can be neglected to leading order. This means that, for the temperatures of interest in this work, one can neglect the contributions from gluon and ghost loops to the gluon self-energy, and consider the quark contribution only, $`\mathrm{\Pi }\mathrm{\Pi }_q`$.
Due to the aforementioned similarity between the generating functional (2), with the quark part (16), and the one at zero chemical potential, it is not difficult to derive the quark contribution to the one-loop gluon self-energy. If there is no superconducting condensate, this contribution is
$$\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }(x,y)\frac{M}{2}g^2\mathrm{Tr}_{s,c,f,NG}\left[\widehat{\mathrm{\Gamma }}_a^\mu 𝒮_0(x,y)\widehat{\mathrm{\Gamma }}_b^\nu 𝒮_0(y,x)\right].$$
(21)
Here, the factor $`M/2`$ arises from the fact that there are $`M/2`$ identical species of quarks described by spinors $`\mathrm{\Psi }_r`$ in Eq. (16), which contribute to the gluon self-energy. In the following, set $`M=1`$, to recover the original theory. The trace in Eq. (21) is taken over $`4`$-dimensional spinor space, $`N_c`$-dimensional color space, $`N_f`$-dimensional flavor space, and the $`2`$-dimensional space of regular and charge-conjugate spinors (Nambu–Gor’kov space).
In the following, the self-energy (21) is evaluated in momentum space. Use will be made of translational invariance, $`𝒮_0(x,y)𝒮_0(xy)`$, cf. Eq. (13), and of the Fourier transforms
$`𝒮_0(x)`$ $`=`$ $`{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}e^{iKx}𝒮_0(K),`$ (23)
$`i\delta ^{(4)}(x)`$ $``$ $`\delta ^{(3)}(𝐱)\delta (\tau )={\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}e^{iKx},`$ (24)
$`{\displaystyle _x}e^{iKx}`$ $`=`$ $`{\displaystyle \frac{V}{T}}\delta _{K,0}^{(4)},`$ (25)
where $`_K_nVd^3𝐤/(2\pi )^3`$. Here, the quark propagator in momentum space is
$$𝒮_0(K)\left(\begin{array}{cc}G_0^+(K)& 0\\ 0& G_0^{}(K)\end{array}\right),G_0^\pm (K)(\gamma ^\mu K_\mu \pm \mu \gamma _0m)^1.$$
(26)
Then, the gluon self-energy in momentum space is
$$\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }(P)=\frac{1}{2}g^2\frac{T}{V}\underset{K}{}\mathrm{Tr}_{s,c,f,NG}\left[\widehat{\mathrm{\Gamma }}_a^\mu 𝒮_0(K)\widehat{\mathrm{\Gamma }}_b^\nu 𝒮_0(KP)\right].$$
(27)
As a warm-up exercise, and to confirm that the method of the Nambu–Gor’kov propagators indeed gives the correct answer, let us derive from Eq. (27) the standard hard-dense-loop (HDL) result for the quark contribution to the gluon self-energy. To see the analogy to the computation in the superconducting phase, cf. Sec. IV, Eq. (27) will be evaluated in several steps.
### A Trace over Nambu–Gor’kov space
First perform the trace over Nambu–Gor’kov space. With Eqs. (14) and (26), one obtains
$$\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }(P)=\frac{1}{2}g^2\frac{T}{V}\underset{K}{}\mathrm{Tr}_{s,c,f}\left[\mathrm{\Gamma }_a^\mu G_0^+(K)\mathrm{\Gamma }_b^\nu G_0^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu G_0^{}(K)\overline{\mathrm{\Gamma }}_b^\nu G_0^{}(KP)\right].$$
(28)
### B Trace over flavor space
The vertices $`\mathrm{\Gamma }_a^\mu `$ and $`\overline{\mathrm{\Gamma }}_a^\mu `$ are diagonal in flavor space,
$$\left(\mathrm{\Gamma }_a^\mu \right)_{fg}=\delta _{fg}\mathrm{\Gamma }_a^\mu ,\left(\overline{\mathrm{\Gamma }}_a^\mu \right)_{fg}=\delta _{fg}\overline{\mathrm{\Gamma }}_a^\mu .$$
(29)
The free propagators $`G_0^\pm `$ are also diagonal in flavor space, but for $`\mu _f\mu _g`$, $`fg`$, $`f,g\{1,\mathrm{},N_f\}`$, the diagonal components are in general not equal. To proceed, assume that all chemical potentials are equal, $`\mu _1=\mu _2=\mathrm{}=\mu _{N_f}\mu `$, such that
$$\left(G_0^\pm \right)_{fg}=\delta _{fg}G_0^\pm .$$
(30)
(For notational convenience, I am somewhat sloppy with indices here and throughout the rest of the paper: I use the same symbol, $`G_0^\pm `$, for the $`4N_cN_f\times 4N_cN_f`$ matrix on the left-hand side of this equation and for the $`4N_c\times 4N_c`$ matrix on the right-hand side.) Thus, the trace over flavor space simply gives a factor $`N_f`$,
$$\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }(P)=\frac{1}{2}g^2N_f\frac{T}{V}\underset{K}{}\mathrm{Tr}_{s,c}\left[\mathrm{\Gamma }_a^\mu G_0^+(K)\mathrm{\Gamma }_b^\nu G_0^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu G_0^{}(K)\overline{\mathrm{\Gamma }}_b^\nu G_0^{}(KP)\right].$$
(31)
This expression is easily generalized to the case where the chemical potentials are not equal. Then, instead of the prefactor $`N_f`$ one would have a sum over flavors $`f`$, where the value of the chemical potential in the propagators $`G_0^\pm `$ in the $`f`$th term of the sum is equal to $`\mu _f`$.
### C Trace over color space
The free quark propagator is diagonal in (fundamental) color space,
$$\left(G_0^\pm \right)_{ij}=\delta _{ij}G_0^\pm .$$
(32)
The only nontrivial color structure thus arises from the generators of $`SU(3)_c`$. On account of
$$\mathrm{Tr}_c(T_aT_b)=\mathrm{Tr}_c(T_aT_b)^T=\mathrm{Tr}_c(T_a^TT_b^T)=\frac{1}{2}\delta _{ab},$$
(33)
one obtains
$`\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }(P)`$ $`=`$ $`\delta _{ab}\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }(P),`$ (35)
$`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s\left[\gamma ^\mu G_0^+(K)\gamma ^\nu G_0^+(KP)+\gamma ^\mu G_0^{}(K)\gamma ^\nu G_0^{}(KP)\right].`$ (36)
### D Mixed representations for the quark propagators
To proceed, let us assume that the quarks are massless, $`m=0`$. Then, write the quark propagator as
$$G_0^\pm (K)=\underset{e=\pm }{}\frac{k_0(\mu ek)}{k_0^2[ϵ_{𝐤0}^e]^2}\mathrm{\Lambda }_𝐤^{\pm e}\gamma _0,$$
(37)
where
$$ϵ_{𝐤0}^e|\mu ek|,$$
(38)
and
$$\mathrm{\Lambda }_𝐤^e\frac{1}{2}\left(1+e\gamma _0𝜸\widehat{𝐤}\right)$$
(39)
are projectors onto states of positive ($`e=+`$) or negative ($`e=1`$) energy. Now introduce a mixed representation for the quark propagator,
$$G_0^\pm (\tau ,𝐤)T\underset{k_0}{}e^{k_0\tau }G_0^\pm (K),G_0^\pm (K)_0^{1/T}𝑑\tau e^{k_0\tau }G_0^\pm (\tau ,𝐤).$$
(40)
After performing the Matsubara sum in terms of a contour integral in the complex $`k_0`$ plane, one obtains
$`G_0^+(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{e=\pm }{}}\mathrm{\Lambda }_𝐤^e\gamma _0\left\{{\displaystyle \frac{}{}}(1n_{𝐤0}^e)\left[\theta (\tau )N(ϵ_{𝐤0}^e)\right]\mathrm{exp}(ϵ_{𝐤0}^e\tau )n_{𝐤0}^e\left[\theta (\tau )N(ϵ_{𝐤0}^e)\right]\mathrm{exp}(ϵ_{𝐤0}^e\tau )\right\},`$ (42)
$`G_0^{}(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{e=\pm }{}}\gamma _0\mathrm{\Lambda }_𝐤^e\left\{{\displaystyle \frac{}{}}n_{𝐤0}^e\left[\theta (\tau )N(ϵ_{𝐤0}^e)\right]\mathrm{exp}(ϵ_{𝐤0}^e\tau )(1n_{𝐤0}^e)\left[\theta (\tau )N(ϵ_{𝐤0}^e)\right]\mathrm{exp}(ϵ_{𝐤0}^e\tau )\right\}.`$ (43)
Here, $`N(x)(e^{x/T}+1)^1`$, and
$$n_{𝐤0}^e\frac{ϵ_{𝐤0}^e+\mu ek}{2ϵ_{𝐤0}^e}$$
(44)
are the occupation numbers of particles ($`e=+1`$) or antiparticles ($`e=1`$) at zero temperature. Consequently, $`1n_{𝐤0}^e`$ are the occupation numbers for particle-holes or antiparticle-holes.
Note that
$$G_0^\pm (\tau ,𝐤)=\gamma _0G_0^{}(\tau ,𝐤)\gamma _0.$$
(45)
For $`0\tau 1/T`$, one derives with $`1N(x)=N(x)e^{x/T}`$
$$G_0^\pm (\frac{1}{T}\tau ,𝐤)=G_0^\pm (\tau ,𝐤),$$
(46)
the well-known Kubo–Martin–Schwinger relation for fermions .
Using the fact that $`n_{𝐤0}^e\theta (\mu ek)`$, and $`N(x)=1N(x)`$, the propagators (40) can be cast into the more familiar form
$`G_0^+(\tau ,𝐤)`$ $`=`$ $`\mathrm{\Lambda }_𝐤^+\gamma _0\left[\theta (\tau )N_F^+(k)\right]e^{(k\mu )\tau }+\mathrm{\Lambda }_𝐤^{}\gamma _0\left[\theta (\tau )N_F^{}(k)\right]e^{(k+\mu )\tau },`$ (48)
$`G_0^{}(\tau ,𝐤)`$ $`=`$ $`\gamma _0\mathrm{\Lambda }_𝐤^+\left[\theta (\tau )N_F^+(k)\right]e^{(k\mu )\tau }\gamma _0\mathrm{\Lambda }_𝐤^{}\left[\theta (\tau )N_F^{}(k)\right]e^{(k+\mu )\tau },`$ (49)
where $`N_F^\pm (k)N(k\mu )`$ is the Fermi–Dirac distribution function for particles (antiparticles). However, in view of the application to the superconducting phase in Sec. IV, it is advantageous to continue to use the form (40).
Denoting $`K_1K`$ and $`K_2KP`$, one computes the expressions
$$T\underset{k_0}{}\mathrm{Tr}_s\left[\gamma ^\mu G_0^\pm (K_1)\gamma ^\nu G_0^\pm (K_2)\right]=T\underset{k_0}{}_0^{1/T}𝑑\tau _1𝑑\tau _2e^{k_0\tau _1+(k_0p_0)\tau _2}\mathrm{Tr}_s\left[\gamma ^\mu G_0^\pm (\tau _1,𝐤_1)\gamma ^\nu G_0^\pm (\tau _2,𝐤_2)\right]$$
(50)
as follows. To perform the Matsubara sum over $`k_0`$, use the identity
$$T\underset{n}{}e^{k_0\tau }=\underset{m=\mathrm{}}{\overset{\mathrm{}}{}}(1)^m\delta \left(\tau \frac{m}{T}\right),$$
(51)
valid for fermionic Matsubara frequencies, $`k_0=i(2n+1)\pi T`$. Since $`0\tau _1,\tau _21/T`$ in Eq. (50), the delta function in Eq. (51) has support only for $`m=1`$, i.e., $`\tau _2=1/T\tau _1`$. With the help of Eqs. (45) and (46), as well as $`e^{p_0/T}=1`$ for bosonic Matsubara frequencies $`p_0=i2n\pi T`$, one obtains
$$T\underset{k_0}{}\mathrm{Tr}_s\left[\gamma ^\mu G_0^\pm (K_1)\gamma ^\nu G_0^\pm (K_2)\right]=_0^{1/T}𝑑\tau e^{p_0\tau }\mathrm{Tr}_s\left[\gamma ^\mu G_0^\pm (\tau ,𝐤_1)\gamma ^\nu \gamma _0G_0^{}(\tau ,𝐤_2)\gamma _0\right].$$
(52)
One now inserts the expressions (40), and integrates over $`\tau `$. Putting everything together, one obtains for the gluon self-energy:
$`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }(P)={\displaystyle \frac{1}{4}}g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}{\displaystyle \underset{e_1,e_2=\pm }{}}\{{\displaystyle \frac{}{}}𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)`$ (53)
$`\times `$ $`\left[({\displaystyle \frac{n_1^0(1n_2^0)}{p_0+ϵ_1^0+ϵ_2^0}}{\displaystyle \frac{(1n_1^0)n_2^0}{p_0ϵ_1^0ϵ_2^0}})(1N_1^0N_2^0)+({\displaystyle \frac{(1n_1^0)(1n_2^0)}{p_0ϵ_1^0+ϵ_2^0}}{\displaystyle \frac{n_1^0n_2^0}{p_0+ϵ_1^0ϵ_2^0}})(N_1^0N_2^0)\right]`$ (55)
$`+𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)`$
$`\times `$ $`[({\displaystyle \frac{(1n_1^0)n_2^0}{p_0+ϵ_1^0+ϵ_2^0}}{\displaystyle \frac{n_1^0(1n_2^0)}{p_0ϵ_1^0ϵ_2^0}})(1N_1^0N_2^0)+({\displaystyle \frac{n_1^0n_2^0}{p_0ϵ_1^0+ϵ_2^0}}{\displaystyle \frac{(1n_1^0)(1n_2^0)}{p_0+ϵ_1^0ϵ_2^0}})(N_1^0N_2^0)]\}.`$ (56)
Here,
$$𝒯_\pm ^{\mu \nu }(𝐤_1,𝐤_2)\mathrm{Tr}_s\left(\gamma _0\gamma ^\mu \mathrm{\Lambda }_{𝐤_1}^{\pm e_1}\gamma _0\gamma ^\nu \mathrm{\Lambda }_{𝐤_2}^{\pm e_2}\right),$$
(57)
and I introduced the somewhat compact notation
$$ϵ_i^0ϵ_{𝐤_i0}^{e_i},n_i^0n_{𝐤_i0}^{e_i},N_i^0N(ϵ_i^0).$$
(58)
An (appropriately generalized) expression of the form (56) will also appear in Sec. IV, when the self-energy is computed in the superconducting phase. In the normal phase, however, one can use $`n_i^0\theta (\mu e_ik_i)`$ to show that
$`n_i^0(1N_i^0)`$ $`=`$ $`n_i^0\left\{\theta (e_i)N_F^+(k_i)+\theta (e_i)\left[1N_F^{}(k_i)\right]\right\},`$ (60)
$`(1n_i^0)N_i^0`$ $`=`$ $`(1n_i^0)\left\{\theta (e_i)N_F^+(k_i)+\theta (e_i)\left[1N_F^{}(k_i)\right]\right\}`$ (61)
$`(1n_i^0)(1N_i^0)`$ $`=`$ $`(1n_i^0)\left\{\theta (e_i)\left[1N_F^+(k_i)\right]+\theta (e_i)N_F^{}(k_i)\right\},`$ (62)
$`n_i^0N_i^0`$ $`=`$ $`n_i^0\left\{\theta (e_i)\left[1N_F^+(k_i)\right]+\theta (e_i)N_F^{}(k_i)\right\}.`$ (63)
Equation (56) then simplifies to
$`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}\left[\frac{𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)}{p_0e_1k_1+e_2k_2}\frac{𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)}{p_0+e_1k_1e_2k_2}\right]}`$ (65)
$`\times \left\{{\displaystyle \frac{}{}}\theta (e_1)\left[1N_F^+(k_1)\right]+\theta (e_1)N_F^{}(k_1)\theta (e_2)\left[1N_F^+(k_2)\right]\theta (e_2)N_F^{}(k_2)\right\}.`$
### E Trace over spinor space
The traces (57) are best computed for temporal and spatial components separately,
$`𝒯_\pm ^{00}`$ $`=`$ $`1+e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2,`$ (67)
$`𝒯_\pm ^{0i}=𝒯_\pm ^{i0}`$ $`=`$ $`\pm e_1\widehat{k}_1^i\pm e_2\widehat{k}_2^i,i=x,y,z,`$ (68)
$`𝒯_\pm ^{ij}`$ $`=`$ $`\delta ^{ij}\left(1e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2\right)+e_1e_2\left(\widehat{k}_1^i\widehat{k}_2^j+\widehat{k}_1^j\widehat{k}_2^i\right),i,j=x,y,z.`$ (69)
Equation (56), or Eq. (65), together with Eqs. (III E), completes the computation of the quark contribution to the gluon self-energy to one-loop order in the normal phase. At temperatures $`T\mu `$, this is the dominant contribution to the gluon self-energy. In the following, I study the so-called hard-dense-loop (HDL) limit.
### F The HDL limit
To derive the HDL limit, it is advantageous to shift the integration over 3-momentum in Eqs. (56) or (65), $`𝐤𝐤+𝐩/2`$, such that $`𝐤_1=𝐤+𝐩/2`$ and $`𝐤_2=𝐤𝐩/2`$. The HDL limit is obtained by taking $`p_0,p`$ to be of order $`g\mu `$ (“soft”), while $`k`$ is of order $`\mu `$ (“hard”) . As the gluon self-energy (56) is already proportional to $`g^2`$, it is permissible to compute the integral in Eq. (56) to order $`O(p^0)`$. However, since some of the energy denominators are of order $`O(p)`$, one has to keep terms up to order $`O(p)`$ in the numerators, too. For the traces (III E) one then obtains
$`𝒯_\pm ^{00}`$ $``$ $`1+e_1e_2+O\left({\displaystyle \frac{p^2}{k^2}}\right),`$ (71)
$`𝒯_\pm ^{0i}=𝒯_\pm ^{i0}`$ $``$ $`\pm (e_1+e_2)\widehat{k}^i\pm (e_1e_2)\left(\delta ^{ij}\widehat{k}^i\widehat{k}^j\right){\displaystyle \frac{p^j}{2k}}+O\left({\displaystyle \frac{p^2}{k^2}}\right),`$ (72)
$`𝒯_\pm ^{ij}`$ $``$ $`\delta ^{ij}(1e_1e_2)+2e_1e_2\widehat{k}^i\widehat{k}^j+O\left({\displaystyle \frac{p^2}{k^2}}\right).`$ (73)
In the following, the temporal and spatial components of the gluon self-energy are evaluated separately.
(i) $`\mu =\nu =0`$: In the HDL limit, Eq. (71) shows that only particle-particle ($`e_1=e_2=+1)`$, or antiparticle-antiparticle ($`e_1=e_2=1`$) excitations contribute to the electric components of the gluon self-energy. In this case, only the difference $`k_1k_2`$ occurs in the energy denominators in Eq. (65), which, in the HDL limit, is
$$k_1k_2𝐩\widehat{𝐤}.$$
(74)
In the numerators, the difference of the thermal occupation numbers is
$$N_F^\pm (k_1)N_F^\pm (k_2)𝐩\widehat{𝐤}\frac{dN_F^\pm (k)}{dk}.$$
(75)
Equation (65) with Eq. (71) then yields
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(P)g^2N_f\frac{d^3𝐤}{(2\pi )^3}\left(1\frac{p_0}{p_0+𝐩\widehat{𝐤}}\right)\left[\frac{dN_F^+(k)}{dk}+\frac{dN_F^{}(k)}{dk}\right].$$
(76)
With some effort, one can also do the integration over $`k`$ exactly for all temperatures and chemical potentials . In this case, the final answer encompasses not only the hard-dense-loop limit, but also the “hard-thermal-loop” (HTL) limit. That much effort is, however, not necessary in the present case. For superconductivity, one is interested in temperatures of the order of the zero-temperature gap, $`T\varphi _0\mu \mathrm{exp}(c_{\mathrm{QCD}}/g)\mu `$. On this basis it was argued above that contributions from the gluon and ghost loops to the gluon self-energy can be neglected, as they are $`g^2T^2`$, while the dominant contribution from the quark loop is $`g^2\mu ^2`$.
In essence this means that effects from nonzero temperature can be neglected to leading order. Consequently,
$$\frac{dN_F^+(k)}{dk}\frac{d\theta (\mu k)}{dk}=\delta (k\mu ),\frac{dN_F^{}(k)}{dk}\frac{d\theta (k+\mu )}{dk}=0.$$
(77)
From the physical point of view this is an important relation: only quark excitations at the Fermi surface contribute to the gluon self-energy.
With these approximations one obtains the well-known result
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(P)3m_g^2\frac{d\mathrm{\Omega }}{4\pi }\left(1\frac{p_0}{p_0+𝐩\widehat{𝐤}}\right),$$
(78)
where $`d\mathrm{\Omega }`$ is the integration over solid angle and
$$m_g^2g^2\frac{N_f}{6\pi ^2}\mu ^2$$
(79)
is the gluon mass at $`T=0`$. Equation (78) remains valid in the HTL limit, when Eq. (79) is properly generalized to nonzero temperature .
In the static limit, $`p_0=0`$, the dependence on $`𝐩`$ vanishes, and one simply has
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(0)3m_g^2,$$
(80)
the usual result for Debye screening.
(ii) $`\mu =0,\nu =i`$: For $`\mathrm{\Pi }_{0}^{}{}_{}{}^{0i}`$, one concludes from Eqs. (65) and (72) that particle-antiparticle excitations are at least of order $`O(p^2)`$, i.e., to leading order in the HDL limit only particle-particle or antiparticle-antiparticle excitations contribute to the gluon self-energy. Then, with Eqs. (74) and (75) one obtains
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{0i}(P)g^2N_f\frac{d^3𝐤}{(2\pi )^3}\frac{p_0\widehat{k}^i}{p_0+𝐩\widehat{𝐤}}\left[\frac{dN_F^+(k)}{dk}+\frac{dN_F^{}(k)}{dk}\right].$$
(81)
For the temperatures of interest, one can again make the approximation (77) to obtain
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{0i}(P)3m_g^2\frac{d\mathrm{\Omega }}{4\pi }\frac{p_0\widehat{k}^i}{p_0+𝐩\widehat{𝐤}},$$
(82)
which coincides with .
In the static limit,
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{0i}(0)0.$$
(83)
(iii) $`\mu =i,\nu =j`$: For $`\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}`$, Eq. (73) shows that not only particle-particle ($`e_1=e_2=+1`$) and antiparticle-antiparticle ($`e_1=e_2=1`$) excitations contribute, as for the other components of $`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }`$, but also particle-antiparticle ($`e_1=e_2=\pm 1`$) excitations. In the former, one encounters again the difference of momenta (74) and thermal occupation numbers (75). In the latter, however, the sum of momenta and thermal occupation numbers occurs. To leading order in $`p`$,
$$k_1+k_22k,N_F^\pm (k_1)+N_F^\pm (k_2)2N_F^\pm (k).$$
(84)
Then,
$`\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(P)`$ $``$ $`g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}\{\widehat{k}^i\widehat{k}^j(1{\displaystyle \frac{p_0}{p_0+𝐩\widehat{𝐤}}})[{\displaystyle \frac{dN_F^+(k)}{dk}}+{\displaystyle \frac{dN_F^{}(k)}{dk}}]`$ (86)
$`(\delta ^{ij}\widehat{k}^i\widehat{k}^j){\displaystyle \frac{1}{k}}[1N_F^+(k)N_F^{}(k)]\}.`$
The 1 in the last term is an ultraviolet-divergent vacuum contribution and has to be removed by renormalization. The angular integration can be performed for the parts which do not depend on $`𝐩`$, $`(d\mathrm{\Omega }/4\pi )\widehat{k^i}\widehat{k^j}\delta ^{ij}/3`$. One then realizes after an integration by parts that the $`𝐩`$-independent part of the first line in Eq. (86) cancels the second line,
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(P)g^2N_f\frac{d^3𝐤}{(2\pi )^3}\widehat{k}^i\widehat{k}^j\frac{p_0}{p_0+𝐩\widehat{𝐤}}\left[\frac{dN_F^+(k)}{dk}+\frac{dN_F^{}(k)}{dk}\right].$$
(87)
With the gluon mass (79) this can be written in the form
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(P)3m_g^2\frac{d\mathrm{\Omega }}{4\pi }\widehat{k}^i\widehat{k}^j\frac{p_0}{p_0+𝐩\widehat{𝐤}},$$
(88)
which is the standard result . Static magnetic gluons are not screened,
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(0)0.$$
(89)
## IV The gluon self-energy in the superconducting phase
In the superconducting phase, the ground state is a condensate of quark Cooper pairs, $`\overline{\psi }_C\psi 0`$. As was shown in , in mean-field approximation the quark propagator (26) becomes
$$𝒮(K)=\left(\begin{array}{cc}G^+(K)& \mathrm{\Xi }^{}(K)\\ \mathrm{\Xi }^+(K)& G^{}(K)\end{array}\right),$$
(90)
where the quasiparticle and charge conjugate quasiparticle propagators are
$$G^\pm \left(G_0^\pm \mathrm{\Sigma }^\pm \right)^1,\mathrm{\Sigma }^\pm \mathrm{\Phi }^{}G_0^{}\mathrm{\Phi }^\pm .$$
(91)
$`\mathrm{\Sigma }^\pm `$ is the quark self-energy generated by exchanging particles or charge conjugate particles with the condensate. For $`\mathrm{\Sigma }^+`$, a particle annihilates with an antiparticle in the condensate $`\mathrm{\Phi }^+\psi _C\overline{\psi }`$, and a charge conjugate particle is created. This charge conjugate particle propagates via $`G_0^{}`$, until it annihilates in the condensate $`\mathrm{\Phi }^{}\psi \overline{\psi }_C`$ with a charge conjugate antiparticle, whereby a particle is created . The meaning of $`\mathrm{\Sigma }^{}`$ can be explained analogously, except that the roles of particles and charge conjugate particles are interchanged.
The off-diagonal components of the quark propagator (90) are
$$\mathrm{\Xi }^\pm G_0^{}\mathrm{\Phi }^\pm G^\pm .$$
(92)
The physical interpretation is that particles (charge conjugate particles) annihilate with an antiparticle (a charge conjugate antiparticle) in the condensate, upon which a charge conjugate particle (a particle) is created.
In mean-field approximation, the condensate $`\mathrm{\Phi }^+`$ obeys the gap equation
$$\mathrm{\Phi }^+(K)g^2\frac{T}{V}\underset{Q}{}\mathrm{\Delta }_{\mu \nu }^{ab}(KQ)\overline{\mathrm{\Gamma }}_a^\mu \mathrm{\Xi }^+(Q)\mathrm{\Gamma }_b^\nu ,$$
(93)
and $`\mathrm{\Phi }^{}`$ can be obtained from
$$\mathrm{\Phi }^{}(K)\gamma _0\left[\mathrm{\Phi }^+(K)\right]^{}\gamma _0.$$
(94)
The solution of the gap equation (93) has been extensively discussed in .
The gluon self-energy (27) becomes
$$\mathrm{\Pi }_{ab}^{\mu \nu }(P)=\frac{1}{2}g^2\frac{T}{V}\underset{K}{}\mathrm{Tr}_{s,c,f,NG}\left[\widehat{\mathrm{\Gamma }}_a^\mu 𝒮(K)\widehat{\mathrm{\Gamma }}_b^\nu 𝒮(KP)\right].$$
(95)
As in the normal phase, this expression is computed in several steps.
### A Trace over Nambu–Gor’kov space
The trace over the 2-dimensional Nambu–Gor’kov space is readily performed with Eqs. (14) and (90),
$`\mathrm{\Pi }_{ab}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{2}}g^2{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_{s,c,f}[\mathrm{\Gamma }_a^\mu G^+(K)\mathrm{\Gamma }_b^\nu G^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu G^{}(K)\overline{\mathrm{\Gamma }}_b^\nu G^{}(KP)`$ (97)
$`+\mathrm{\Gamma }_a^\mu \mathrm{\Xi }^{}(K)\overline{\mathrm{\Gamma }}_b^\nu \mathrm{\Xi }^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu \mathrm{\Xi }^+(K)\mathrm{\Gamma }_b^\nu \mathrm{\Xi }^{}(KP)].`$
When the temperature approaches the critical temperature, $`TT_c`$, the condensate melts, $`\mathrm{\Phi }^\pm 0`$, i.e., $`\mathrm{\Xi }^\pm 0`$ and $`G^\pm G_0^\pm `$, and the gluon self-energy assumes the form of the normal phase, $`\mathrm{\Pi }_{ab}^{\mu \nu }\mathrm{\Pi }_{0}^{}{}_{ab}{}^{\mu \nu }`$, which was discussed in the previous Sec. III.
### B Trace over flavor space
For a condensate with total spin $`J=0`$ and $`N_f=2`$, the condensate is totally antisymmetric in flavor space ,
$$\left(\mathrm{\Phi }^\pm \right)_{fg}\pm ϵ_{fg}\mathrm{\Phi }^\pm ,$$
(98)
where use has been made of $`ϵ_{fg}^T=ϵ_{gf}=ϵ_{fg}`$. Consequently, since the free quark propagator is diagonal in flavor space, the quark self-energy is also diagonal in flavor space,
$$\left(\mathrm{\Sigma }^\pm \right)_{fg}=\left(\mathrm{\Phi }^{}\right)_{fh}\left(G_0^{}\right)_{hm}\left(\mathrm{\Phi }^\pm \right)_{mg}=ϵ_{hf}ϵ_{hg}\mathrm{\Phi }^{}G_0^{}\mathrm{\Phi }^\pm =\delta _{fg}\mathrm{\Sigma }^\pm .$$
(99)
Then, also the quasiparticle propagator is diagonal in flavor space,
$$\left(G^\pm \right)_{fg}=\delta _{fg}G^\pm .$$
(100)
On the other hand, the off-diagonal components of $`𝒮`$ are antisymmetric in flavor space,
$$\left(\mathrm{\Xi }^\pm \right)_{fg}=\left(G_0^{}\right)_{fh}\left(\mathrm{\Phi }^\pm \right)_{hm}\left(G^\pm \right)_{mg}=\pm ϵ_{fg}\mathrm{\Xi }^\pm .$$
(101)
As the vertices $`\mathrm{\Gamma }_a^\mu `$ and $`\overline{\mathrm{\Gamma }}_a^\mu `$ are flavor-blind (proportional to the unit matrix in flavor space), the trace over flavor space in Eq. (97) results in
$`\mathrm{\Pi }_{ab}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_{s,c}[\mathrm{\Gamma }_a^\mu G^+(K)\mathrm{\Gamma }_b^\nu G^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu G^{}(K)\overline{\mathrm{\Gamma }}_b^\nu G^{}(KP)`$ (103)
$`+\mathrm{\Gamma }_a^\mu \mathrm{\Xi }^{}(K)\overline{\mathrm{\Gamma }}_b^\nu \mathrm{\Xi }^+(KP)+\overline{\mathrm{\Gamma }}_a^\mu \mathrm{\Xi }^+(K)\mathrm{\Gamma }_b^\nu \mathrm{\Xi }^{}(KP)],`$
where, of course, $`N_f=2`$.
### C Trace over color space
A $`N_f=2`$, $`J=0`$ condensate is also totally antisymmetric in color space ,
$$\left(\mathrm{\Phi }^\pm \right)_{ij}\pm ϵ_{ij3}\mathrm{\Phi }^\pm ,$$
(104)
where use has been made of $`ϵ_{ij3}^T=ϵ_{ji3}=ϵ_{ij3}`$, and where a global color rotation has been performed to orient the condensate into the (anti-)3–direction in color space. (The notation is again sloppy: the “3” is actually not a triplet, but an anti-triplet index.)
The free quark propagator is diagonal in color space, so that one computes for the quark self-energy:
$$\left(\mathrm{\Sigma }^\pm \right)_{ij}=\left(\mathrm{\Phi }^{}\right)_{ik}\left(G_0^{}\right)_{kl}\left(\mathrm{\Phi }^\pm \right)_{lj}=ϵ_{ki3}ϵ_{kj3}\mathrm{\Phi }^{}G_0^{}\mathrm{\Phi }^\pm =\left(\delta _{ij}\delta _{i3}\delta _{j3}\right)\mathrm{\Sigma }^\pm .$$
(105)
This result is physically easy to interpret, remembering the above discussion of how the quark self-energy arises. Quarks with color 3 do not condense, consequently there is no antiquark in the condensate which a color–3 quark could annihilate with. Thus, color–3 quarks do not attain a self-energy .
The color structure of the quasiparticle propagator is therefore
$$\left(G^\pm \right)_{ij}=\left(\delta _{ij}\delta _{i3}\delta _{j3}\right)G^\pm +\delta _{i3}\delta _{j3}G_0^\pm .$$
(106)
For the off-diagonal components of $`𝒮`$ one then finds
$$\left(\mathrm{\Xi }^\pm \right)_{ij}=\left(G_0^{}\right)_{ik}\left(\mathrm{\Phi }^\pm \right)_{kl}\left(G^\pm \right)_{lj}=\pm ϵ_{ij3}\mathrm{\Xi }^\pm .$$
(107)
One now computes the trace over color space with the explicit form of the Gell-Mann matrices. After a somewhat tedious, but straightforward calculation one obtains for $`a=b=1,2,3`$:
$`\mathrm{\Pi }_{11}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s[\gamma ^\mu G^+(K)\gamma ^\nu G^+(KP)+\gamma ^\mu G^{}(K)\gamma ^\nu G^{}(KP)`$ (110)
$`+\gamma ^\mu \mathrm{\Xi }^{}(K)\gamma ^\nu \mathrm{\Xi }^+(KP)+\gamma ^\mu \mathrm{\Xi }^+(K)\gamma ^\nu \mathrm{\Xi }^{}(KP)],`$
for $`a=b=4,5,6,7`$:
$`\mathrm{\Pi }_{44}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{8}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s[\gamma ^\mu G_0^+(K)\gamma ^\nu G^+(KP)+\gamma ^\mu G^+(K)\gamma ^\nu G_0^+(KP)`$ (112)
$`+\gamma ^\mu G_0^{}(K)\gamma ^\nu G^{}(KP)+\gamma ^\mu G^{}(K)\gamma ^\nu G_0^{}(KP)],`$
and for $`a=b=8`$:
$`\mathrm{\Pi }_{88}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{2}{3}}\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }(P)+{\displaystyle \frac{1}{3}}\stackrel{~}{\mathrm{\Pi }}^{\mu \nu }(P),`$ (113)
$`\stackrel{~}{\mathrm{\Pi }}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s[\gamma ^\mu G^+(K)\gamma ^\nu G^+(KP)+\gamma ^\mu G^{}(K)\gamma ^\nu G^{}(KP)`$ (115)
$`\gamma ^\mu \mathrm{\Xi }^{}(K)\gamma ^\nu \mathrm{\Xi }^+(KP)\gamma ^\mu \mathrm{\Xi }^+(K)\gamma ^\nu \mathrm{\Xi }^{}(KP)],`$
where $`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }`$ is the gluon self-energy in the normal phase, Eq. (36).
Apart from the diagonal elements (110) – (115), after performing the color-trace one also finds the off-diagonal elements
$`\mathrm{\Pi }_{45}^{\mu \nu }(P)`$ $`=`$ $`\mathrm{\Pi }_{54}^{\mu \nu }(P)=\mathrm{\Pi }_{67}^{\mu \nu }(P)=\mathrm{\Pi }_{76}^{\mu \nu }(P)i\widehat{\mathrm{\Pi }}^{\mu \nu }(P),`$ (116)
$`\widehat{\mathrm{\Pi }}^{\mu \nu }(P)`$ $``$ $`{\displaystyle \frac{1}{8}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s[\gamma ^\mu G_0^+(K)\gamma ^\nu G^+(KP)\gamma ^\mu G^+(K)\gamma ^\nu G_0^+(KP)`$ (118)
$`\gamma ^\mu G_0^{}(K)\gamma ^\nu G^{}(KP)+\gamma ^\mu G^{}(K)\gamma ^\nu G_0^{}(KP)].`$
The occurrence of these off-diagonal elements bears no special physical meaning. It simply indicates that the inverse gluon propagator $`\mathrm{\Delta }^1`$ is not diagonal in the original basis of adjoint colors. For instance, in the $`(45)`$-subspace of adjoint colors $`\mathrm{\Delta }^1`$ has the form
$$\left(\begin{array}{cc}\mathrm{\Delta }_0^1+\mathrm{\Pi }_{44}& i\widehat{\mathrm{\Pi }}\\ i\widehat{\mathrm{\Pi }}& \mathrm{\Delta }_0^1+\mathrm{\Pi }_{44}\end{array}\right).$$
(119)
This hermitean matrix is easily diagonalized by the unitary matrix
$$U\frac{1}{\sqrt{2}}\left(\begin{array}{cc}1& i\\ i& 1\end{array}\right).$$
(120)
In the new (diagonal) basis of adjoint colors,
$$\left(\begin{array}{cc}\mathrm{\Delta }_0^1+\mathrm{\Pi }_{44}+\widehat{\mathrm{\Pi }}& 0\\ 0& \mathrm{\Delta }_0^1+\mathrm{\Pi }_{44}\widehat{\mathrm{\Pi }}\end{array}\right).$$
(121)
Similar arguments hold for the $`(67)`$-subspace. Therefore, rotating into the new (diagonal) basis,
$$\mathrm{\Pi }_{44}+\widehat{\mathrm{\Pi }}=\mathrm{\Pi }_{66}+\widehat{\mathrm{\Pi }}\mathrm{\Pi }_{44}=\mathrm{\Pi }_{66},\mathrm{\Pi }_{44}\widehat{\mathrm{\Pi }}=\mathrm{\Pi }_{66}\widehat{\mathrm{\Pi }}\mathrm{\Pi }_{55}=\mathrm{\Pi }_{77}.$$
(122)
In the following, only these diagonal gluon self energies will be considered. They read explicitly
$`\mathrm{\Pi }_{44}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s\left[\gamma ^\mu G_0^+(K)\gamma ^\nu G^+(KP)+\gamma ^\mu G^{}(K)\gamma ^\nu G_0^{}(KP)\right],`$ (124)
$`\mathrm{\Pi }_{55}^{\mu \nu }(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{T}{V}}{\displaystyle \underset{K}{}}\mathrm{Tr}_s\left[\gamma ^\mu G^+(K)\gamma ^\nu G_0^+(KP)+\gamma ^\mu G_0^{}(K)\gamma ^\nu G^{}(KP)\right].`$ (125)
Remembering the explicit form of the Gell-Mann matrices, the results (110), (115), and (122) are simple to interpret. Gluons of adjoint colors 1, 2, and 3 see only quarks in the condensate, with fundamental colors 1 and 2. Their self-energy has therefore contributions from the diagonal ($`G^\pm `$), as well as the off-diagonal ($`\mathrm{\Xi }^\pm `$) components of the quark propagator (90).
Gluons of colors 4 and 5 “see” the uncondensed quark with fundamental color 3, but also the condensed quarks of color 1. Analogously, gluons of colors 6 and 7 see the uncondensed quark and the condensed quark of color 2. Therefore, the fermion loop in the self-energy contains one free propagator $`G_0^\pm `$, corresponding to the uncondensed quark, and one quasiparticle (charge conjugate quasiparticle) propagator $`G^\pm `$, corresponding to the quark in the condensate. As there is no way to annihilate a color-3 quark in the condensate, there is no contribution from the off-diagonal components of Eq. (90).
Finally, gluons of color 8 see the condensed quarks of colors 1 and 2, but also the uncondensed color-3 quark. The contribution to the gluon self-energy from the latter is equal to that in the normal phase, $`\mathrm{\Pi }_{0}^{}{}_{}{}^{\mu \nu }`$, the factor $`2/3`$ comes from the $`(33)`$-element of $`T^8`$. Apart from the prefactor $`1/3`$, the contribution from the condensed quarks, $`\stackrel{~}{\mathrm{\Pi }}^{\mu \nu }`$, looks similar to $`\mathrm{\Pi }_{11}^{\mu \nu }`$, except that the sign of the last two terms is different. As will be seen below, this difference is important to keep gluons of colors 1, 2, and 3 massless, while the eighth gluon becomes massive. Note that, for QED, $`\overline{\mathrm{\Gamma }}_a^\mu \overline{\mathrm{\Gamma }}^\mu =\gamma ^\mu `$, $`ge`$. Thus, for $`N_f=2`$, the contribution from the condensed quarks to the self-energy of gluons of color 8, $`\stackrel{~}{\mathrm{\Pi }}^{\mu \nu }`$, is exactly $`g^2/e^2`$ of what one expects for the photon self-energy in an ordinary superconductor.
### D Mixed representations for the quark propagators
For $`m=0`$, the quasiparticle propagator can be written in terms of chirality and energy projectors ,
$$G^\pm (K)=\underset{h=r,\mathrm{}}{}\underset{e=\pm }{}\frac{𝒫_{\pm h}\mathrm{\Lambda }_𝐤^{\pm e}}{k_0^2[ϵ_𝐤^e(\varphi _h^e)]^2}\left[G_0^{}(K)\right]^1,$$
(126)
where $`𝒫_{r,\mathrm{}}=(1\pm \gamma _5)/2`$ are chirality projectors (the notation $`h`$ stands for $`\mathrm{}`$, if $`h=r`$, and $`r`$, if $`h=\mathrm{}`$). The quasiparticle energies are
$$ϵ_𝐤^e(\varphi _h^e)\sqrt{(\mu ek)^2+|\varphi _h^e|^2},$$
(127)
where $`\varphi _h^e`$ is the gap function for pairing of quarks ($`e=+1`$) or antiquarks ($`e=1`$) with chirality $`h`$.
An analysis of the gap functions in mean-field approximation shows that left- and right-handed gap functions differ only by a complex phase factor,
$$\varphi _r^e=\varphi ^e\mathrm{exp}(i\theta ^e),\varphi _{\mathrm{}}^e=\varphi ^e\mathrm{exp}(i\theta ^e),$$
(128)
with $`\varphi ^e𝐑`$. Moreover, the phase factor is independent of the energy projection, $`\theta ^+=\theta ^{}\theta `$. Condensation fixes the value of $`\theta `$, and breaks the $`U_A(1)`$ symmetry (which is effectively restored at high densities) spontaneously. If $`\theta =0`$ or $`\pi /2`$, condensation occurs in a spin-zero channel with good parity, $`J^P=0^+`$ or $`J^P=0^{}`$, respectively. For $`\theta 0`$, there is always a $`J^P=0^{}`$ admixture, thus condensation breaks also parity . For the sake of simplicity, in the following we only consider $`\theta =0`$.
From Eq. (128), $`|\varphi _r^e||\varphi _{\mathrm{}}^e|\varphi ^e`$, and the sum over chiralities in Eq. (126) is superfluous. Writing the inverse free propagator as $`\left[G_0^{}(K)\right]^1=\left[k_0(\mu ek)2ek\mathrm{\Lambda }_𝐤^e\right]\gamma _0`$, Eq. (126) can be brought in the form
$$G^\pm (K)=\underset{e=\pm }{}\frac{k_0(\mu ek)}{k_0^2[ϵ_𝐤^e]^2}\mathrm{\Lambda }_𝐤^{\pm e}\gamma _0,$$
(129)
which should be compared with Eq. (37). Obviously, all that has changed is that the free quark excitation energies (38) have been replaced by the quasiparticle excitation energies (127), $`ϵ_{𝐤0}^eϵ_𝐤^eϵ_𝐤^e(\varphi ^e)`$.
After realizing this, by comparison with Eqs. (40) one can immediately write down the mixed representation for the quasiparticle propagators,
$`G^+(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{e=\pm }{}}\mathrm{\Lambda }_𝐤^e\gamma _0\left\{{\displaystyle \frac{}{}}(1n_𝐤^e)\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )n_𝐤^e\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )\right\},`$ (131)
$`G^{}(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{e=\pm }{}}\gamma _0\mathrm{\Lambda }_𝐤^e\left\{{\displaystyle \frac{}{}}n_𝐤^e\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )(1n_𝐤^e)\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )\right\}.`$ (132)
Here,
$$n_𝐤^e\frac{ϵ_𝐤^e+\mu ek}{2ϵ_𝐤^e}$$
(133)
are the occupation numbers for quasiparticles ($`e=+1`$) or quasi-antiparticles ($`e=1`$) at zero temperature . Consequently, $`1n_𝐤^e`$ are the occupation numbers for quasiparticle holes or quasi-antiparticle holes. Due to the presence of a gap $`\varphi ^e`$ in the quasiparticle excitation spectrum, these occupation numbers are no longer simple theta functions in momentum space, as in the noninteracting case; the theta functions become “smeared” over a range $`\varphi ^e`$ around the Fermi surface (cf. Fig. 2 in ). The relations (45) and (46) are also fulfilled by $`G^\pm (\tau ,𝐤)`$.
From a comparison of Eqs. (IV D) and (40), one can immediately deduce from Eq. (56) the result for the traces $`\mathrm{Tr}_s\left[\gamma ^\mu G^\pm (K)\gamma ^\nu G^\pm (KP)\right]`$, $`\mathrm{Tr}_s\left[\gamma ^\mu G_0^\pm (K)\gamma ^\nu G^\pm (KP)\right]`$, or $`\mathrm{Tr}_s\left[\gamma ^\mu G^\pm (K)\gamma ^\nu G_0^\pm (KP)\right]`$. All one has to do is replace
$$ϵ_i^0ϵ_iϵ_{𝐤_i}^{e_i},n_i^0n_in_{𝐤_i}^{e_i},N_i^0N_iN(ϵ_i),$$
(134)
when a propagator $`G^\pm `$ occurs in place of $`G_0^\pm `$.
One also needs a mixed representation for the off-diagonal components of $`𝒮(K)`$. First, write $`\mathrm{\Xi }^\pm (K)`$, Eq. (92), in terms of projectors,
$$\mathrm{\Xi }^+(K)=\underset{h=r,\mathrm{}}{}\underset{e=\pm }{}\frac{\varphi _h^e(K)}{k_0^2[ϵ_𝐤^e]^2}𝒫_h\mathrm{\Lambda }_𝐤^e,\mathrm{\Xi }^{}(K)=\underset{h=r,\mathrm{}}{}\underset{e=\pm }{}\frac{\left[\varphi _h^e(K)\right]^{}}{k_0^2[ϵ_𝐤^e]^2}𝒫_h\mathrm{\Lambda }_𝐤^e.$$
(135)
As in , assume that $`\varphi _h^e(k_0)`$ has no poles or cuts in the complex $`k_0`$-plane and that $`\varphi _h^e(k_0)=\varphi _h^e(k_0)`$. In this case, one obtains the mixed representations
$`\mathrm{\Xi }^+(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{h=r,\mathrm{}}{}}{\displaystyle \underset{e=\pm }{}}𝒫_h\mathrm{\Lambda }_𝐤^e{\displaystyle \frac{\varphi _h^e(ϵ_𝐤^e,𝐤)}{2ϵ_𝐤^e}}\left\{{\displaystyle \frac{}{}}\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )+\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )\right\},`$ (137)
$`\mathrm{\Xi }^{}(\tau ,𝐤)`$ $`=`$ $`{\displaystyle \underset{h=r,\mathrm{}}{}}{\displaystyle \underset{e=\pm }{}}𝒫_h\mathrm{\Lambda }_𝐤^e{\displaystyle \frac{\left[\varphi _h^e(ϵ_𝐤^e,𝐤)\right]^{}}{2ϵ_𝐤^e}}\left\{{\displaystyle \frac{}{}}\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )+\left[\theta (\tau )N(ϵ_𝐤^e)\right]\mathrm{exp}(ϵ_𝐤^e\tau )\right\}.`$ (138)
Note that the energy in the gap functions $`\varphi _h^e`$ is on the quasiparticle mass shell, $`k_0\pm ϵ_𝐤^e`$.
The traces in Eqs. (107) involving $`\mathrm{\Xi }^\pm `$ are now straightforwardly computed as
$`T{\displaystyle \underset{k_0}{}}\mathrm{Tr}_s\left[\gamma ^\mu \mathrm{\Xi }^{}(K_1)\gamma ^\nu \mathrm{\Xi }^\pm (K_2)\right]={\displaystyle \underset{e_1,e_2=\pm }{}}𝒰_\pm ^{\mu \nu }(𝐤_1,𝐤_2){\displaystyle \frac{\varphi _1\varphi _2}{4ϵ_1ϵ_2}}`$ (139)
$`\times `$ $`\left[\left({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}}\right)\left(1N_1N_2\right)\left({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}}\right)\left(N_1N_2\right)\right],`$ (140)
where $`K_1K`$, $`K_2KP`$, as before, while
$$\varphi _i\varphi ^{e_i}(ϵ_i,𝐤_i),$$
(141)
and
$$𝒰_\pm ^{\mu \nu }(𝐤_1,𝐤_2)\mathrm{Tr}_s\left[\gamma ^\mu \mathrm{\Lambda }_{𝐤_1}^{\pm e_1}\gamma ^\nu \mathrm{\Lambda }_{𝐤_2}^{e_2}\right].$$
(142)
On account of $`𝒫_h\gamma ^\mu =\gamma ^\mu 𝒫_h`$ and $`𝒫_r𝒫_{\mathrm{}}=0`$, the sum over chiralities $`h_1`$ and $`h_2`$ originating from the mixed representations (135) could be performed trivially.
Putting everything together, the self-energy for gluons of color 1, 2, and 3 is
$`\mathrm{\Pi }_{11}^{\mu \nu }(P)={\displaystyle \frac{1}{4}}g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}{\displaystyle \underset{e_1,e_2=\pm }{}}\{{\displaystyle \frac{}{}}𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)`$ (144)
$`\times `$ $`\left[\left({\displaystyle \frac{n_1(1n_2)}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{(1n_1)n_2}{p_0ϵ_1ϵ_2}}\right)(1N_1N_2)+\left({\displaystyle \frac{(1n_1)(1n_2)}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{n_1n_2}{p_0+ϵ_1ϵ_2}}\right)(N_1N_2)\right]`$ (146)
$`+𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)`$
$`\times `$ $`\left[\left({\displaystyle \frac{(1n_1)n_2}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{n_1(1n_2)}{p_0ϵ_1ϵ_2}}\right)(1N_1N_2)+\left({\displaystyle \frac{n_1n_2}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{(1n_1)(1n_2)}{p_0+ϵ_1ϵ_2}}\right)(N_1N_2)\right]`$ (148)
$`\left[𝒰_+^{\mu \nu }(𝐤_1,𝐤_2)+𝒰_{}^{\mu \nu }(𝐤_1,𝐤_2)\right]{\displaystyle \frac{\varphi _1\varphi _2}{4ϵ_1ϵ_2}}`$
$`\times `$ $`[({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2)({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2)]\},`$ (149)
for gluon colors 4 and 6,
$`\mathrm{\Pi }_{44}^{\mu \nu }(P)={\displaystyle \frac{1}{4}}g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}{\displaystyle \underset{e_1,e_2=\pm }{}}\{{\displaystyle \frac{}{}}𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)`$ (150)
$`\times `$ $`\left[\left({\displaystyle \frac{n_1^0(1n_2)}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)n_2}{p_0ϵ_1^0ϵ_2}}\right)(1N_1^0N_2)+\left({\displaystyle \frac{(1n_1^0)(1n_2)}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0n_2}{p_0+ϵ_1^0ϵ_2}}\right)(N_1^0N_2)\right]`$ (152)
$`+𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)`$
$`\times `$ $`[({\displaystyle \frac{(1n_1)n_2^0}{p_0+ϵ_1+ϵ_2^0}}{\displaystyle \frac{n_1(1n_2^0)}{p_0ϵ_1ϵ_2^0}})(1N_1N_2^0)+({\displaystyle \frac{n_1n_2^0}{p_0ϵ_1+ϵ_2^0}}{\displaystyle \frac{(1n_1)(1n_2^0)}{p_0+ϵ_1ϵ_2^0}})(N_1N_2^0)]\},`$ (153)
for gluon colors 5 and 7,
$`\mathrm{\Pi }_{55}^{\mu \nu }(P)={\displaystyle \frac{1}{4}}g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}{\displaystyle \underset{e_1,e_2=\pm }{}}\{{\displaystyle \frac{}{}}𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)`$ (154)
$`\times `$ $`\left[\left({\displaystyle \frac{n_1(1n_2^0)}{p_0+ϵ_1+ϵ_2^0}}{\displaystyle \frac{(1n_1)n_2^0}{p_0ϵ_1ϵ_2^0}}\right)(1N_1N_2^0)+\left({\displaystyle \frac{(1n_1)(1n_2^0)}{p_0ϵ_1+ϵ_2^0}}{\displaystyle \frac{n_1n_2^0}{p_0+ϵ_1ϵ_2^0}}\right)(N_1N_2^0)\right]`$ (156)
$`+𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)`$
$`\times `$ $`[({\displaystyle \frac{(1n_1^0)n_2}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0(1n_2)}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{n_1^0n_2}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)(1n_2)}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)]\},`$ (157)
and for gluon color 8
$`\stackrel{~}{\mathrm{\Pi }}^{\mu \nu }(P)={\displaystyle \frac{1}{4}}g^2N_f{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}{\displaystyle \underset{e_1,e_2=\pm }{}}\{{\displaystyle \frac{}{}}𝒯_+^{\mu \nu }(𝐤_1,𝐤_2)`$ (158)
$`\times `$ $`\left[\left({\displaystyle \frac{n_1(1n_2)}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{(1n_1)n_2}{p_0ϵ_1ϵ_2}}\right)(1N_1N_2)+\left({\displaystyle \frac{(1n_1)(1n_2)}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{n_1n_2}{p_0+ϵ_1ϵ_2}}\right)(N_1N_2)\right]`$ (160)
$`+𝒯_{}^{\mu \nu }(𝐤_1,𝐤_2)`$
$`\times `$ $`\left[\left({\displaystyle \frac{(1n_1)n_2}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{n_1(1n_2)}{p_0ϵ_1ϵ_2}}\right)(1N_1N_2)+\left({\displaystyle \frac{n_1n_2}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{(1n_1)(1n_2)}{p_0+ϵ_1ϵ_2}}\right)(N_1N_2)\right]`$ (162)
$`+\left[𝒰_+^{\mu \nu }(𝐤_1,𝐤_2)+𝒰_{}^{\mu \nu }(𝐤_1,𝐤_2)\right]{\displaystyle \frac{\varphi _1\varphi _2}{4ϵ_1ϵ_2}}`$
$`\times `$ $`[({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2)({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2)]\}.`$ (163)
### E Trace over spinor space
The traces $`𝒯_\pm ^{\mu \nu }`$ have been computed in Sec. III E. What remains to be done is to compute $`𝒰_\pm ^{\mu \nu }`$. One finds
$`𝒰_\pm ^{00}`$ $`=`$ $`𝒯_\pm ^{00},`$ (165)
$`𝒰_\pm ^{0i}=𝒰_\pm ^{i0}`$ $`=`$ $`𝒯_\pm ^{0i},i=x,y,x,`$ (166)
$`𝒰_\pm ^{ij}`$ $`=`$ $`𝒯_\pm ^{ij},i,j=x,y,z.`$ (167)
In the following, the results for the different components of the gluon self-energy in the superconducting phase will be collected.
### F Gluons of color 1, 2, and 3
(i) $`\mu =\nu =0`$: Defining
$$\xi _ie_ik_i\mu ,$$
(168)
the self-energy of electric gluons of color 1, 2, and 3 is determined from Eqs. (III E), (149), and (165) as
$`\mathrm{\Pi }_{11}^{00}(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(1+e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)}`$ (172)
$`\times [({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2){\displaystyle \frac{ϵ_1ϵ_2\xi _1\xi _2\varphi _1\varphi _2}{2ϵ_1ϵ_2}}`$
$`+({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2){\displaystyle \frac{ϵ_1ϵ_2+\xi _1\xi _2+\varphi _1\varphi _2}{2ϵ_1ϵ_2}}].`$
(ii) $`\mu =0,\nu =i`$: For the $`(0i)`$-components of the self-energy of gluons with colors 1, 2, or 3 one obtains
$`\mathrm{\Pi }_{11}^{0i}(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(e_1\widehat{k}_1^i+e_2\widehat{k}_2^i)}`$ (175)
$`\times [({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}+{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2)({\displaystyle \frac{\xi _2}{2ϵ_2}}{\displaystyle \frac{\xi _1}{2ϵ_1}})`$
$`+({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}+{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2)({\displaystyle \frac{\xi _1}{2ϵ_1}}+{\displaystyle \frac{\xi _2}{2ϵ_2}})].`$
(iii) $`\mu =i,\nu =j`$: The self-energy of magnetic gluons of colors 1, 2, and 3 is
$`\mathrm{\Pi }_{11}^{ij}(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}\left[\delta ^{ij}\left(1e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2\right)+e_1e_2\left(\widehat{k}_1^i\widehat{k}_2^j+\widehat{k}_1^j\widehat{k}_2^i\right)\right]}`$ (178)
$`\times [({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2){\displaystyle \frac{ϵ_1ϵ_2\xi _1\xi _2+\varphi _1\varphi _2}{2ϵ_1ϵ_2}}`$
$`+({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2){\displaystyle \frac{ϵ_1ϵ_2+\xi _1\xi _2\varphi _1\varphi _2}{2ϵ_1ϵ_2}}].`$
### G Gluons of color 4 and 6
(i) $`\mu =\nu =0`$: Using the symmetry of Eq. (153) under $`𝐤_1𝐤_2`$, $`e_1e_2`$, the self-energy of electric gluons of colors 4 and 6 can be written as
$`\mathrm{\Pi }_{44}^{00}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(1+e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)}`$ (180)
$`\times `$ $`\left[({\displaystyle \frac{n_1^0(1n_2)}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)n_2}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{(1n_1^0)(1n_2)}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0n_2}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (181)
(ii) $`\mu =0,\nu =i`$: The same symmetry arguments lead to
$`\mathrm{\Pi }_{44}^{0i}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(e_1\widehat{k}_1^i+e_2\widehat{k}_2^i)}`$ (182)
$`\times `$ $`\left[({\displaystyle \frac{n_1^0(1n_2)}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)n_2}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{(1n_1^0)(1n_2)}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0n_2}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (183)
(iii) $`\mu =i,\nu =j`$: For the self-energy of magnetic gluons of color 4 and 6 one obtains
$`\mathrm{\Pi }_{44}^{ij}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}\left[\delta ^{ij}(1e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)+e_1e_2\left(\widehat{k}_1^i\widehat{k}_2^j+\widehat{k}_1^j\widehat{k}_2^i\right)\right]}`$ (184)
$`\times `$ $`\left[({\displaystyle \frac{n_1^0(1n_2)}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)n_2}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{(1n_1^0)(1n_2)}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0n_2}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (185)
### H Gluons of color 5 and 7
(i) $`\mu =\nu =0`$: Again using the symmetry of Eq. (157) under $`𝐤_1𝐤_2`$, $`e_1e_2`$, the self-energy of electric gluons of colors 5 and 7 can be written as
$`\mathrm{\Pi }_{55}^{00}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(1+e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)}`$ (187)
$`\times `$ $`\left[({\displaystyle \frac{(1n_1^0)n_2}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0(1n_2)}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{n_1^0n_2}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)(1n_2)}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (188)
(ii) $`\mu =0,\nu =i`$: The $`(0i)`$-components are
$`\mathrm{\Pi }_{55}^{0i}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(e_1\widehat{k}_1^i+e_2\widehat{k}_2^i)}`$ (189)
$`\times `$ $`\left[({\displaystyle \frac{(1n_1^0)n_2}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0(1n_2)}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{n_1^0n_2}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)(1n_2)}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (190)
(iii) $`\mu =i,\nu =j`$: For the self-energy of magnetic gluons of color 5 and 7 one gets
$`\mathrm{\Pi }_{55}^{ij}(P)={\displaystyle \frac{1}{2}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}\left[\delta ^{ij}(1e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)+e_1e_2\left(\widehat{k}_1^i\widehat{k}_2^j+\widehat{k}_1^j\widehat{k}_2^i\right)\right]}`$ (191)
$`\times `$ $`\left[({\displaystyle \frac{(1n_1^0)n_2}{p_0+ϵ_1^0+ϵ_2}}{\displaystyle \frac{n_1^0(1n_2)}{p_0ϵ_1^0ϵ_2}})(1N_1^0N_2)+({\displaystyle \frac{n_1^0n_2}{p_0ϵ_1^0+ϵ_2}}{\displaystyle \frac{(1n_1^0)(1n_2)}{p_0+ϵ_1^0ϵ_2}})(N_1^0N_2)\right].`$ (192)
### I Gluons of color 8
(i) $`\mu =\nu =0`$: For $`\stackrel{~}{\mathrm{\Pi }}^{00}`$ one obtains
$`\stackrel{~}{\mathrm{\Pi }}^{00}(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}(1+e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2)}`$ (196)
$`\times [({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2){\displaystyle \frac{ϵ_1ϵ_2\xi _1\xi _2+\varphi _1\varphi _2}{2ϵ_1ϵ_2}}`$
$`+({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2){\displaystyle \frac{ϵ_1ϵ_2+\xi _1\xi _2\varphi _1\varphi _2}{2ϵ_1ϵ_2}}].`$
(ii) $`\mu =0,\nu =i`$: For $`\stackrel{~}{\mathrm{\Pi }}^{0i}`$ one simply has
$$\stackrel{~}{\mathrm{\Pi }}^{0i}(P)\mathrm{\Pi }_{11}^{0i}(P).$$
(197)
(iii) $`\mu =i,\nu =j`$: The magnetic components $`\stackrel{~}{\mathrm{\Pi }}^{ij}`$ are
$`\stackrel{~}{\mathrm{\Pi }}^{ij}(P)`$ $`=`$ $`{\displaystyle \frac{1}{4}}g^2N_f{\displaystyle \frac{d^3𝐤}{(2\pi )^3}\underset{e_1,e_2=\pm }{}\left[\delta ^{ij}\left(1e_1e_2\widehat{𝐤}_1\widehat{𝐤}_2\right)+e_1e_2\left(\widehat{k}_1^i\widehat{k}_2^j+\widehat{k}_1^j\widehat{k}_2^i\right)\right]}`$ (200)
$`\times [({\displaystyle \frac{1}{p_0+ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0ϵ_1ϵ_2}})(1N_1N_2){\displaystyle \frac{ϵ_1ϵ_2\xi _1\xi _2\varphi _1\varphi _2}{2ϵ_1ϵ_2}}`$
$`+({\displaystyle \frac{1}{p_0ϵ_1+ϵ_2}}{\displaystyle \frac{1}{p_0+ϵ_1ϵ_2}})(N_1N_2){\displaystyle \frac{ϵ_1ϵ_2+\xi _1\xi _2+\varphi _1\varphi _2}{2ϵ_1ϵ_2}}].`$
Equations (168) – (IV I) are the central result of this work. Starting from these equations, one can derive explicit expressions for the gluon self-energy in a two-flavor color superconductor for arbitrary $`p_0`$ and $`𝐩`$. As a first step, in the remainder of this work I compute the color-electric (Debye) screening mass, as well as the color-magnetic (Meissner) mass. These are obtained from the gluon self-energy in the static limit, $`p_0=0`$, for $`p0`$. Then I compute the self-energy for $`p_0=0`$, but $`p\varphi _0`$.
## V Debye screening and Meissner effect
In the following, I shall always assume that antiparticle gaps are small, $`\varphi ^{}0`$, and consequently that
$$ϵ_𝐤^{}ϵ_{𝐤0}^{},n_𝐤^{}n_{𝐤0}^{}1,N(ϵ_𝐤^{})0.$$
(201)
Therefore, thermal antiparticle occupation numbers and their derivatives will be neglected. As in the previous section, the different color sectors will be discussed separately.
### A Gluons with colors 1, 2, and 3
(i) $`\mu =\nu =0`$: I show several calculational steps in greater detail to illustrate the main approximations used throughout the following. For $`p_0=0`$, $`p0`$, $`𝐤_2𝐤_1𝐤`$, and only particle-particle ($`e_1=e_2=+1`$), or antiparticle-antiparticle ($`e_1=e_2=1`$) excitations contribute in the sum over $`e_1`$ and $`e_2`$ in (172). This is very similar to what happens in the HDL limit, cf. Sec. III F. Furthermore
$$\frac{ϵ_1ϵ_2\xi _1\xi _2\varphi _1\varphi _2}{2ϵ_1ϵ_2}0,\frac{ϵ_1ϵ_2+\xi _1\xi _2+\varphi _1\varphi _2}{2ϵ_1ϵ_2}1.$$
(202)
In the limit $`𝐤_2𝐤_1`$, $`(N_1N_2)/(ϵ_1ϵ_2)dN/dϵ`$, and neglecting the variation of $`N(ϵ_𝐤^{})`$, as discussed above, one obtains
$$\mathrm{\Pi }_{11}^{00}(0)\frac{g^2N_f}{2\pi ^2}_0^{\mathrm{}}𝑑kk^2\frac{dN(ϵ_𝐤^+)}{dϵ_𝐤^+}.$$
(203)
As the thermal occupation number varies appreciably only close to the Fermi surface, it is permissible to approximate $`k^2\mu ^2`$, and to restrict the $`k`$ integration to the region $`0k2\mu `$. Introducing the variable
$$\xi k\mu ,$$
(204)
one obtains with Eq. (79)
$$\mathrm{\Pi }_{11}^{00}(0)3m_g^2_0^\mu \frac{d\xi }{2T}\frac{1}{\mathrm{cosh}^2\left(\sqrt{\xi ^2+\varphi ^2}/2T\right)}.$$
(205)
Now change variables to $`\zeta \xi /2T`$, and remembering that $`\mu \varphi T`$, send the upper limit of the integral to infinity,
$$\mathrm{\Pi }_{11}^{00}(0)3m_g^2_0^{\mathrm{}}𝑑\zeta \frac{1}{\mathrm{cosh}^2\sqrt{\zeta ^2+(\varphi /2T)^2}}.$$
(206)
This expression has two interesting limits. For $`T0`$, the integrand becomes zero everywhere, and
$$T0:\mathrm{\Pi }_{11}^{00}(0)0.$$
(207)
At zero temperature, static, homogeneous electric fields of colors 1, 2, or 3, are not screened.
The other limit is when $`TT_c`$, and $`\varphi 0`$. Then, as $`_0^{\mathrm{}}𝑑\zeta /\mathrm{cosh}^2\zeta 1`$,
$$TT_c:\mathrm{\Pi }_{11}^{00}(0)3m_g^2\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(0).$$
(208)
As expected, $`\mathrm{\Pi }_{11}^{00}(0)`$ approaches the value in the normal phase, Eq. (80).
The interpretation of this result is the following. From the explicit form of the Gell-Mann matrices it is clear that gluons of adjoint colors 1, 2, and 3 “see” only quarks with fundamental colors 1 and 2. However, at $`T=0`$, all these quarks are bound in Cooper pairs to form a condensate of fundamental color (anti-)3, to which these gluons are “blind”. Hence, at $`T=0`$ the color superconductor is transparent with respect to these color fields. There is nothing which could screen these fields, thus there is no Debye mass for the gluons of colors 1, 2, or 3. Of course, this holds only in the limit $`p_0=0`$, $`p0`$, because only then are the gluons unable to resolve the individual quarks (with colors that can be “seen”) inside a Cooper pair.
When $`T`$ is nonzero, quasiparticles are thermally excited, and screening sets in. As $`T`$ approaches $`T_c`$, the condensate melts completely, and all quarks with the right colors to screen gluon fields with colors 1, 2, and 3 are freed. Then, the gluon self-energy approaches its value in the normal phase.
(ii) $`\mu =0,\nu =i`$: From Eq. (175) it is clear that
$$\mathrm{\Pi }_{11}^{0i}(0,𝐩)0.$$
(209)
This is similar to the normal phase in the static limit, Eq. (83).
(iii) $`\mu =i,\nu =j`$: As in the HDL limit, the magnetic components of the gluon self-energy receive contributions not only from particle-particle and antiparticle-antiparticle, but also from particle-antiparticle excitations. With Eq. (201) and $`𝑑\mathrm{\Omega }\widehat{k}^i\widehat{k}^j/(4\pi )=\delta ^{ij}/3`$, one obtains from Eq. (178)
$`\mathrm{\Pi }_{11}^{ij}(0)`$ $``$ $`\delta ^{ij}{\displaystyle \frac{g^2N_f}{6\pi ^2}}{\displaystyle _0^{\mathrm{}}}dkk^2\{{\displaystyle \frac{[\varphi _𝐤^+]^2}{2[ϵ_𝐤^+]^3}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_𝐤^+}{2T}}\right){\displaystyle \frac{dN(ϵ_𝐤^+)}{dϵ_𝐤^+}}{\displaystyle \frac{\xi ^2}{[ϵ_𝐤^+]^2}}`$ (211)
$`+{\displaystyle \frac{4[1N(ϵ_𝐤^+)](1n_𝐤^+)}{ϵ_𝐤^++k+\mu }}{\displaystyle \frac{4N(ϵ_𝐤^+)n_𝐤^+}{ϵ_𝐤^+k\mu }}{\displaystyle \frac{2}{k}}\},`$
where the last term was added to subtract the (UV-divergent) vacuum contribution, and where $`\varphi _𝐤^+\varphi ^+(ϵ_𝐤^+,𝐤)`$.
At zero temperature, and after an integration by parts ($`dn_𝐤^+/dk=[\varphi _𝐤^+]^2/2[ϵ_𝐤^+]^3`$),
$$\mathrm{\Pi }_{11}^{ij}(0)\delta ^{ij}\frac{g^2N_f}{6\pi ^2}_0^{\mathrm{}}𝑑kk\frac{4ϵ_𝐤^+}{ϵ_𝐤^++k+\mu }n_𝐤^+(1n_𝐤^+).$$
(212)
The term $`n_𝐤^+(1n_𝐤^+)`$ is proportional to $`[\varphi _𝐤^+]^2`$. The momentum dependence of the gap function is $`\varphi _𝐤^+=\varphi _0\mathrm{sin}(\overline{g}x_𝐤)`$ , where $`\overline{g}=g/(3\sqrt{2}\pi )`$ and $`x_𝐤\mathrm{ln}[2b\mu /(ϵ_𝐤^++|\xi |)]`$, with $`\xi `$ defined in Eq. (204) and $`b256\pi ^4[2/(N_fg^2)]^{5/2}`$. The gap function peaks at the Fermi surface, and is small far away from the Fermi surface. Therefore, the region $`k2\mu `$ can be neglected.
In the remaining integral over the region $`0k2\mu `$, take $`k\mu `$ in the slowly varying factor $`k/(ϵ_𝐤^++k+\mu )`$, and change the integration variable to $`\xi `$:
$$\mathrm{\Pi }_{11}^{ij}(0)\delta ^{ij}\frac{g^2N_f}{6\pi ^2}_0^\mu \frac{d\xi }{ϵ_𝐤^+}\left[\varphi _𝐤^+\right]^2.$$
(213)
Inserting the solution of the gap equation (including the momentum dependence), and changing the integration variable to $`x=\mathrm{ln}[2b\mu /(ϵ_𝐤^++\xi )]`$, this integral can be solved analytically. However, it turns out that this is unnecessary, if one only wants to know the parametric dependence on the gap and the QCD coupling constant in weak coupling, $`g1`$. One can simply neglect the momentum dependence of the gap function, and approximate $`\varphi _𝐤^+`$ by its value at the Fermi surface, $`\varphi _0`$, to obtain
$$\mathrm{\Pi }_{11}^{ij}(0)\delta ^{ij}m_g^2\frac{\varphi _0^2}{\mu ^2}\mathrm{ln}\left(\frac{2\mu }{\varphi _0}\right).$$
(214)
As $`\varphi _0\mu \mathrm{exp}(c_{\mathrm{QCD}}/g)`$, $`\mathrm{\Pi }_{11}^{ij}`$ is formally of order $`g\varphi _0^2`$. To this order, I cannot exclude that there are cancellations from other terms I have neglected (for instance the antiparticle gaps). To leading order, the result (214) is therefore consistent with $`\mathrm{\Pi }_{11}^{ij}(0)0`$.
Finally, as $`TT_c`$, an integration by parts shows that the expression (211) approaches the HDL limit, Eq. (89).
### B Gluons with colors 4 and 6
(i) $`\mu =\nu =0`$: For $`p_0=0`$, $`p0`$, and with the approximations (201), Eq. (181) becomes
$$\mathrm{\Pi }_{44}^{00}(0)\frac{g^2N_f}{2\pi ^2}_0^{\mathrm{}}𝑑kk^2\left\{\frac{1n_𝐤^+}{ϵ_𝐤^+\xi }\left[N_F^+(k)N(ϵ_𝐤^+)\right]+\frac{n_𝐤^+}{ϵ_𝐤^++\xi }\left[1N_F^+(k)N(ϵ_𝐤^+)\right]\right\}.$$
(215)
At $`T=0`$, and restricting the $`k`$ integration to the range $`0k2\mu `$ (as before, the momentum dependence of the gap function suppresses any contribution from the region $`k2\mu `$), this can be transformed into
$$\mathrm{\Pi }_{44}^{00}(0)3m_g^2_0^\mu \frac{d\xi }{ϵ_𝐤^+}\left(1+\frac{\xi ^2}{\mu ^2}\right)\frac{ϵ_𝐤^+\xi }{ϵ_𝐤^++\xi }.$$
(216)
Neglecting the momentum dependence of the gap function, the remaining integral can be done introducing the variable
$$y\mathrm{ln}\left(\frac{ϵ_𝐤^++\xi }{\varphi _0}\right).$$
(217)
To leading order, the result is
$$\mathrm{\Pi }_{44}^{00}(0)\frac{3}{2}m_g^2.$$
(218)
The Debye mass is reduced by a factor 2 as compared to the value in the normal phase.
The limit $`TT_c`$ cannot be studied with Eq. (215), and one has to go back to Eq. (181). It is obvious that one will reproduce the HDL result (80).
(ii) $`\mu =0,\nu =i`$: With Eq. (183), and the same approximations as before, one obtains
$`\mathrm{\Pi }_{44}^{0i}(0)`$ $``$ $`{\displaystyle \frac{g^2N_f}{2\pi ^2}}{\displaystyle _0^{\mathrm{}}}𝑑kk^2{\displaystyle \frac{d\mathrm{\Omega }}{4\pi }\widehat{k}^i\left\{\frac{1n_𝐤^+}{ϵ_𝐤^+\xi }\left[N_F^+(k)N(ϵ_𝐤^+)\right]+\frac{n_𝐤^+}{ϵ_𝐤^++\xi }\left[1N_F^+(k)N(ϵ_𝐤^+)\right]\right\}}`$ (219)
$``$ $`0,`$ (220)
by symmetry.
(iii) $`\mu =i,\nu =j`$: From Eq. (185) one derives under the same approximations
$`\mathrm{\Pi }_{44}^{ij}(0)`$ $``$ $`\delta ^{ij}{\displaystyle \frac{g^2N_f}{6\pi ^2}}{\displaystyle _0^{\mathrm{}}}dkk^2\{{\displaystyle \frac{1n_𝐤^+}{ϵ_𝐤^+\xi }}[N_F^+(k)N(ϵ_𝐤^+)]+{\displaystyle \frac{n_𝐤^+}{ϵ_𝐤^++\xi }}[1N_F^+(k)N(ϵ_𝐤^+)]`$ (222)
$`+{\displaystyle \frac{1}{k}}[1N_F^+(k)]+2{\displaystyle \frac{1n_𝐤^+}{ϵ_𝐤^++k+\mu }}[1N(ϵ_𝐤^+)]2{\displaystyle \frac{n_𝐤^+}{ϵ_𝐤^+k\mu }}N(ϵ_𝐤^+){\displaystyle \frac{2}{k}}\},`$
where the last term is a vacuum subtraction.
At $`T=0`$, the integral over the first two terms in the integrand has already been computed for $`\mathrm{\Pi }_{44}^{00}(0)`$, with the result (218). This is cancelled by a part of the vacuum subtraction. The remainder is
$$\mathrm{\Pi }_{44}^{ij}(0)\delta ^{ij}\frac{g^2N_f}{6\pi ^2}_0^{\mathrm{}}𝑑k\frac{k}{ϵ_𝐤^+}\frac{\mu (ϵ_𝐤^+\xi )+\left[\varphi _𝐤^+\right]^2}{ϵ_𝐤^++\xi +2\mu }.$$
(223)
Because the momentum dependence of the gap function suppresses the contribution from momenta far from the Fermi surface, the integral can be restricted to the region $`0k2\mu `$. To leading order, one may neglect $`\left[\varphi _𝐤^+\right]^2`$ in the numerator. \[The respective contribution is of order $`\varphi _0^2\mathrm{ln}(2\mu /\varphi _0)`$.\] Then, introduce the integration variable $`z=ϵ_𝐤^+k+\mu `$. Neglecting the momentum dependence of the gap function, as well as terms of order $`[\varphi _𝐤^+]^2`$, one obtains
$$\mathrm{\Pi }_{44}^{ij}(0)\delta ^{ij}\frac{g^2N_f}{12\pi ^2}\mu _0^{2\mu }𝑑z\left(1\frac{z}{2\mu }\right)=\delta ^{ij}\frac{m_g^2}{2}.$$
(224)
The limit $`TT_c`$ is not well-defined for Eq. (222); using Eq. (185) it is, however, straightforward to show that $`\mathrm{\Pi }_{44}^{ij}(0)\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(0)`$, as expected.
### C Gluons with color 5 and 7
In the limit $`p_0=0`$, $`p0`$, i.e., $`𝐤_2𝐤_1`$, it is obvious from comparing Eqs. (153) and (157) that
$$\mathrm{\Pi }_{44}^{\mu \nu }(0)\mathrm{\Pi }_{55}^{\mu \nu }(0),$$
(225)
hence, the results from the previous subsection can be carried over.
### D Gluons with color 8
(i) $`\mu =\nu =0`$: From Eq. (196) one obtains with the approximations (201)
$$\stackrel{~}{\mathrm{\Pi }}^{00}(0)\frac{g^2N_f}{2\pi ^2}_0^{\mathrm{}}𝑑kk^2\left\{\frac{dn_𝐤^+}{dk}\left[12N(ϵ_𝐤^+)\right]+\frac{dN(ϵ_𝐤^+)}{dk}\left(12n_𝐤^+\right)\right\}.$$
(226)
The integrand is vanishingly small except close to the Fermi surface. One can therefore restrict the $`k`$ integration to the range $`0k2\mu `$. Then, introducing $`\xi `$ as integration variable and using the symmetry of the integrand around $`\xi =0`$,
$$\stackrel{~}{\mathrm{\Pi }}^{00}(0)3m_g^2_0^\mu 𝑑\xi \frac{d}{d\xi }\left[\frac{\xi }{ϵ_𝐤^+}\mathrm{tanh}\left(\frac{ϵ_𝐤^+}{2T}\right)\right],$$
(227)
where higher order terms ($`\xi ^2/\mu ^2`$) in the integrand have been neglected. The remaining integral is unity (remember that $`\mu T`$), and the final result is
$$\stackrel{~}{\mathrm{\Pi }}^{00}(0)3m_g^2.$$
(228)
Note that this result is independent of the temperature. One concludes that
$$\mathrm{\Pi }_{88}^{00}(0)\frac{2}{3}\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(0)+\frac{1}{3}\stackrel{~}{\mathrm{\Pi }}^{00}(0)3m_g^2$$
(229)
does not change with temperature in the superconducting phase; it always has the same value as in the normal phase.
(ii) $`\mu =0,\nu =i`$: On account of Eqs. (197) and (209),
$$\stackrel{~}{\mathrm{\Pi }}^{0i}(0,𝐩)0.$$
(230)
Consequently, also $`\mathrm{\Pi }_{88}^{0i}(0)0`$.
(iii) $`\mu =i,\nu =j`$: For $`\stackrel{~}{\mathrm{\Pi }}^{ij}(0)`$ one derives from Eq. (200) with the standard approximations
$$\stackrel{~}{\mathrm{\Pi }}^{ij}(0)\delta ^{ij}\frac{g^2N_f}{6\pi ^2}_0^{\mathrm{}}𝑑kk^2\left\{\frac{dN(ϵ_𝐤^+)}{dϵ_𝐤^+}+\frac{4[1N(ϵ_𝐤^+)](1n_𝐤^+)}{ϵ_𝐤^++k+\mu }\frac{4N(ϵ_𝐤^+)n_𝐤^+}{ϵ_𝐤^+k\mu }\frac{2}{k}\right\},$$
(231)
where the last term is a vacuum subtraction.
At $`T=0`$, Eq. (231) becomes twice the integral in Eq. (223), hence
$$\stackrel{~}{\mathrm{\Pi }}^{ij}(0)\delta ^{ij}m_g^2.$$
(232)
As a consequence,
$$\mathrm{\Pi }_{88}^{ij}(0)\delta ^{ij}\frac{m_g^2}{3}.$$
(233)
As $`TT_c`$, an integration by parts shows that $`\stackrel{~}{\mathrm{\Pi }}^{ij}(0)0`$, as it should be. Consequently, also $`\mathrm{\Pi }_{88}^{ij}(0)0`$.
This concludes the discussion of Debye screening and the Meissner effect. In the next section, it will be demonstrated that for momenta $`p\varphi _0`$, i.e., when the gluon momentum is large enough to resolve the quarks in a Cooper pair, the gluon self-energy approaches the value in the normal phase.
## VI Nonzero gluon momentum
In this section, the gluon self-energy will be computed in the static limit, but for gluon momenta $`\varphi _0p\mu `$. In the condensed matter literature, this limit is known as the Pippard limit . The actual calculation follows closely that for ordinary superconductors (see for instance ). It will be convenient to consider the difference between the self energies in the superconducting and normal phases,
$$\delta \mathrm{\Pi }\mathrm{\Pi }\mathrm{\Pi }_0.$$
(234)
For large gluon momenta, effects from the pairing of quarks have to vanish, as the gluon wave length is short enough to resolve individual quarks in a Cooper pair. Consequently, the Debye mass for gluons of color 1, 2, and 3 can no longer vanish, but must approach the value in the normal phase. Simultaneously, for gluons of color 8 the Meissner effect has to vanish. These are the two cases studied in this section.
Of course, also the electric and magnetic masses of gluons with colors 4, 5, 6, and 7 have to approach their values in the normal phase. I was, however, not able to derive simple analytical expressions for the self-energy of these gluons in the limit $`\varphi _0p\mu `$. An explicit numerical study will be deferred to the future.
First note that for $`p\mu `$, $`k\mu `$,
$$k_{1,2}k\pm \frac{\widehat{𝐤}𝐩}{2}.$$
(235)
This then leads to the same expressions (71) – (73) for the spin traces as in the HDL limit. As in the previous section, quasi-antiparticles will be treated as real antiparticles, cf. Eq. (201). Furthermore, for the sake of notational convenience, let us introduce
$$\xi _\pm \xi \pm \frac{\widehat{𝐤}𝐩}{2},ϵ_\pm ϵ_{𝐤_{1,2}}^+,\varphi _\pm \varphi ^+(ϵ_\pm ),n_\pm n_{𝐤_{1,2}}^+,N_\pm N(ϵ_\pm ).$$
(236)
### A Electric gluons of color 1, 2, and 3
Writing $`N_\pm =[1\mathrm{tanh}(ϵ_\pm /2T)]/2`$, the self-energy of electric gluons of colors 1, 2, and 3 is from Eq. (172)
$`\mathrm{\Pi }_{11}^{00}(0,𝐩)`$ $``$ $`{\displaystyle \frac{g^2N_f}{2}}{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}\{{\displaystyle \frac{1}{ϵ_++ϵ_{}}}[\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right)+\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)]{\displaystyle \frac{1}{2}}(1{\displaystyle \frac{\xi _+\xi _{}+\varphi _+\varphi _{}}{ϵ_+ϵ_{}}})`$ (238)
$`+{\displaystyle \frac{1}{ϵ_+ϵ_{}}}[\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right)\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)]{\displaystyle \frac{1}{2}}(1+{\displaystyle \frac{\xi _+\xi _{}+\varphi _+\varphi _{}}{ϵ_+ϵ_{}}})\},`$
where terms of order $`p^2/k^2`$ have been neglected. The self-energy in the normal phase can be obtained either from Eq. (65), for $`p_0=0`$ and with the approximations (201), or directly from Eq. (238) in the limit $`\varphi _\pm 0`$:
$$\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(0,𝐩)\frac{g^2N_f}{2}\frac{d^3𝐤}{(2\pi )^3}\frac{1}{\xi _+\xi _{}}\left[\mathrm{tanh}\left(\frac{\xi _+}{2T}\right)\mathrm{tanh}\left(\frac{\xi _{}}{2T}\right)\right].$$
(239)
Now consider the difference $`\delta \mathrm{\Pi }_{11}^{00}(0,𝐩)`$ between (238) and (239). As the main contribution to the integral over $`𝐤`$ comes from the region around the Fermi surface, it is admissible to neglect the momentum dependence of the gap function, $`\varphi _+\varphi _{}\varphi `$. Then one rearranges the integrand to separate terms of the form
$$\frac{1}{\xi _+\xi _{}}\left[\frac{\xi _\pm }{ϵ_\pm }\mathrm{tanh}\left(\frac{ϵ_\pm }{2T}\right)\mathrm{tanh}\left(\frac{\xi _\pm }{2T}\right)\right].$$
(240)
As argued in , these terms vanish by symmetry when integrating over $`\xi `$. (A careful analysis shows that this is correct to leading order in $`\varphi /p`$.) The result is
$$\delta \mathrm{\Pi }_{11}^{00}(0,𝐩)\frac{g^2N_f}{2}\frac{d^3𝐤}{(2\pi )^3}\frac{\varphi ^2}{\xi \widehat{𝐤}𝐩}\left[\frac{1}{ϵ_+}\mathrm{tanh}\left(\frac{ϵ_+}{2T}\right)\frac{1}{ϵ_{}}\mathrm{tanh}\left(\frac{ϵ_{}}{2T}\right)\right].$$
(241)
As the integrand peaks at the Fermi surface, $`\xi 0`$, and for $`\widehat{𝐤}𝐩0`$, one can approximate the hyperbolic tangens by $`\mathrm{tanh}(ϵ_\pm /2T)\mathrm{tanh}(\varphi /2T)`$, and obtains to leading order
$$\delta \mathrm{\Pi }_{11}^{00}(0,𝐩)3m_g^2\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right)_0^{\mu /\varphi }\frac{dx}{x}_0^{p/2\varphi }\frac{dy}{y}\left(\frac{1}{\sqrt{(x+y)^2+1}}\frac{1}{\sqrt{(xy)^2+1}}\right),$$
(242)
where $`x\xi /\varphi `$, $`y\widehat{𝐤}𝐩/(2\varphi )`$. The $`y`$ integral can be done exactly. In the limit $`\mu p\varphi `$,
$$\delta \mathrm{\Pi }_{11}^{00}(0,𝐩)3m_g^2\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right)_0^{\mathrm{}}𝑑u\frac{2u}{\mathrm{sinh}u}3m_g^2\frac{\pi ^2}{2}\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right).$$
(243)
The self-energy in the normal phase is approximately constant for momenta $`p\mu `$, such that
$$\mathrm{\Pi }_{11}^{00}(0,𝐩)3m_g^2\left[1\frac{\pi ^2}{2}\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right)\right].$$
(244)
This shows that the absolute value of the self-energy in the superconducting phase is reduced as compared to the normal phase. For increasing $`p/\varphi `$, the correction becomes smaller, such that electric fields for adjoint colors 1, 2, and 3 are screened over an only slightly longer distance than in the normal phase. In this case, the gluons “see” the individual fundamental color charges inside the Cooper pairs.
For decreasing $`p/\varphi `$, however, the correction becomes larger. This is in agreement with the results of Sec. V, where the self-energy of gluons with colors 1, 2, and 3 was found to vanish in the limit $`p0`$, i.e., when the gluon momentum is too small to resolve individual quarks inside a Cooper pair. Although strictly valid only for $`p\varphi `$, by extrapolating Eq. (244) to $`p\varphi `$ one would conclude that, at $`T=0`$, this happens once $`p`$ is smaller than $`5\varphi _0`$.
### B Magnetic gluons of color 8
For magnetic gluons, one derives from Eq. (200)
$`\stackrel{~}{\mathrm{\Pi }}^{ij}(0,𝐩)`$ $``$ $`{\displaystyle \frac{g^2N_f}{2}}{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}(\widehat{k}^i\widehat{k}^j\{{\displaystyle \frac{1}{\xi _+\xi _{}}}[{\displaystyle \frac{\xi _+}{ϵ_+}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right){\displaystyle \frac{\xi _{}}{ϵ_{}}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)]`$ (248)
$`+{\displaystyle \frac{\varphi ^2}{\xi \widehat{𝐤}𝐩}}[{\displaystyle \frac{1}{ϵ_+}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right){\displaystyle \frac{1}{ϵ_{}}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)]\}`$
$`+(\delta ^{ij}\widehat{k}^i\widehat{k}^j){\displaystyle \frac{1}{2k}}\{2+{\displaystyle \frac{\xi _+}{ϵ_+}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right)+{\displaystyle \frac{\xi _{}}{ϵ_{}}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)`$
$`{\displaystyle \frac{\varphi ^2}{2\mu }}[{\displaystyle \frac{1}{ϵ_+}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_+}{2T}}\right)+{\displaystyle \frac{1}{ϵ_{}}}\mathrm{tanh}\left({\displaystyle \frac{ϵ_{}}{2T}}\right)]\}).`$
Here, the momentum dependence of the gap function was neglected, $`\varphi _\pm \varphi `$. Moreover, in denominators which contain terms $`\mu ^2`$, $`ϵ_\pm ^2`$ was approximated by $`\xi _\pm ^2`$.
In the normal phase, the corresponding expression reads
$`\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(0,𝐩)`$ $``$ $`{\displaystyle \frac{g^2N_f}{2}}{\displaystyle }{\displaystyle \frac{d^3𝐤}{(2\pi )^3}}\{\widehat{k}^i\widehat{k}^j{\displaystyle \frac{1}{\xi _+\xi _{}}}[\mathrm{tanh}\left({\displaystyle \frac{\xi _+}{2T}}\right)\mathrm{tanh}\left({\displaystyle \frac{\xi _{}}{2T}}\right)]`$ (250)
$`+(\delta ^{ij}\widehat{k}^i\widehat{k}^j){\displaystyle \frac{1}{2k}}[2+\mathrm{tanh}\left({\displaystyle \frac{\xi _+}{2T}}\right)+\mathrm{tanh}\left({\displaystyle \frac{\xi _{}}{2T}}\right)]\}.`$
In the difference $`\delta \stackrel{~}{\mathrm{\Pi }}^{ij}`$, there are again terms like (240), which vanish by symmetry arguments. There is also a term $`\varphi ^2/(4\mu k)`$ which is of higher order and thus can be neglected. The remainder can be written as
$`\delta \stackrel{~}{\mathrm{\Pi }}^{ij}(0,𝐩)3m_g^2{\displaystyle \frac{\varphi }{p}}`$ (251)
$`\times `$ $`{\displaystyle _0^{\mu /\varphi }}{\displaystyle \frac{dx}{x}}{\displaystyle _0^{p/2\varphi }}{\displaystyle \frac{dy}{y}}\mathrm{tanh}\left({\displaystyle \frac{\varphi \sqrt{y^2+1}}{2T}}\right)\left({\displaystyle \frac{1}{\sqrt{(x+y)^2+1}}}{\displaystyle \frac{1}{\sqrt{(xy)^2+1}}}\right){\displaystyle _0^{2\pi }}{\displaystyle \frac{d\phi }{2\pi }}\widehat{k}^i\widehat{k}^j.`$ (252)
As before, $`x\xi /\varphi `$, $`y\widehat{𝐤}𝐩/2\varphi `$. Since the $`x`$ integral is dominated by the region around the Fermi surface, $`x0`$, I have set $`x=0`$ in the argument of the hyperbolic tangens.
For $`ij`$, the integration over the polar angle $`\phi `$ vanishes, thus $`\delta \stackrel{~}{\mathrm{\Pi }}^{ij}`$ is diagonal. However, not all diagonal elements are equal. Let $`𝐩=(0,0,p)`$. Then $`\widehat{k}_x^2=[1(2\varphi y/p)^2]\mathrm{cos}^2\phi `$, $`\widehat{k}_y^2=[1(2\varphi y/p)^2]\mathrm{sin}^2\phi `$, and the transverse components of $`\delta \stackrel{~}{\mathrm{\Pi }}^{ij}`$ are
$$\delta \stackrel{~}{\mathrm{\Pi }}^{xx}(0,𝐩)\delta \stackrel{~}{\mathrm{\Pi }}^{yy}(0,𝐩)m_g^2\frac{3\pi ^2}{4}\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right).$$
(253)
To obtain this result, I have used the fact that the $`y`$ integration is dominated by the region $`y0`$, and consequently have set $`y=0`$ in the hyperbolic tangens as well as in $`\widehat{k}_{x,y}^2`$. The remaining integral is then the same as in Eq. (242).
The longitudinal component can be shown to be of higher order in $`\varphi /p`$, such that to leading order,
$$\delta \stackrel{~}{\mathrm{\Pi }}^{zz}(0,𝐩)0.$$
(254)
This result is not unexpected: the self-energy for gluons in the normal phase is transverse, $`\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(0,𝐩)(\delta ^{ij}\widehat{p}^i\widehat{p}^j)p^2m_g^2/(12\mu ^2)`$. \[Note that this expressions is of order $`g^2p^2g^2\mu ^2`$, and thus not in contradiction to the HDL result (89).\] Equations (253) and (254) now combine to give a transverse self-energy for the eighth gluon, too,
$$\mathrm{\Pi }_{88}^{ij}(0,𝐩)\left(\delta ^{ij}\widehat{p}^i\widehat{p}^j\right)m_g^2\left[\frac{p^2}{12\mu ^2}+\frac{\pi ^2}{4}\frac{\varphi }{p}\mathrm{tanh}\left(\frac{\varphi }{2T}\right)\right].$$
(255)
## VII Summary, conclusions, and outlook
In color-superconducting quark matter with $`N_f=2`$ degenerate quark flavors, the condensate can be oriented in (anti-)3 direction in fundamental color space by means of a global color rotation. Then, only quarks with fundamental colors 1 and 2 form Cooper pairs, while quarks of the third fundamental color remain unpaired, and act as a background to neutralize the color-charged condensate. Since the unpaired quarks carry the same color charge, two of them are in the (repulsive) sextet representation of $`SU(3)_c`$. Consequently, they do not form Cooper pairs and the system is stable.
The condensate breaks the $`SU(3)_c`$ color symmetry to $`SU(2)_c`$. With the above color choice, the generators of the unbroken $`SU(2)_c`$ subgroup are the $`SU(3)_c`$ generators $`T^1,T^2`$, and $`T^3`$, with $`T^a=\lambda ^a/2`$ and the standard convention for the Gell-Mann matrices $`\lambda ^a`$. The gluons corresponding to the remaining generators $`T^4`$ through $`T^8`$ all receive a mass via the Anderson–Higgs mechanism.
What are the expected values for these masses? The effective Lagrangian for the low-energy excitations of the condensate fields minimally coupled to gauge fields has the kinetic term
$$_{\mathrm{eff}}^{\mathrm{kin}}=\alpha _\mathrm{e}\left(D_0\mathrm{\Phi }\right)^{}D^0\mathrm{\Phi }+\alpha _\mathrm{m}\left(D_i\mathrm{\Phi }\right)^{}D^i\mathrm{\Phi }.$$
(256)
The presence of a heat and particle bath at nonzero $`T`$ and/or $`\mu `$ breaks Lorentz invariance, so that the coefficient $`\alpha _\mathrm{e}`$ of the part containing the time derivatives can in principle be different from the one of the part containing the spatial derivatives, $`\alpha _\mathrm{m}`$.
For a two-flavor color-superconductor, $`\mathrm{\Phi }`$ is a $`SU(3)_c`$ (anti-)triplet, $`\mathrm{\Phi }(\mathrm{\Phi }_1,\mathrm{\Phi }_2,\mathrm{\Phi }_3)^T`$ . Consequently, the covariant derivative is $`D_\mu =_\mu igA_\mu ^aT^a`$, with the generators $`T^a`$ being in the fundamental representation. If $`\mathrm{\Phi }`$ attains a non-vanishing expectation value $`\mathrm{\Phi }=(0,0,\varphi _0)^T`$, $`\varphi _0𝐑`$, this generates a mass term for the gluon fields of the form
$`_1^\mathrm{M}`$ $`=`$ $`g^2\varphi _0^2\left(\alpha _\mathrm{e}A_0^aA_b^0+\alpha _\mathrm{m}A_i^aA_a^i\right)\delta _{3i}T_{ij}^aT_{jk}^b\delta _{k3}`$ (257)
$``$ $`g^2\varphi _0^2\left[{\displaystyle \frac{1}{4}}{\displaystyle \underset{a=4}{\overset{7}{}}}\left(\alpha _\mathrm{e}A_0^aA_a^0+\alpha _\mathrm{m}A_i^aA_a^i\right)+{\displaystyle \frac{1}{3}}\left(\alpha _\mathrm{e}A_0^8A_8^0+\alpha _\mathrm{m}A_i^8A_8^i\right)\right].`$ (258)
The expected electric and magnetic gluon masses are
$$M_{\mathrm{e},\mathrm{m}}^1=M_{\mathrm{e},\mathrm{m}}^2=M_{\mathrm{e},\mathrm{m}}^3=0,M_{\mathrm{e},\mathrm{m}}^4=M_{\mathrm{e},\mathrm{m}}^5=M_{\mathrm{e},\mathrm{m}}^6=M_{\mathrm{e},\mathrm{m}}^7=\sqrt{\frac{\alpha _{\mathrm{e},\mathrm{m}}}{2}}g\varphi _0,M_\mathrm{e}^8=\sqrt{\frac{2\alpha _{\mathrm{e},\mathrm{m}}}{3}}g\varphi _0,$$
(259)
such that the ratio
$$R_{\mathrm{e},\mathrm{m}}\left(\frac{M_{\mathrm{e},\mathrm{m}}^8}{M_{\mathrm{e},\mathrm{m}}^4}\right)^2=4/3.$$
(260)
In this work, the gluon self-energy in a $`N_f=2`$ color superconductor has been derived. Due to the pattern of symmetry breaking, one has to study the individual gluon colors separately. The central result are equations (168) – (IV I). Various limits of these expressions are of interest. Here, the self-energy was computed in the static, homogeneous limit, $`p_0=0`$, $`p0`$, which yields the Debye mass for electric and the Meissner mass for magnetic gluons. The main results are summarized in Table I.
For the three gluons of the unbroken $`SU(2)_c`$ subgroup (gluon colors 1, 2, and 3), the Debye mass as well as the Meissner mass vanish. While this is in agreement with (259), it is at first physically unclear, and therefore quite surprising, why gluon fields with colors 1, 2, and 3 are not screened. To explain this, I argued as follows. Gluons with adjoint colors 1, 2, and 3 couple to fundamental colors 1 and 2. At $`T=0`$, however, all quarks with these color charges are bound in Cooper pairs which have fundamental color (anti-)3. Thus, these gluons cannot “see” the quark charges, and hence are unscreened. At nonzero $`T`$, quasiparticles are thermally excited. They have the “right” fundamental color (1 and 2) to screen gluon fields with adjoint colors 1, 2, and 3, and consequently lead to screening and a nonzero Debye mass. At $`T=T_c`$, when the condensate melts, the Debye mass assumes its standard value in the normal phase.
Of course, at $`T=0`$ the gluon self-energy vanishes only in the zero-energy, zero-momentum limit, since then the gluon field cannot resolve individual quarks inside the Cooper pair. For large gluon momentum $`p\varphi _0`$, electric gluon fields are screened; the self-energy is the same as in the normal phase, up to a correction of order $`m_g^2\varphi _0/p`$, as computed in Sec. VI A.
The gluons corresponding to the broken generators of $`SU(3)_c`$ all attain a mass through the Anderson–Higgs mechanism. While the Debye mass for electric gluons of color 8 is the same as in the normal phase, the Debye mass squared for colors 4 through 7 is only half as large. As $`T`$ approaches $`T_c`$, however, the melting of the condensate leads to an increase of the Debye mass to its standard value. At zero temperature, the ratio of the Debye masses squared of gluon color 8 and 4 is $`R_\mathrm{e}\mathrm{\Pi }_{88}^{00}(0)/\mathrm{\Pi }_{44}^{00}(0)=2`$.
The Meissner mass squared for gluons of color 8 is $`1/3`$ of the gluon mass squared, $`m_g^2`$, while that for gluons of colors 4 through 7 is $`1/2`$ of the gluon mass squared. The Meissner effect vanishes as $`T`$ approaches $`T_c`$, or when the gluon momentum $`p\varphi _0`$, as computed in Sec. VI B. The ratio of the Meissner masses squared of gluon color 4 and 8 is $`R_\mathrm{m}\mathrm{\Pi }_{88}^{ii}(0)/\mathrm{\Pi }_{44}^{ii}(0)=2/3`$.
Both $`R_\mathrm{e}`$ and $`R_\mathrm{m}`$ differ from the expectation (260). What is the origin of this discrepancy? The kinetic term (257) is not the only possible invariant in an effective Lagrangian, where the condensate fields are minimally coupled to the gauge fields. Another possibility is the term
$$_{\mathrm{eff}}^{}=\beta _\mathrm{e}\left(\mathrm{\Phi }^{}D_0\mathrm{\Phi }\right)^{}\mathrm{\Phi }^{}D^0\mathrm{\Phi }+\beta _\mathrm{m}\left(\mathrm{\Phi }^{}D_i\mathrm{\Phi }\right)^{}\mathrm{\Phi }^{}D^i\mathrm{\Phi },$$
(261)
which has mass dimension six \[consequently, $`\beta _{\mathrm{e},\mathrm{m}}`$ have dimension (mass)<sup>-2</sup>\]. Note that in the nonlinear version of the effective theory , where the modulus of $`\mathrm{\Phi }`$ does not change, only the phase, this term is identical to the standard kinetic term (257).
Upon condensation, $`\mathrm{\Phi }=(0,0,\varphi _0)^T`$, the term (261) contributes to the mass of the eighth gluon,
$$_2^\mathrm{M}=g^2\varphi _0^4\frac{1}{3}\left(\beta _\mathrm{e}A_0^8A_8^0+\beta _\mathrm{m}A_i^8A_8^i\right).$$
(262)
With this term, one reproduces the zero-temperature magnetic masses given in Table I with the choice
$$\alpha _\mathrm{m}\frac{m_g^2}{g^2\varphi _0^2}=\frac{N_f}{6\pi ^2}\frac{\mu ^2}{\varphi _0^2},\beta _\mathrm{m}\frac{1}{2}\frac{m_g^2}{g^2\varphi _0^4}=\frac{N_f}{12\pi ^2}\frac{\mu ^2}{\varphi _0^4}.$$
(263)
Note that the prefactor of the kinetic term (257) has the $`1/\varphi _0^2`$ behavior typical for effective theories of superconductivity . To reproduce the electric masses, the coefficients $`\alpha _\mathrm{e}`$ and $`\beta _\mathrm{e}`$ have to be chosen as
$$\alpha _\mathrm{e}3\alpha _\mathrm{m},\beta _\mathrm{e}=3\beta _\mathrm{m}.$$
(264)
The expressions (263) and (264) fix the prefactors of the kinetic term (257) and the higher-order term (261) in the effective low-energy theory of condensate fields coupled to gluons. Up to mass dimension four, the effective theory for an $`SU(3)_c`$ vector $`\mathrm{\Phi }`$ has, apart from the gauge field part, two more terms which are invariant under $`SU(3)_c`$ transformations : a mass term for the condensate field
$$_{\mathrm{eff}}^{\mathrm{mass}}=^2\mathrm{\Phi }^{}\mathrm{\Phi },$$
(265)
and a quartic self interaction
$$_{\mathrm{eff}}^{\mathrm{int}}=\lambda \left(\mathrm{\Phi }^{}\mathrm{\Phi }\right)^2.$$
(266)
Work is in progress to determine the condensate mass $``$ and the coupling constant $`\lambda `$ .
What is the impact of these results for the solution of the gap equations? Remember that, after taking into account the color and flavor structure, the gap matrix in spinor space obeys the gap equation
$$\mathrm{\Phi }^+(K)=\frac{3}{4}g^2\frac{T}{V}\underset{Q}{}\left[\mathrm{\Delta }_{11}^{\mu \nu }(KQ)\frac{1}{9}\mathrm{\Delta }_{88}^{\mu \nu }(KQ)\right]\gamma _\mu G_0^{}(Q)\mathrm{\Phi }^+(Q)G^+(Q)\gamma _\nu .$$
(267)
Previously , the gap equation was solved using the HDL propagator for both $`\mathrm{\Delta }_{11}`$ and $`\mathrm{\Delta }_{88}`$,
$$\mathrm{\Phi }^+(K)=\frac{2}{3}g^2\frac{T}{V}\underset{Q}{}\mathrm{\Delta }_{\mathrm{HDL}}^{\mu \nu }(KQ)\gamma _\mu G_0^{}(Q)\mathrm{\Phi }^+(Q)G^+(Q)\gamma _\nu ,$$
(268)
where $`\mathrm{\Delta }_{\mathrm{HDL}}^1\mathrm{\Delta }_0^1+\mathrm{\Pi }_0`$. The integral on the right-hand side is dominated by gluons with small momenta, $`KQ0`$. In the HDL limit, however, static electric gluons are screened by the Debye mass, $`\mathrm{\Pi }_{0}^{}{}_{}{}^{00}(0)3m_g^2`$, cf. Eq. (80). Their contribution is therefore suppressed as compared to that of magnetic gluons which are not screened in the static limit, $`\mathrm{\Pi }_{0}^{}{}_{}{}^{ij}(0)0`$, cf. Eq. (89). The dominant contribution to the gap integral therefore comes from (nearly) static magnetic gluons. A careful analysis shows that the gluon energy is not exactly zero, but $`p_0\varphi _0`$, while the gluon momentum is $`p(m_g^2\varphi _0)^{1/3}`$, and thus, in weak coupling, actually much larger than $`\varphi _0`$. The coefficient $`c_{\mathrm{QCD}}=3\pi ^2/\sqrt{2}`$ is determined by how many nearly static magnetic modes contribute, and by the precise form of the magnetic HDL propagator.
As shown in this paper, the gluon propagator in a two-flavor color superconductor is, at least in the static limit, $`p_0=0`$, and for small gluon momenta, $`p\varphi _0`$, drastically different from the HDL propagator. For instance, for gluon colors 1, 2, and 3, which constitute the main contribution to the gap equation (267), both magnetic and electric modes remain unscreened. For gluon color 8, previously unscreened static magnetic gluons attain a Meissner mass.
In order to assess the effect of these results on the solution of the gap equation, one needs to solve the gap equation with the full energy and momentum dependence of the gluon propagator in the superconducting phase, to decide which energies and momenta constitute the dominant contribution to the gap integral. If gluon energy and momentum are much larger than the zero-temperature gap, the impact will be rather small, because, as was shown in Sec. VI, the effect of the superconducting medium is only a small correction of order $`O(\varphi _0/p)`$ to the standard HDL propagator. This might influence the prefactor of the exponential $`\mathrm{exp}(c_{\mathrm{QCD}}/g)`$, but not $`c_{\mathrm{QCD}}`$ itself. On the other hand, if the dominant range of energies and momenta is $`p_0,p\varphi _0`$, the impact could be large and might even change $`c_{\mathrm{QCD}}`$. A detailed analysis of this problem is under investigation .
## Acknowledgements
I thank W. Brown, G. Carter, M. Gyulassy, R. Pisarski, K. Rajagopal, H.C. Ren, T. Schäfer, I. Shovkovy, and D.T. Son for discussions. I am especially indebted to G. Carter, for discussions on the ratio $`\mathrm{\Pi }_{88}(0)/\mathrm{\Pi }_{44}(0)`$, to R. Pisarski, for pointing out that $`_{\mathrm{eff}}^{}`$ explains the perturbative results for the Debye and Meissner masses, and to D.T. Son, for indicating the similarity of $`\mathrm{\Pi }_{88}`$ with the photon self-energy in ordinary superconductors. My thanks go to RIKEN, BNL and the U.S. Dept. of Energy for providing the facilities essential for the completion of this work, and to Columbia University’s Nuclear Theory Group for continuing access to their computing facilities. Finally, I would like to express my everlasting gratitude for the hospitality extended to me at Sherwood Castle, where part of this work was done. |
warning/0001/astro-ph0001467.html | ar5iv | text | # Quark phases in neutron stars and a “third family” of compact stars as a signature for phase transitions11footnote 1Supported by DFG and GSI Darmstadt.
## 1 Introduction
The properties of nuclear matter under extreme conditions are subject to intense experimental and theoretical studies in heavy-ion physics as well as in nuclear astrophysics. Especially neutron stars provide a unique astrophysical environment to study the properties of matter in the region of low temperatures and high baryonic densities. The densities deep inside neutron stars might be sufficient to populate numerous new forms of matter which may compete with each other. It is expected that at about two times normal nuclear matter density hyperons appear in neutron star matter . Also the formation of Bose-Einstein condensates or the phase transition to deconfined quark matter have been considered . For a recent review on possible phase transitions in neutron stars see . However, our understanding of the properties of neutron star matter is still restricted due to theoretical limits and uncertainties in modeling the equation of state (EOS). Measurements of neutron star masses and radii may put important constraints on the EOS at large density. The masses of about twenty neutron stars (mainly pulsars in binary systems) have been measured and found to be quite near the “canonical” mass of $`M=1.4M_{}`$ ($`M_{}`$ is the solar mass) . Unfortunately, the currently known masses do not constrain the EOS unambiguously. While some soft EOS – which can only account for smaller masses – can be ruled out, it is not possible to differentiate between the class of stiffer EOS which give higher neutron star masses. However, these EOS can be classified by their different predictions for neutron star radii. Precise measurements of neutron star radii can therefore give important information on the property of matter and possible exotic phases at large density. Some indirect radius estimates from observations of neutron stars seem to indicate surprisingly small radii. Recent radius estimates for Geminga (PSR 0630+17) by Golden and Shearer suggest an upper bound of the apparent radius of $`R_{\mathrm{}}9.5_{2.0}^{+3.5}`$km for an assumed blackbody source and $`R_{\mathrm{}}10.0_{2.1}^{+3.8}`$km with the presence of a magnetized $`H`$ atmosphere. Walter and Matthews obtained $`R_{\mathrm{}}\genfrac{}{}{0pt}{}{<}{}14`$km for the recently discovered isolated neutron star RXJ1856.5-3745 . Generally the apparent radius $`R_{\mathrm{}}`$ at a distance $`d\mathrm{}`$ is related to the local stellar radius $`R`$ by
$$R_{\mathrm{}}=R/\sqrt{12GM/Rc^2}.$$
(1)
If we assume a mass of $`M=1.4M_{}`$ an apparent radius $`R_{\mathrm{}}`$ of about $`1113`$km would restrict the radius to the quite low values of $`R710`$km. (Due to (1), radii below $`R_{\mathrm{}}11`$km would imply masses below $`1.4M_{}`$.) A radius of $`R\genfrac{}{}{0pt}{}{<}{}10`$km for a typical neutron star mass is furthermore consistent with the findings by Haberl and Titarchuk for the X-ray burster 4U 1820-30, with the findings by Pavlov and Zavlin for the millisecond pulsar J0437-4715 and by Li et al. for Her X-1. (In the context of Ref. see also Ref. and Refs. .) If radius estimates of $`R\genfrac{}{}{0pt}{}{<}{}10`$km indeed should prove to be reliable in connection with typical neutron star masses, the careful interpretation of these results in terms of neutron star matter properties would be an intriguing challenge for nuclear astrophysics. Since the appearance of a deconfined quark matter phase inside neutron stars is one possible mechanism to soften the EOS (and therefore to reduce the radius of the star), we want systematically address the question to which amount the compactness of a neutron star can be attributed to the presence of a quark phase. Such investigations are necessary also in view of more “exotic” interpretations of small radii e.g. in terms of “strange stars” . It is important to study whether such interpretations are a consequence of ruling out other mechanisms suitable to soften the EOS at high densities (e.g. by hyperons, kaon condensates or quark phases) or provide only one possible explanation of the observed neutron star properties. For this purpose we will apply various hadronic EOS in the relativistic mean-field model (including hyperons) and a wide range of model parameters of the quark phase which we will describe in an extended MIT bag model.
The EOS of the hadronic and the quark phase will be discussed in Sec. 2. In Sec. 3 we discuss the construction of a first-order phase transition from hadronic to quark matter and present the results for the EOS. In Sec. 4 we apply the EOS to the calculation of the neutron star structure and discuss the influence of quark phases. In Sec. 5 we discuss the properties of a possible “third family” of compact stars. Such hypothetical family of compact stars might exist besides the two known families of white dwarfs and neutron stars . In this work we show how future measurements of masses and radii of only two compact stars may prove the existence of a third family. Moreover, we argue that this would provide a novel signature for a phase transition inside neutron stars, revealing a particular behavior of the EOS in the almost unknown regime of supernuclear densities. Finally, we summarize our findings in Sec. 6.
## 2 Equations of state
We want to start with a discussion of the different models used to describe the equation of state of “neutron star matter” i.e. cold, charge neutral matter in beta equilibrium. Neutron star matter covers a wide range of densities. Starting from the density of iron ($`ϵ8`$g/cm<sup>3</sup>) which builds the surface of a neutron star, densities up to few times normal nuclear matter density ($`ϵ_0=140`$MeV/fm<sup>3</sup> $`2.5\times 10^{14}`$g/cm<sup>3</sup>) can be achieved in the center of the star. Unfortunately there is yet no single theory capable to meet the demands of the various degrees of freedom opened up in neutron star matter in different densities regimes. We are therefore forced to use different models at different density ranges.
### 2.1 Subnuclear densities
For subnuclear densities we use the Baym-Pethick-Sutherland EOS , which describes the crust of the neutron star. Up to densities of $`ϵ\genfrac{}{}{0pt}{}{<}{}10^7`$g/cm<sup>3</sup> matter is composed of a Coulomb lattice of $`{}_{28}{}^{}{}_{}{}^{56}`$Fe nuclei. The pressure is dominated by degenerate electrons. At higher densities nuclei become more and more neutron rich and at $`ϵ_{drip}`$ $`4\times 10^{11}`$g/cm<sup>3</sup> the most weakly bound neutrons start to drip out of the nuclei. For a detailed discussion of the Baym-Pethick-Sutherland EOS see . A review of the properties of neutron rich matter at subnuclear densities can be found in .
### 2.2 Nuclear densities
The Baym-Pethick-Sutherland EOS is matched at $`ϵ10^{14}`$g/cm$`{}_{}{}^{3}ϵ_0`$ to a hadronic EOS calculated in the framework of the relativistic mean-field (RMF) model. At these densities nuclei begin to dissolve and nucleons become the relevant degrees of freedom in this phase. The RMF model is widely used for the description of dense nuclear matter , especially in neutron stars. For an introduction into the RMF model see e.g. . We use three EOS calculated by Schaffner and Mishustin in the extended RMF model (i.e. TM1, TM2, GL85) and one by Ghosh, Phatak and Sahu . For the latter one we use GPS as an abbreviation. These models include hyperonic degrees of freedom which typically appear at $`ϵ23ϵ_0`$. Table 1 shows the nuclear matter properties and the particle composition of the four EOS.
Even if the relevant degrees of freedom are specified (in our case basically nucleons and hyperons) the high density range of the EOS is still not well understood. The use of different hadronic models should reflect this uncertainty to some degree. In the following we will denote the phase described by the Baym-Pethick-Sutherland EOS and by the RMF model as the hadronic phase (HP) of the neutron star.
### 2.3 Quark matter
At densities above $`ϵ_0`$ we allow the HP to undergo a phase transition to a deconfined quark matter phase (QP). This section describes how we model the QP in weak equilibrium. The QP consists of $`u`$, $`d`$, $`s`$ quarks and electrons in weak equilibrium i.e. the weak reactions
$`d`$ $``$ $`u+e^{}+\overline{\nu }_e^{},`$ (2)
$`s`$ $``$ $`u+e^{}+\overline{\nu }_e^{},`$
$`s+u`$ $``$ $`d+u,`$
imply relations between the four chemical potentials $`\mu _u`$, $`\mu _d`$, $`\mu _s`$, $`\mu _e`$ which read
$$\mu _s=\mu _d=\mu _u+\mu _e.$$
(3)
Since the neutrinos can diffuse out of the star their chemical potentials are taken to be zero. The number of chemical potentials necessary for the description of the QP in weak equilibrium (the number of components) is therefore reduced to two independent ones. We choose the pair ($`\mu _n`$, $`\mu _e`$) with the neutron chemical potential
$$\mu _n\mu _u+2\mu _d.$$
In a pure QP (in contrast to a QP in a mixed phase, which we will discuss later) we have to require the QP to be charge neutral. This gives us an additional constraint on the chemical potentials
$$\rho _c^{QP}=\frac{2}{3}\rho _u\frac{1}{3}\rho _d\frac{1}{3}\rho _s\rho _e=0.$$
(4)
Here $`\rho _c^{QP}`$ denotes the charge density of the QP and $`\rho _f`$ ($`fu,d,s`$), $`\rho _e`$ the particle densities of the quarks and the electrons, respectively. The EOS can now be parametrized by only one chemical potential, say $`\mu _n`$. At this point it should be noted that the arguments given here for the QP also holds for the HP. There one also ends up with two independent chemical potentials (e.g. $`\mu _n`$, $`\mu _e`$) if one only requires weak equilibrium between the constituents of the HP, and with one chemical potential (e.g. $`\mu _n`$) if one additionally requires charge neutrality. As we will discuss later, the number of independent chemical potentials plays a crucial role in the formulation of the Gibbs condition for chemical and mechanical equilibrium between the HP and the QP.
To calculate the EOS of the QP we apply the “effective mass bag model” . In this model, medium effects are taken into account in the framework of the MIT bag model by introducing density-dependent effective quark masses.
The effective quark masses follow from the poles of the resummed one-loop quark propagator at finite chemical potential (see Fig. 1). Details of the calculation of the effective quark masses in the so called hard dense loop approximation can be found in . One finds
$$m_f^{}(\mu _f)=\frac{m_f}{2}+\sqrt{\frac{m_f^2}{4}+\frac{g^2\mu _f^2}{6\pi ^2}},$$
(5)
with the current quark mass $`m_f`$ of the quark flavor $`f`$ and the strong coupling constant $`g`$. These effective quark masses show the behavior of increasing mass with increasing density (chemical potential). For the current quark masses we choose $`m_u=m_d=0`$ and $`m_s=150`$MeV. The effective masses (5) can now be used in the Fermi gas expression for the particle density
$$\rho _f(\mu _f)=\frac{d_f}{6\pi ^2}\left[\mu _f^2m_{f}^{}{}_{}{}^{2}(\mu _f)\right]^{3/2},$$
(6)
where $`d_f`$ is the degree of degeneracy . The pressure $`p_f`$ and energy density $`ϵ_f`$ can be extracted from the thermodynamic relations
$$\rho _f(\mu _f)=\frac{dp_f(\mu _f)}{d\mu _f}\text{and}ϵ_f(\mu _f)+p_f(\mu _f)=\mu _f\rho _f(\mu _f).$$
(7)
If $`g0`$ the effective quark masses (5) approach the current quark mass $`m_f`$. In this limit, the EOS equals to a free MIT bag Fermi gas EOS. We therefore can use $`g`$ to study the deviations from the free Fermi gas owing to the influence of medium effects. For this purpose we take $`g`$ as a parameter ranging from $`g=0`$ (no medium effects) to $`g23`$ ($`\alpha _s0.310.72`$). (It was found in that a phase transition to a pure QP does not occur inside a typical neutron star for $`g\genfrac{}{}{0pt}{}{>}{}2`$. Larger values are therefore of minor interest for our purpose of studying the influence of quark phases on the properties of neutron stars.) The overall energy density $`ϵ_{QP}`$ and pressure $`p_{QP}`$ of the QP is the sum over all flavors plus the Fermi gas contribution $`ϵ_e`$, $`p_e`$ of the uniform background of electrons plus the contribution of the phenomenological bag constant $`B`$.
$`ϵ_{QP}`$ $`=`$ $`ϵ_u+ϵ_d+ϵ_s+ϵ_e+B,`$ (8)
$`p_{QP}`$ $`=`$ $`p_u+p_d+p_s+p_eB,`$
with
$$ϵ_e=\frac{\mu _e^4}{4\pi ^2}\text{and}p_e=\frac{\mu _e^4}{12\pi ^2}.$$
(9)
The bag constant $`B`$ is introduced in the usual way by adding it to the energy density and subtracting it from the pressure. Due to the phenomenological nature of the bag model, an exact value of $`B`$ is not known. Therefore we will take $`B`$ as a free parameter ranging from $`B^{1/4}=165`$MeV ($`B96`$MeV/fm<sup>3</sup>) to $`B^{1/4}200`$MeV ($`B208`$MeV/fm<sup>3</sup>). The lower bound of the bag constant is reasonable if one requires that the deconfinement phase transition to the QP should not occur at densities below $`ϵϵ_0`$. We do not consider here the possibility of the existence of pure QP stars, so called “strange stars” or the related case of an absolutely stable QP (stable strange quark matter) with respect to the HP as it was supposed by Witten . These possibilities are in detail discussed in the literature and we refer to for an overview. To find the QP to be absolutely stable, their energy per baryon $`E/A`$ in equilibrium ($`p_{QP}=0`$) must be lower than the energy per baryon of <sup>56</sup>Fe which is about $`930`$MeV. This requires small bag constants of $`B^{1/4}\genfrac{}{}{0pt}{}{<}{}155`$MeV (if we neglect medium effects). At our lower bound of the bag constant of $`B^{1/4}=165`$MeV we obtain an energy per baryon of $`E/A990`$MeV which is about $`60`$MeV unbound with respect to <sup>56</sup>Fe. As one can imagine from (8), the bag constant plays a crucial role in the question whether or not a QP can exist in the interior of neutron stars.
## 3 Phase transition
Using the four EOS of the HP discussed in section 2.1 and 2.2 and the EOS of the QP discussed in the last section (2.3), we study the deconfinement phase transition from the HP to the QP and calculate the corresponding total EOS. In particular we are interested in the influence of the model parameters of the QP (i.e. the bag constant $`B`$ and coupling constant $`g`$) on the phase transition densities. (The existence of a QP inside the neutron star clearly requires the transition density to be smaller than the central density of the star.) Before we come to this, we have first to discuss how to construct a first-order phase transition from the confined HP to the deconfined QP.
### 3.1 Construction of the phase transition
We calculate the phase transition according to Glendenning who has first realized the possibility of the occurrence of a mixed phase (MP) of hadronic and quark matter in a finite density range inside neutron stars . For simplicity we neglect Coulomb and surface effects in the MP as studied in Ref. . The essential point in the calculation of the MP is that total charge neutrality can be achieved by a positively charged amount of hadronic matter and a negatively charged amount of quark matter. As already discussed in Sec. 2.3 we have to deal with two independent chemical potentials ($`\mu _n`$, $`\mu _e`$) if we impose the condition of weak equilibrium. Such a system is called a two-component system. The Gibbs condition for mechanical and chemical equilibrium at zero temperature between the both phases reads
$$p_{HP}(\mu _n,\mu _e)=p_{QP}(\mu _n,\mu _e).$$
(10)
Using Eq. (10) we can calculate the equilibrium chemical potentials of the MP where $`p_{HP}=p_{QP}p_{MP}`$ holds. Fig. 2 illustrates this calculation for a specific choice of the model parameters (see figure caption).
The HP$``$MP phase transition takes place if the pressure of the charge neutral HP (white line) meets the pressure surface of the QP. This happens at point 1 in Fig. 2. Up to this point the pressure of the QP is below the pressure of the HP, making the HP the physically realized one. Up to point 2 the phase follows the MP curve which is given by the Gibbs condition (10). At point 2 the MP curve meets the charge neutral QP curve (white line) and the pressure of the QP is above the pressure of the HP, making the QP the physically realized one. The resulting chemical potentials are again shown in Fig. 3.
Below the charge neutral HP curve and above the charge neutral QP curve the HP is positively charged ($`\rho _c^{HP}>0`$, $`\rho _c`$ denotes the charge density) and the QP is negatively charged ($`\rho _c^{QP}<0`$). Therefore, the charge of hadronic matter can be neutralized in the MP by an appropriate amount of quark matter. Up to the critical chemical potential $`\mu _n^{[1]}`$ the electron chemical potential of the charge neutral HP is increasing. (This is mainly because electrons have to neutralize the increasing amount of protons appearing in the HP.) At $`\mu _n^{[1]}`$ the phase transition to the MP takes place and $`\mu _e`$ can be reduced due to the first appearance of negatively charged quark phase droplets immersed in the hadronic phase. (For a detailed discussion of the geometrical structure of the MP see .) For every point on the MP curve one has to calculate the volume proportion
$$\chi =\frac{V_{QP}}{V_{QP}+V_{HP}}$$
(11)
occupied by the quark phase by imposing the condition of global charge neutrality in the MP
$$\chi \rho _c^{QP}+(1\chi )\rho _c^{HP}=0.$$
(12)
From this, the energy density $`ϵ_{MP}`$ of the MP follows as
$$ϵ_{MP}=\chi ϵ_{QP}+(1\chi )ϵ_{HP}.$$
(13)
In the MP the volume proportion of the quark phase is monotonically increasing from $`\chi =0`$ at $`\mu _n^{[1]}`$ to $`\chi =1`$ at $`\mu _n^{[2]}`$ where the transition into the pure QP takes place. Taking the charge neutral EOS of the HP (Sec. 2.2) for $`\mu \mu _n^{[1]}`$, Eq. (10), (12) and (13) for the MP for $`\mu _n^{[1]}<\mu _n<\mu _n^{[2]}`$ and the charge neutral EOS of the QP (Sec. 2.3) for $`\mu _n\mu _n^{[2]}`$ we can construct the full EOS in the form $`p=p(ϵ)`$. For simplicity we denote the complete EOS as the hybrid star EOS.
As one realizes from Fig. 2, the phase transition construction in a two-component system leads to a monotonically increasing pressure of the MP with increasing density . This is quite different from the constant pressure in a one-component system (with only one independent chemical potential) which neutron star matter is occasionally assumed to be. For the existence of a MP inside the star an increasing pressure is crucial. In turn, such a MP allows for a softer EOS over a wide density range which would be excluded in the one-component treatment. We will discuss this point in more detail in the appendix. There we show that a simplified treatment of the phase transition with only one independent component results in an overestimation of the phase transition density.
### 3.2 Results for the equation of state
In this section we present the results of the phase transition construction discussed in the last section. We study in particular the influence of different hadronic EOS and the influence of the model parameters of the QP on the properties of the phase transition. Fig. 4 shows the low density range of the hybrid star EOS in the form $`p=p(ϵ)`$ for the four hadronic EOS and two coupling constants of the QP.
The bag constant is chosen to be $`B^{1/4}=170`$MeV. The MP part of the EOS is shaded gray. As already noted, even in the MP the pressure is monotonically increasing with increasing density. For that reason the MP does exist in the interior of the neutron star if its central energy density exceeds the HP$``$MP transition density. One can see that the HP$``$MP transition density depends only slightly on the choice of the hadronic EOS. This reflects the rather small uncertainty in the hadronic EOS in the density range up to about $`1.5ϵ_0`$ where hyperonic degrees of freedom are still not present. (In the HP the hyperons typically appear at $`23ϵ_0`$ .) The HP$``$MP transition density also depends only weakly on the influence of medium effects which shift the transition densities to slightly higher values. This was already found in . Obviously the EOS gets softer compared to the HP due to the onset of the phase transition. (By “softer” we denote a smaller pressure at a fixed energy density $`ϵ`$). In Fig. 5 we show the same hybrid star EOS in the high density region.
In contrast to the HP$``$MP transition density we see that the MP$``$QP transition density is quite sensitive to the choice of the hadronic EOS as well as to the influence of medium effects in the QP. Assuming e.g. a star with a central energy density of $`ϵ=5ϵ_0`$ we see that no conclusive statement can be made about the composition of the center of the star. Neglecting medium effects ($`g=0`$), TM1 and TM2 predict a QP core (QC) while GPS and GL85 predict a MP core (MC). Taking medium effects into account ($`g=2`$) only TM2 predicts a QC. The strong sensitivity of the transition densities can qualitatively be understood if we look again at Fig. 2. In the high pressure region where the MP$``$QP transition takes place, the HP and the QP pressure surfaces are nearly parallel. Therefore a small change in one EOS is able to produce considerably large changes in the MP$``$QP transition density. Furthermore one should note in Fig. 5 that (despite a change in the phase transition densities) the influence of medium effects on the pure QP EOS in the form $`p=p(ϵ)`$ is negligible. This was already found in .
Up to now we have neglected the influence of the bag constant $`B`$ which enters in the QP EOS. Since $`B`$ enters negatively into the QP pressure in Eq. (8) a larger $`B`$ will shift the pressure surface $`p_{QP}(\mu _n,\mu _e)`$ in Fig. 2 down which moves the MP intersection curve to higher pressures. This finally also shifts the transition densities to higher values. Fig. 6 and Fig. 7 demonstrate this behavior for different hybrid star EOS using GPS and TM2 for the HP (the other hybrid star EOS using TM1 and GL85 show a very similar behavior).
In these figures the gray shaded area corresponds to the density region of the MP for $`g=0`$ (i.e. no medium effects). Obviously the HP$``$MP and the MP$``$QP transition densities are increasing with increasing $`B`$ and increasing $`g`$ where the MP$``$QP transition density again is more sensitive. An increase of $`B^{1/4}`$ by only $`5`$MeV is able to increase the MP$``$QP transition density by about one $`ϵ_0`$. From Fig. 6 it is furthermore interesting to note that the HP$``$MP transition density seems to be only slightly sensitive to a change of the model parameters as long as its density is below $`ϵ2ϵ_0`$ and strongly sensitive above (comparable to the MP$``$QP transition density). This effect seems to be related to the appearance of hyperons (mainly $`\mathrm{\Sigma }^{}`$ and $`\mathrm{\Lambda }`$) in the HP EOS at densities of $`ϵ23ϵ_0`$ which has already been noted in .
It is important to note that the increase of the transition densities with increasing $`B`$ does not necessarily disfavor the existence of deconfined phases inside the star since also the central energy density can be increased by changing this parameter. This is illustrated by the thick lines in Fig. 6 and Fig. 7 which show the central energy density of a star at fixed mass as a function of the bag constant. (The calculation of neutron star masses is discussed in the next section. In Fig. 7 we choose a mass of $`M=1.3M_{}`$ instead of the “canonical” value of $`M=1.4M_{}`$ since it turns out that for some ($`B`$, $`g`$) parameters the maximum mass ($`M_{\mathrm{max}}`$) is slightly below $`1.4M_{}`$.) At least for $`g=0`$ and $`g=1`$ the central energy density is increasing with $`B`$ for low values. From these figures we can already see which kind of phase (HP, MP or QP) is realized in the center of the star. E.g. from Fig. 6 we find that at $`g=0`$ the central energy density (thick solid line) of a $`M=1.4M_{}`$ star is above the MP$``$QP transition density for bag constants below $`B^{1/4}180`$MeV. In this parameter range the neutron star possesses a pure QP core (QC). At higher $`B`$ up to $`B^{1/4}195`$MeV the star possesses a MP core (MC). At still higher $`B`$ the star is purely hadronic (HC). We will come back to a discussion of the ($`B`$, $`g`$) parameter range leading to one of the possibilities (QC, MC, HC) in the next section.
In the following we first want to focus in some more detail on the interesting behavior of the central energy density with increasing $`B`$. The central energy density of a star at fixed mass depends on the softness or stiffness of the underlying EOS. Soft EOS possess a larger central energy density as compared to stiff EOS. In a simple way one could imagine that stars corresponding to soft EOS are compressed to smaller radii and therefore possess a larger central density. In our case (e.g. Fig. 6, thick line $`g=0`$) two competing effects are important. By increasing $`B`$, i) the low density range of the hybrid star EOS gets stiffer due to an enlarged density range of the HP (which is stiffer than the MP) while ii) the density range of the QP gets softened directly by a larger bag constant. Effect ii) dominates for smaller $`B`$ since large parts of the star are made of the QP. Therefore the central energy density is increasing with $`B`$. In both figures this holds up to about $`B^{1/4}175`$MeV for $`g=0`$. We will see in the next section that in this range the internal structure of the star is only slightly sensitive to a change of $`B`$ because both the MP$``$QP transition density and the central energy density are equally increasing with $`B`$. The situation is reversed for bigger $`B`$ and the central energy density is decreasing towards the value of the pure hadronic star. This behavior demonstrates two things. Firstly, the central energy density depends quite sensitively on the softness or stiffness of the EOS which effect the composition of neutron star matter. Thus it is not possible to talk about a “typical central energy density” of a neutron star without having a particular EOS in mind. Secondly, the increase of the bag constant – which makes the QP energetically less favorable – does not automatically disfavor the existence of a QP inside the star since also the central energy density is increased. Indeed, there is a region of the bag constant ($`B^{1/4}\genfrac{}{}{0pt}{}{<}{}175`$MeV) where the internal structure of the star depends only slightly on it. We will come back to this point below. We furthermore should note that the discontinuous fall of the central energy density in Fig. 7 for $`g=0`$ at $`B^{1/4}180`$MeV and for $`g=1`$ at slightly lower $`B`$ is related to the existence of “neutron star twins” which we will discuss in Sec. 5.
## 4 Neutron star structure
With the evaluated hybrid star EOS presented in the last section we now turn to analyse the structure of the corresponding non-rotating neutron stars by solving the Tolman-Oppenheimer-Volkoff (TOV) equations . These equations describe the balance between the gravitational force and the pressure given by the Fermi pressure of the particular EOS. This leads to a relation between the mass $`M`$ and the radius $`R`$ of the neutron star in general relativistic hydrostatic equilibrium.
### 4.1 Mass-radius relations
Fig. 8 and Fig. 9 show the mass-radius (MR) relations for two hybrid star EOS (GPS and GL85) and several bag constants and coupling constants of the QP. The MR-relation denoted by H corresponds to the solution for the pure hadronic star (without any phase transition). The central energy density $`ϵ_c`$ of the stars is increasing from large radii and small masses (down to the right) to the upper left edge where the critical central energy density $`ϵ_c^{crit}`$ is reached at the maximum mass and the minimal radius of the star. From the two differently shaded regions denoted by QC (QP core) and MC (MP core) one sees whether $`ϵ_c`$ is large enough to possess a QC or a MC in the center of the star. For small bag constants ($`B^{1/4}\genfrac{}{}{0pt}{}{<}{}180`$MeV) and small coupling constants (lower curves) $`ϵ_c`$ is sufficient to exceed the MP$``$QP transition density before $`ϵ_c^{crit}`$ is reached (cf. Fig. 5). This leads to a pure QC inside the star. For a more detailed discussion of MR-relations and in particular of the influence of medium effects on the MR-relation see .
It is now our major aim to see what the effect of a pure QC on the mass and the radius of a neutron star is. Concerning the mass of the star we find that $`M_{\mathrm{max}}`$ of a pure hadronic star (H) is reduced by about $`0.10.3M_{}`$ due to the phase transition and the corresponding softening of the EOS. Nevertheless, both hybrid star EOS are able to explain a typical neutron star mass of about $`1.4M_{}`$. The most compact objects seem to have radii of about $`910`$km. In our model calculations these radii are only reached in stars possessing a pure QC (especially for $`g0`$). Very similar results are found for the two other hybrid star EOS (TM1 and TM2) which are not shown here. If we furthermore compare the radii of stars located in the QC region with the radii of the pure hadronic stars (H) of the same mass we can see that the QC stars are about $`24`$km smaller than the corresponding hadronic stars. Hence, a neutron star is about $`2030\%`$ more compact due to the presence of a pure QP in its center.
We define the compactness of a neutron star possessing a QC as
$$\text{compactness(M)}\frac{R_H(M)R_{QC}(M)}{R_H(M)},$$
(14)
with the radius $`R_{QC}`$ of the QC star and the radius $`R_H`$ of the pure hadronic star of the same mass (cf. Figs. 8, 9). The compactness therefore corresponds to the ratio to which the radius of a pure hadronic star can be reduced due to the presence of a pure QP. The shaded areas in Fig. 10 come from calculating the compactness for all stars located in the QC region of the corresponding MR-relation (cf. Figs. 8, 9) for different bag constants (which are depicted in Fig. 10 by different shadings). It is remarkable that the compactness of $`0.20.3`$ depend only slightly on the particular HP EOS and on the model parameters applied in the QP. Especially an increase of the bag constant $`B`$ (darker shading) only narrows the mass range leading to a QC. At the same time the range of the compactness is only slightly altered. A reduction of the radius of an hadronic star by $`2030\%`$ due to the presence of a pure QP seems to be a typical value. Only for $`g0`$ (which in general leads to the most compact stars) and the TM2 hybrid star EOS a radius reduction of about $`40\%`$ seems possible (cf. Fig. 10). This might be related to a rather large incompressibility $`K`$, a small saturation density $`\rho _0`$, and a small effective mass $`m_N^{}`$ of the TM2 EOS compared to the other EOS (cf. Tab. 1).
### 4.2 Internal structure
In this section we further study how the internal structure of the star e.g. the radius of the quark core (QC) or the thickness of the MP depends on the model parameters of the QP. Fig. 11 and Fig. 12 show the neutron star cross sections for a range of bag constants $`B`$ and a fixed neutron star mass.
The respective left panels show the results without medium effects ($`g=0`$) while the right panels include medium effects in the QP. The three differently shaded regions QP, MP and HP again correspond to the parts of the star made of quark phase, mixed phase and hadronic phase, respectively. The thick line to the right of the HP marks the surface of the star. It is obvious that basically only the presence of a pure QP is able to reduce the radius of the star significantly. For $`g=0`$ (left panel) we can see in both figures that only for $`B^{1/4}\genfrac{}{}{0pt}{}{<}{}180`$MeV a pure QC can exist. For $`B^{1/4}\genfrac{}{}{0pt}{}{<}{}175`$MeV the radius of the QC can be as large as $`47`$km at an overall radius of about $`10`$km. For these values of $`B`$ we can see that the internal structure of the star (especially the overall radius) only slightly depends on a change of $`B`$. This behavior can be traced back to the findings of Sec. 3.2. There we found that despite of an increase of $`B`$ from $`B^{1/4}=165`$MeV to $`175`$MeV (which increases the MP$``$QP transition density in the GPS case by about $`3ϵ_0`$) a pure QP is not excluded due to a similar increase of the central energy density $`ϵ_c`$ (cf. Figs. 6 and 7).
On the right panels we can see how the influence of medium effects change the internal structure of the star. Since the HP$``$MP and MP$``$QP transition densities at fixed $`B`$ increase with increasing $`g`$ (cf. Fig. 6 and 7) medium effects are able to transform a star with MC into an pure hadronic star (e.g. Fig. 11, $`B^{1/4}=190`$MeV) or a star with QC to one with a MC (e.g. Fig. 11, $`B^{1/4}=175`$MeV). Only a lower bag constant can compensate for this effect to a certain degree. The solutions for the TM1 and GL85 hybrid star EOS (which are not shown here) lead to very similar results.
Finally, we illustrate the complete ($`B`$, $`g`$) parameter range in Fig. 13 and 14 to see which values lead to which kind of internal structure (QC, MC or HC (pure hadronic star)).
Again we can see that - if we want to keep the internal structure of the star nearly fixed - the influence of medium effects can only be compensated by lowering the bag constant. The dashed lines show the neutron star radii in terms of the radius of the pure hadronic star. These lines correspond to lines of constant compactness (14). In both figures the borderline between the QC and the MC region appears at about $`80\%`$ (corresponding to a compactness of $`0.2`$). This demonstrates once more that - almost independently of the model parameters of the QP - a neutron star with a MC can at most be about $`20\%`$ more compact than a hadronic star of the same mass while a star with a QC is typically $`2030\%`$ more compact than a hadronic star.
## 5 Phase transitions and the third family
### 5.1 Gerlachs criterion
Almost 30 years ago, it was found by Gerlach that a third family of stable equilibrium configurations of compact stars are not forbidden by general relativity. Such a third family could therefore in principle exist besides the two known families of white dwarfs and neutron stars. (See Fig. 15 for a schematic view of the three families in a MR-radius relation. The criteria for stability and instability of the shown ranges will be discussed in Sec. 5.3.).
Gerlach found that a necessary condition for the theoretical possibility of this family is a sufficiently large discontinuity in the speed of sound ($`c_s^2=dp/dϵ`$) of the corresponding EOS<sup>3</sup><sup>3</sup>3This was proved by means of an inversion of the Tolman-Oppenheimer-Volkoff equations. In this context see also the work by Lindblom .. As the density is increased, the speed of sound has to increase abruptly at a density exceeding the central energy density of the neutron star of limiting mass (point C in Fig. 15). Rather smooth EOS (as comprehensively studied by Wheeler et al. ) does not possess the possibility of a third family. Thus a verification of the existence of a third family in nature is equivalent to the verification of a phase transition in the EOS of matter at large densities.
### 5.2 Quark phases and the third family
From Gerlachs criterion it remains however unclear i) which physical mechanism could produce a discontinuity (a phase transition) sufficient to support the stability of a third family and ii) whether a formation process exists in nature to physically realize these objects. Concerning i) it was recently found by Glendenning and Kettner that certain EOS that describe a first-order phase transition to deconfined quark matter (like the EOS also considered in this work) could indeed lead to a third family. Due to partially overlapping mass regions of the neutron star branch and the branch of the third family it is possible that non-identical stars of the same mass can exist. Such pairs are refered to as “neutron star twins” . In such kind of EOS the necessary discontinuity in the speed of sound is produced by the MP$``$QP transition of the deconfinement phase transition. As can be seen in our model from Fig. 5, especially the TM2 hybrid star EOS possesses a large discontinuity due to a rather low speed of sound at the high density end of the MP. In a particular parameter range of the bag constant we will see that the TM2 hybrid star EOS indeed allows for the existence of a third family<sup>4</sup><sup>4</sup>4For the other hybrid star EOS we have checked more than 200 MR-relations following from different combinations of $`B`$ and $`g`$ without finding any further solutions including a third family. Since the parameter range for a third family seems to be quite small we might however ignored some solutions.. Fig. 16 shows MR-relations using the TM2 hybrid star EOS. They are shown for six different bag constants $`B^{1/4}`$ from $`175`$ to $`183`$MeV.
We furthermore assume $`g=0`$ (no medium effects in the QP). The solutions for $`B^{1/4}=175`$MeV and $`B^{1/4}=183`$MeV do not lead to a third family and thus are similar to QC and MC solutions found in Figs. 8 and 9. Increasing the central density beyond the critical point does not restore stability (corresponding to range C-H-I in Fig. 15). This is not the case for bag constants in the small intermediate range $`176`$MeV $`<B^{1/4}182`$MeV. There the mass-radius relations are splitted by an instable (dotted) range which corresponds to C-D in Fig. 15. This enables the possibility of neutron star twins as shown e.g. for the low-density and high-density twin of mass $`M=1.36M_{}`$. While the low-density twin (on the neutron star branch) has a MC the high-density twin of the third family is more compact and possesses a QC. This is a general characteristic of neutron star twins . The neutron star branch terminates in the MP owing to a small adiabatic index $`\mathrm{\Gamma }=(ϵ+p)/pdp/dϵ`$ (cf. Fig. 5) . Thus, all stars on the neutron star branch are MC stars (or HC stars at lower $`ϵ_c`$). Only the larger adiabatic index of the QP can restore stability and therefore enables the QC stars of the third family. The internal structure of the neutron star twins with $`M=1.36M_{}`$ is schematically shown in Fig. 17.
In this section we have seen, that from the deconfinement phase transition described in our model a sufficiently large discontinuity in the speed of sound arises to allow a third family of compact stars. This possibility was first realized by Glendenning and Kettner . Despite the open questions concerning the existence of a formation process, we now want to ask for the theoretical possibility of identifying a third family by means of mass and radius measurements of neutron stars. To do so we have first to discuss the stability criteria which ultimately separate the third family branch of a MR-relation (range D-E in Fig. 15) from the neutron star branch (range B-C) in the same way as neutron stars are separated from white dwarfs.
### 5.3 Stability criteria
While hydrostatic stability of stellar configurations is assured by means of the Tolman-Oppenheimer-Volkoff equations a prove of dynamical stability requires an additional analysis of the radial vibration modes (acoustical modes) of the star. Due to dynamical instabilities no stable stars can have central densities in the range of about $`10^910^{14}`$g/cm<sup>3</sup>. This separates the family of the white dwarfs from the neutron star family. A configuration is stable against small perturbations around hydrostatic equilibrium if and only if the squared frequency eigenvalue $`\omega _0^2`$ of the fundamental vibrational mode is positive . Dynamical stable and instable parts of a schematic MR-relation are shown in Fig. 15. The letters A, B, …, I refer to critical points (turning points) where $`dM/dϵ_c=0`$ holds. The central energy density $`ϵ_c`$ is rising from right to left along the curve A-C-I and A-C-G respectively. Reaching a critical point one radial vibrational mode (characterized by its number of nodes $`n`$) has to change stability. This means that the corresponding squared frequency eigenvalue $`\omega _n^2`$ changes its sign . (A stable mode requires a positive squared frequency $`\omega _n^2>0`$). At critical points where $`dR/dϵ_c<0`$ holds (all points except H and F) an even mode changes stability while for $`dR/dϵ_c>0`$ (points H and F) an odd mode changes stability . Starting from a stable neutron star located in the range B-C (all $`\omega _n^2>0`$) we reach the critical point C by increasing $`ϵ_c`$. At point C an even mode changes stability due to $`dR/dϵ_c<0`$ and therefore gets instable. Since
$$\omega _0^2<\omega _1^2<\omega _2^2<\mathrm{}<\omega _n^2$$
(15)
holds this only can be the fundamental $`n=0`$ mode. A perturbation would cause such instable star to explode or collapse to a black hole. The only way to recover stability and therefore to make a third family possible at larger density is a further change of stability of the even $`n=0`$ mode at point D. (Note that point H can only change an odd mode). Due to (15) no even mode expect $`n=0`$ can change stability at this critical point since all higher modes are bound by $`\omega _1^2`$ which is still positive. For that reason the region D-E corresponding to the third family is again stable against radial vibrations. Without a sufficiently large discontinuity in the speed of sound (i.e. for any reasonably smooth EOS) the mass-radius relation follows the curve C-H-I . Increasing the central density then leads to a successive excitation of more and more unstable modes. (Only modes up to $`n=2`$ are shown in Fig. 15).
As a important consequence of the stability analysis above, the mass and radius of point C has to be larger than the mass and radius of point D. Therefore it is possible that stars of the neutron star branch can in principle exist which are larger and heavier than stars of the third family.
### 5.4 The third family as a possible signature for a phase transition
Within our model we see from Fig. 16 that the masses and radii of the third family are comparable to the ones of the neighbouring $`B^{1/4}=175`$MeV MR-relation which does not possess a third family. Moreover, for a radius and mass of $`R10`$km and $`M1.35M_{}`$ we can see the interesting behavior that all branches of the third family intersect with the branch ($`B^{1/4}=175`$MeV) of the neutron star family in just one point. Only the existence of such an example (even though found in a particular model) illustrates that the measurement of the mass and the radius of only one star (even if exact) could in general not provide the information to decide whether this star is a neutron star or an object of the third family. But how much points on a MR-relation do we in principle have to know (to measure) to decide whether or not a third family exists? The answer is two<sup>5</sup><sup>5</sup>5For this it is important to note that the two masses and radii refer to two stars located on the same MR-relation. Therefore we can e.g. not refer to two stars with largely different temperature, magnetic field or rotational period.. Comparing two arbritay points on a neutron star MR-relation (e.g. $`B^{1/4}=175`$MeV) we see that the heavier star always possesses the smaller radius. But this must not hold for MR-relations including a third family as discussed in the previous section. Then it is possible that the heavier star possesses the larger radius. Such pair of stars always exists if a third family exists. For that it is necessary that the smaller star is located on the third family branch while the larger star is on the neutron star branch. This is schematically shown in Fig. 18 where we have used the same notation as in Fig. 15.
We will refer to such pair as a rising twin. The idea of a rising twin as a signature for a third family is therefore based on the decrease of the mass with increasing radius of all MR-relations without a third family. This holds for almost all gravitational bound stars in which an increase of mass leads to a decrease of the radius due to the increasing gravitational attraction (see also the Figs. 8, 9). An exception of this is given by self-bound stars like the hypothetical strange stars which show the behavior of increasing radius with increasing mass (for small masses $`MR^3`$ holds) . Also in some hadronic EOS (see e.g. ) mass ranges of slightly increasing radius with increasing mass can exist. However, we can exclude these exceptions by additionally require that $`\mathrm{\Delta }M\mathrm{\Delta }R`$ ($`G=c=1`$) has to hold for a rising twin. While in the case of a third family even $`\mathrm{\Delta }M=0`$ is allowed (the neutron star twin) this is not possible for the exceptions for which typically $`\mathrm{\Delta }M\mathrm{\Delta }R`$ holds in the mass range of typical neutron stars. In the framework of our model (see Fig. 16) the corresponding radius differences can grow as large as $`\mathrm{\Delta }R3`$km at mass differences smaller than $`\mathrm{\Delta }M0.05M_{}0.07`$km ($`M_{}1.48`$km). For all possible rising twins of Fig. 16 $`\mathrm{\Delta }M`$ is more than one order of magnitude smaller than $`\mathrm{\Delta }R`$. The requirement of small mass differences is furthermore not in contradiction to the experimentally known masses of neutron stars which are (at least for the classes of observed stars) in a remarkably narrow range of $`M=1.35\pm 0.04M_{}`$ . In particular the mass differences of the two stars in some double neutron star binaries seems to be quite small<sup>6</sup><sup>6</sup>6 For example the mass difference of the Hulse-Taylor pulsar PSR B1913+16 and its companion is $`\mathrm{\Delta }M0.05M_{}`$ while the difference of PSR B1534+12 and its companion is even of the order of only $`\mathrm{\Delta }M10^3M_{}`$ . Of course the two stars of a rising twin must not necessarily build a binary system..
In applying the stability considerations of Sec. 5.3 we conclude that the detection of a rising twin (by the measurement of masses and radii of two neutron stars) can prove the existence of a third family of compact stars. Gerlachs criterion (Sec. 5.1) shows that this is equivalent to the existence of a phase transition in the EOS at finite density. The deconfinement phase transition (as discussed in this work) is one possible scenario that might explain the hypothetical third family of compact stars (Sec. 5.2).
## 6 Summary and conclusion
We have studied the properties of non-rotating cold neutron stars including the possibility of a phase transition to a deconfined quark phase (QP). To describe the confined hadronic phase (HP) of the star we have applied several hadronic EOS in the framework of the relativistic mean-field model (cf. Sec. 2.2, ). The QP was modeled by an extended MIT bag model (effective mass bag model) which includes medium effects by means of effective medium dependent quark masses (cf. Sec. 2.3, ). The influence of medium effects was parametrized by the strong coupling constant $`g`$. The construction of a first-order phase transition from confined to deconfined matter was performed by taking into account two independent chemical potentials (components) of neutron star matter (cf. Sec. 3, Fig. 2). This construction allows for the existence of a mixed phase (MP) of quark and hadronic matter over a finite range inside the star. We have systematically studied the influence of the model parameters of the QP (bag constant $`B`$, coupling constant $`g`$) on the EOS (see Figs. 4, 5), the phase transition densities (see Figs. 6, 7), the mass-radius relations of the stars (see Figs. 8, 9) and the corresponding internal structure (see Figs. 11, 12). This analysis provides us with the information which parameters lead to neutron stars with pure quark phase core (QC), mixed phase core (MC) or with no phase transition at all (HC). We found that - almost independent of the model parameters of the quark phase and the applied hadronic EOS - the radius of neutron stars possessing a MC can at most be about $`20\%`$ more compact than a pure hadronic star of the same mass. A neutron star possessing a MC would therefore – at least from its radius – be hardly distinguishable from a pure hadronic star. Stars with a QC, however, are typically $`2030\%`$ more compact than a hadronic star. This is depicted in Fig. 10 and Figs. 13, 14. The limitation to the upper value of $`30\%`$ comes from our reasonable requirement that the deconfinement phase transition density should not appear below normal nuclear matter density. This approximately corresponds to $`B^{1/4}\genfrac{}{}{0pt}{}{>}{}165`$MeV (c.f. Figs 6, 7). Within our model, the minimal radius of a neutron star is about $`910`$km at a corresponding mass of $`1.41.5M_{}`$. Such small radii of $`910`$km are only reached in neutron stars with a pure QC (as large as $`47`$km) and could not be explained using the pure hadronic models applied here. On the other hand, our models fail to explain radii of $`R<9`$km. Although some radius estimates of neutron stars suggest quite small values , the experimental limits (e.g. of distance measurements) and the uncertainties in the interpretation of the available data (e.g. by blackbody fits) do currently not allow for a definite confirmation. Nevertheless, a future experimental confirmation of extreme small radii would offer a unique possibility to place new stringent constraints on the EOS at high densities. It is therefore necessary to study the theoretical limits of the compactness of neutron stars with inclusion of different softening mechanisms to disentangle the various scenarios in view of more restrictive future radius estimates.
In Sec. 5 we have discussed the theoretical possibility of a third family of compact stars which might exist besides the two known families of neutron stars and white dwarfs (c.f. Fig. 15). Within our model, we have shown that a deconfinement phase transition can explain the existence of such third family (c.f. Fig. 16). Compared to stars of the neutron star family, the stars of the third family can have similar masses while their radii can be up to about $`3`$km smaller. Without refering to our particular model for the EOS we argue that the availability of mass and radius measurements of only two compact stars can reveal the existence of a third family if the larger star is the (slightly) more massive one. We refer to such pair of stars as a rising twin (c.f. Fig. 18). By means of Gerlachs criterion (Sec. 5.1) the existence of a third family is equivalent to the existence of a phase transition in the EOS at large densities (which not necessarily need to be the deconfinement phase transition). If reliable mass and radius measurements would be available, the detection of a rising twin could therefore serve as a novel signature for a phase transition inside neutron stars.
Acknowledgments: The authors thank S. Leupold for helpful discussions and for reading the manuscript. We thank P.K. Sahu for providing us with the GPS EOS. K.S. acknowledges the correspondence with U.H. Gerlach concerning the third family.
## Appendix
In this appendix we intend to show that a simplified treatment of the phase transition construction like the one which appears in a one-component system (instead of a two-component system which neutron star matter in fact is) leads to an overestimation of the phase transition density. This is worth to point out since within a simplified treatment which occassionaly is utilized for the construction of the phase transition one might exclude the possibility of quark phases inside neutron stars . In a correct two component treatment, however, the onset of a mixed phase (MP) or a quark phase (QP) might already be occured inside the star. Compared to the simplified treatment this results in more compact neutron star configurations due to a larger density range occupied by the soft MP or QP EOS.
The two-component system of neutron star matter in weak equilibrium can be reduced to a one-component one by requiring charge neutrality of the hadronic phase and the quark phase independently (i.e. locally) and not globally in the MP. As already discussed in Sec. 2.3, this reduces the number of independent chemical potentials (components) from two (e.g. $`\mu _n`$, $`\mu _e`$) to one (e.g. $`\mu _n`$). (In the following one argument ($`\mu _n`$) denotes charge neutral properties while two arguments ($`\mu _n,\mu _e`$) denotes charged ones.) The condition of local charge neutrality is of course unphysical in the sense that charge neutrality in the MP can also be achieved by means of the weaker (global) condition Eq. (12). Furthermore such a construction is thermodynamically incorrect if applied to neutron star matter. Whereas the Gibbs condition
$$p_{HP}(\mu _n)=p_{QP}(\mu _n),$$
(16)
of a one-component system does ensure the mechanical equilibrium and the chemical equilibrium of neutrons (and therefore of all uncharged particles) one finds that the chemical equilibrium of the second component ($`\mu _e`$) is in general not fulfilled in the MP i.e. (cf. Fig. 3)
$$\mu _e^{HP}(\mu _n)\mu _e^{QP}(\mu _n).$$
(17)
Consequently, all charged particles of neutron star matter do not fulfill the condition of chemical equilibrium. The Gibbs condition (16) leads to the familiar first-order phase transition with a constant pressure MP. Since we know from the equations of hydrostatic equilibrium - the Tolman-Oppenheimer-Volkoff equations \- that the pressure has to increase if we go deeper into the star, a constant pressure MP is strictly excluded from the star. That is not the case for a two-component phase transition calculation, since the pressure is increasing with density even in the MP (see e.g. Fig. 2).
To show that the one-component treatment leads to an overestimation of the deconfinement phase transition density we go back to Fig. 3 where $`\mu _n^{[1]}`$ marks the critical chemical potential necessary for a HP$``$MP phase transition of the two-component system. All we have to show is that at $`\mu _n^{[1]}`$ the one-component system is still in its HP. In other words
$$p_{HP}(\mu _n^{[1]})>p_{QP}(\mu _n^{[1]})$$
(18)
must hold since at a fixed chemical potential the phase with the larger pressure is the physically realized one. The point 1 in Fig. 2 and Fig. 3 is defined by the phase equilibrium between the charge neutral HP and the charged QP, which reads
$$p_{HP}(\mu _n^{[1]})=p_{QP}(\mu _n^{[1]},\mu _e^{[1]}).\text{(point }\text{1}\text{ in Fig. }\text{2}\text{)}$$
(19)
Now we know that at fixed neutron chemical potential the charge density of the QP is given (in units of $`|e|`$) by
$$\rho _c^{QP}(\mu _n,\mu _e)=\frac{p_{QP}(\mu _n,\mu _e)}{\mu _e}.$$
(20)
Since the QP is negatively charged at point 1 (and everywhere above the charge neutral QP curve in Fig. 3) we find that the right hand side of (20) must also be negative. Therefore, at fixed $`\mu _n`$ (say $`\mu _n^{[1]}`$) the pressure is decreasing if we decrease $`\mu _e`$ from $`\mu _e^{[1]}`$ to the lower value at which the QP is charge neutral. This means
$$p_{QP}(\mu _n^{[1]},\mu _e^{[1]})>p_{QP}(\mu _n^{[1]}).$$
(21)
Putting (19) and (21) together, we finally obtain (18). This shows that due to the treatment of the phase transition as a simplified transition of a one-component system, the transition densities are shifted to higher values as compared to the ones expected from the correct treatment of neutron star matter as a two-component system. |
warning/0001/nlin0001006.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The Maxwell-Bloch (MB) system of equations have been fundamental to much of theoretical quantum optics and nonlinear optics since they were first introduced in the late 1960’s (some history is in and also in ). These MB systems are of abiding theoretical interest. A feature is that their complete integrability is handed down from the ’reduced MB’ or (RMB) equations to the envelope MB (or self-induced transparency (SIT)) equations, thence, at resonance, to the Sine-Gordon equation (cf. e.g. ). Each member of this hierarchy has important physical applications while the physics of SIT in particular remains a very active field of current research , even into the femto-second pulse regime.
Our recent paper followed up ideas of quantum groups and their relevance to integrable systems theory and derived a $`q`$-deformed lattice version of the envelope MB system together with its zero-curvature representation: in continuum limit these lattice equations become the resonant envelope MB (or SIT) equations. In this paper we now report exact $`N`$-soliton solutions of this $`q`$-deformed dynamical system. Solitons of the lattice equations were promised in , and a Riemann-Hilbert method of solution sketched. But for the pure $`N`$-soliton solution reported in this paper it is more convenient to use a variant of the Darboux-Bäcklund dressing method which (see below) makes an ansatz for the dressing in terms of appropriate $`N`$ bound states eigenvalues.
Historically multi-soliton solutions of the SIT equations were found by the method which become Hirota’s method ; Lamb gave the inverse scattering solutions; the inverse method for the RMB equations was in and obtains the multi-soliton solutions for the SIT equations from it; gave a further account of inverse scattering for these SIT equations. These several results on inverse scattering confirmed the generality of a method first devised to solve the Korteweg-de Vries equation .
Expressed in terms of the complex slowly varying envelopes for the electric field and polarisation $`\epsilon `$, $`\rho `$ and the real inversion $`𝒩`$, the SIT equations can be put in the form (e.g. ).
$$\begin{array}{ccc}_\xi \epsilon =\rho ,& & \\ _\tau \rho +2i\eta \rho =𝒩\epsilon ,& & \\ _\tau 𝒩=\frac{1}{2}(\epsilon ^{}\rho +\epsilon \rho ^{}).& & \end{array}$$
(1)
Propagation is along a coordinate $`z`$ and $`\xi =\mathrm{\Omega }x`$ with $`x=z/c`$. The time $`\tau `$ is a retarded time, $`\tau =\mathrm{\Omega }(tx)`$; $`\eta =(\omega \omega _0)/2\mathrm{\Omega }`$ is the detuning, and $`\mathrm{\Omega }=2\pi n_0\omega _0\mu ^2/\mathrm{}`$ is the coupling constant. The number $`n_0`$ is the density of 2-level atoms with the non-degenerate transition frequency $`\omega _0`$ ($`\text{rad}\text{sec}^1`$); $`\mu `$ is the matrix element for dipole allowed transitions at $`\omega _0`$. The star denotes complex conjugation and $`.=_{\mathrm{}}^{\mathrm{}}h(\eta )(.)d\eta `$ is the average over inhomogeneous broadening: $`h(\eta )`$ is a $`\delta `$-function in the sharp-line limit case .
In we constructed the completely integrable lattice system whose equations of motion for three dynamical variables $`s_n`$, $`H_n`$ and $`\beta _n`$ at each lattice site $`n`$ can be written
$`_t\beta _n`$ $`=`$ $`{\displaystyle \frac{1}{2}}q^{2(N_n+H_n)}(\beta _{n+1}+\beta _n){\displaystyle \frac{i}{2}}q^{2N_n}(s_n+s_{n1})`$
$`_ts_n`$ $`=`$ $`{\displaystyle \frac{i}{2}}(\beta _n+\beta _{n+1})(1+2\gamma s_ns_n^{})+{\displaystyle \frac{1}{2}}q^{2(N_n+H_n)}(s_n+s_{n1})`$ (2)
$`_tH_n`$ $`=`$ $`{\displaystyle \frac{i}{2}}(s_niq^{2H_n}\beta _n)(\beta _n^{}+\beta _{n+1}^{}){\displaystyle \frac{i}{2}}(s_n^{}+iq^{2H_n}\beta _n^{})(\beta _n+\beta _{n+1}).`$
Here $`q^{2N_n}=1+2\gamma \beta _n^{}\beta _n`$, $`q=e^\gamma `$ and $`\gamma >0`$, is a real parameter (a coupling constant – see below). Reference to shows that in Eqs. (2) we use $`s_n=\sqrt{2\gamma }\chi _n+iq^{2H_n}\beta _n`$: in the second equation is for $`_t\chi _n`$. As can be checked (and cf. ) when the lattice spacing $`\mathrm{\Delta }0`$ for a continuum limit with
$$\begin{array}{c}tt\mathrm{\Delta }^1,x=n\mathrm{\Delta },\beta _n=\sqrt{\mathrm{\Delta }}(x),\chi _n=\mathrm{\Delta }S(x),H_n=\mathrm{\Delta }S^3(x)\\ \gamma =\kappa \mathrm{\Delta }/2,\kappa >0.\end{array}$$
(3)
one reaches the resonant sharp-line form of the envelope MB (or SIT) equations Eqs (1) via the definitions
$$\epsilon (\xi ,\tau )=2(x,t),\rho (\xi ,\tau )=2iS(x,t),N(\xi ,\tau )=2S^3(x,t),$$
(4)
with $`\mathrm{\Omega }=\sqrt{\kappa }`$. Our use of ’lattice Maxwell-Bloch system(LMB) equations’ for Eqs. (2) stems from this fact.
A Hamiltonian for this LMB system is
$`^L`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{n=1}{\overset{M}{}}}\{\sqrt{2\gamma }[\chi _n^{}(\beta _{n+1}+\beta _n)+\chi _n(\beta _{n+1}^{}+\beta _n^{})]`$ (5)
$`+`$ $`iq^{2H_n}(\beta _{n+1}^{}\beta _n\beta _{n+1}\beta _n^{})\}.`$
For $`M<\mathrm{}`$ it would be natural to impose periodic boundary conditions. But we shall look for lattice soliton solutions and here think of $`M\mathrm{}`$ with suitable boundary conditions still to be specified. The Poisson brackets of Eq. (5) are
$$\{X_n^{},X_m\}=i\{2H_n\}\delta _{mn},\{H_n,X_m\}=iX_n\delta _{mn},$$
(6)
and the quantities $`X_n^{}`$, $`X_n`$ and $`H_n`$ form the $`su_q(2)`$ Lie-Poisson algebra for by $`\{\}`$ we mean $`\{x\}=(q^xq^x)/(2\gamma )`$. This algebra has a central element $`X_nX_n^{}+\{H_n\}^2=\{S\}^2`$. The variables $`X_n^{}`$, $`X_n`$ enter (2) via $`\chi _n=q^{H_n}X_n`$ . The variables $`\beta _n`$, $`\beta _n^{}`$ (the ’electric fields’, see Eqs. (3),(4) above) satisfy the Lie-Poisson $`q`$-boson algebra
$$\{\beta _n,\beta _m^{}\}=iq^{2N_n}\delta _{mn},\{N_n,\beta _m\}=i\beta _n\delta _{mn}.$$
(7)
## 2 The $`q`$-deformed solitons
In we obtained the zero-curvature representation of the system (2) which means that we constructed an over-determined linear system for a matrix-function $`\mathrm{\Psi }_n(\zeta ,t)`$ such that
$`\mathrm{\Psi }_{n+1}`$ $`=`$ $`L(\zeta |n)\mathrm{\Psi }_n`$ (8)
$`_t\mathrm{\Psi }_n`$ $`=`$ $`V(\zeta |n)\mathrm{\Psi }_n,`$ (9)
where
$$V(\zeta |n)=\underset{j=2}{\overset{2}{}}\zeta ^jV_j(n),L(\zeta |n)=\frac{q^{N_nH_n}}{2\gamma }\underset{j=2}{\overset{2}{}}\zeta ^jL_j(n).$$
(10)
Here
$$V_0(n)=2i\gamma (\beta _ns_{n1}^{}+\beta _n^{}s_{n1})\sigma ^z,V_{\pm 2}=\frac{1}{4}\sigma ^z,$$
(11)
$$V_{+1}(n)=\frac{\sqrt{2\gamma }}{2}\left(\begin{array}{cc}0& i\beta _n^{}\\ s_{n1}& 0\end{array}\right),V_1(n)=\frac{\sqrt{2\gamma }}{2}\left(\begin{array}{cc}0& s_{n1}^{}\\ i\beta _n& 0\end{array}\right),$$
(12)
while
$$L_0(n)=2i\gamma \left(\begin{array}{cc}\beta _ns_n^{}& 0\\ 0& \beta _n^{}s_n\end{array}\right)q^{2(N_n+H_n)}\sigma ^z,L_{\pm 2}=\frac{1}{2}(\sigma ^z\pm I),$$
(13)
$$L_{+1}(n)=\sqrt{2\gamma }\left(\begin{array}{cc}0& i\beta _n^{}\\ s_n& 0\end{array}\right),L_1(n)=\sqrt{2\gamma }\left(\begin{array}{cc}0& s_n^{}\\ i\beta _n& 0\end{array}\right).$$
(14)
The parameter $`\zeta 𝐂`$ which appears in Eqs. (8)-(10) will be thought of as the spectral parameter while in continuum limit (9) is a spectral problem in $`L`$ in the usual $`2\times 2`$ sense (Zakharov-Shabat linear system ); $`\sigma ^{x,y,z}`$ are the Pauli matrices. The compatibility condition of the two linear systems Eqs. (8),(9) under isospectral condition $`_t\zeta =0`$ is
$$_tL(\zeta |n)+L(\zeta |n)V(\zeta |n)V(\zeta |n+1)L(\zeta |n)=0$$
(15)
and this coincides with Eqs. (2), independent of $`\zeta `$. However, $`\zeta =e^{i\gamma \lambda }`$, $`\lambda 𝐂`$ as it was introduced in ; $`\lambda `$ is a second ’spectral parameter’ and the real axis in the $`\lambda `$-plane is the circle of the unit radius in the $`\zeta `$-plane; $`\lambda `$ is the usual spectral parameter for the equations Eqs. (1) derived in continuum limit. Notice that time $`t`$ is suppressed in Eqs. (8),(9): an explicit time dependence will be indicated only where and when it is needed. Reference to Eqs. (8) and (9) may make plain that the function $`\mathrm{\Psi }_n(\zeta )`$ possesses essential singularities of the rank 2 at $`\zeta =0,\mathrm{}`$. It is also important to notice that the linear equations Eqs. (8) and (9) are invariant under the transformations
$$\mathrm{\Psi }_n(\zeta )(1)^{n1}\sigma ^y\mathrm{\Psi }_n^{}\left(\frac{1}{\zeta ^{}}\right)\sigma ^y,\mathrm{\Psi }_n(\zeta )\sigma ^z\mathrm{\Psi }_n(\zeta )\sigma ^z$$
(16)
We can now turn to the derivation of exact solutions of the LMB system Eqs. (2). For this, as mentioned, we develop a variant of the Darboux-Bäcklund dressing procedure rather then any inverse scattering method . The essence of the dressing procedure is to choose a ’seed’ solution of the system Eqs. (2), typically some trivial solution, and construct from it a new solution associated with additional points $`\zeta _\nu `$, $`\nu =1,\mathrm{},N`$ (say) of the discrete spectrum: thus $`det\mathrm{\Psi }_n(\zeta _\nu ,t)=0`$ for the new solution $`\mathrm{\Psi }_n(\zeta ,t)`$.
For initial and boundary conditions observe that for SIT and envelope MB system Eqs. (1), typical experimental situation is the half-space problem: an initial optical pulse enters, supposedly without reflection from $`x<0`$ into the resonant medium $`x0`$ and here breaks up into background radiation and a sequence of soliton pulses. The corresponding mathematical problem is the Cauchy problem at the point $`x=0`$: $`\epsilon (x,t)|_{x=0}=\epsilon _0(t)`$ together with the asymptotic boundary conditions (in $`t`$) that for $`x>0`$ $`𝒩𝒩_{}`$, $`\rho 0`$ as $`t\mathrm{}`$. For the so-called ’attenuator’ $`N_{}`$ is the ground state $`N_{}=1`$ of the inversion density. For the lattice problem we therefore take the half-space problem in which $`\beta _n(t)`$ and $`s_n(t)`$ are sufficiently decreasing for $`|t|\mathrm{}`$, while $`H_n(t)H`$ such that $`H`$ corresponds to $`N_{}`$. In this way we would look for a solution in the half-space $`n>0`$, for which it becomes the Cauchy problem specified by the conditions
$$\beta _n(t)|_{n=1}=\beta _1(t),s_n(t)|_{n=1}=s_1(t),H_n(t)|_{n=1}=H_1(t).$$
(17)
With this as motivation we report in this paper exact $`N`$-soliton solutions derived by the dressing procedure based on the seed solution
$$\beta _n=0,s_n=0,H_n=H.$$
(18)
the corresponding solution of the linear system Eqs (8),(9) is then
$$\mathrm{\Psi }_n^{(0)}(\zeta ,t)=\mathrm{exp}\left\{\frac{\sigma ^zt}{4}\right\}\left(\zeta ^2\frac{1}{\zeta ^2}\right)\left(\begin{array}{cc}z^n(\zeta )& 0\\ 0& \left(z\left(\frac{1}{\zeta }\right)\right)^n\end{array}\right),$$
(19)
where $`z(\zeta )=\frac{1}{2\gamma }\left(\zeta ^2q^Hq^H\right)`$, while the corresponding operator $`V^{(0)}(\zeta |n,t)`$ has $`V_0^{(0)}=V_{\pm 1}^{(0)}=0`$, and $`V_{\pm 2}^{(0)}=\frac{1}{4}\sigma ^z`$. For the $`N`$-soliton solution of Eqs. (2) we construct the new solution $`\mathrm{\Psi }_n^{(N)}(\zeta )`$ of Eqs. (8),(9) through the ansatz
$$\mathrm{\Psi }_n^{(N)}(\zeta )=F(\zeta )\mathrm{\Psi }_n^{(0)}(\zeta ).$$
(20)
The function $`F(\zeta )`$ is to have poles only at the essential singularities of $`\mathrm{\Psi }_n(\zeta )`$. As was indicated above these points are $`0`$ and $`\mathrm{}`$. This suggests the ansatz
$$F(\zeta ,n,t)=F_0(n,t)+\underset{i=1}{\overset{N}{}}\zeta ^iF_{+i}(n,t)+\zeta ^iF_i(n,t).$$
(21)
It is convenient to impose the additional conditions on $`F(\zeta )`$ that
$$\sigma ^yF^{}\left(\frac{1}{\zeta ^{}}\right)\sigma ^y=F(\zeta ),\sigma ^zF(\zeta )\sigma ^z=(1)^NF(\zeta )$$
(22)
which are obviously compatible with the transformation Eq. (16) We can also normalize the matrix $`F(\zeta )`$ so that for each $`(n,t)`$
$$F_N=𝒬\left(\begin{array}{cc}f(n,t)& 0\\ 0& f^1(n,t)\end{array}\right),F_N=𝒬^{}\left(\begin{array}{cc}f^1(n,t)& 0\\ 0& f(n,t)\end{array}\right)_,$$
(23)
where the constant $`𝒬`$ is independent of $`n`$ and $`t`$ and $`f(n,t)`$ is a real function. We now choose a set of $`N`$ points $`\left\{\zeta _\nu \right\}_{\nu =1}^N`$ where $`det\mathrm{\Psi }_n^{(N)}(\zeta )`$ is to vanish. This means
$$F(\zeta _\nu )\mathrm{\Phi }(\zeta _\nu )=F\left(\frac{1}{\zeta _\nu ^{}}\right)\sigma ^y\mathrm{\Phi }^{}(\zeta _\nu )=0,$$
(24)
where
$$\mathrm{\Phi }(\zeta _\nu )=\left(\begin{array}{c}\mathrm{\Phi }_1(\zeta _\nu )\\ \mathrm{\Phi }_2(\zeta _\nu )\end{array}\right)=\mathrm{\Psi }_n^{(0)}(\zeta _\nu )\left(\begin{array}{c}1\\ c_\nu \end{array}\right)_,$$
and $`c_\nu `$ are constants independent on $`n`$ and $`t`$. The set $`\{\zeta _\nu ,c_\nu \}_{\nu =1}^N`$ together constitute a necessary and complete set of parameters (spectral data) for an $`N`$-soliton solution .
The system of equations Eqs. (24) has a unique solution satisfying conditions Eqs. (22),(23) if we choose
$$𝒬=e^{i\frac{\pi }{2}N+i_{\nu =1}^N\alpha _\nu },\zeta _\nu =e^{\gamma _\nu +i\alpha _\nu },$$
(25)
where $`\gamma _\nu ,\alpha _\nu 𝐑.`$ In so far as $`\zeta =e^{i\gamma \lambda }=e^{\gamma _\nu +i\alpha _\nu }`$ and $`\lambda `$ is the spectral parameter for Eqs. (1) we are interested in zeros $`\zeta _\nu `$ defined by the half $`\lambda `$-plane $`\text{Im}\lambda 0`$ which lie inside the circle $`|\zeta |=1`$ in the $`\zeta `$-plane. The linear system Eqs. (8),(9) is invariant under the gauge transformation Eqs. (20) with the potentials written as
$`F_{N+1}F_N^1`$ $`=`$ $`\sqrt{2\gamma }\left(\begin{array}{cc}0& s_{n1}^{}\\ i\beta _n& 0\end{array}\right)`$ (28)
$`q^H{\displaystyle \frac{f(n+1,t)}{f(n,t)}}`$ $`=`$ $`q^{N_n+H_n}`$ (29)
We turn next to a determination of the matrices $`F_{\pm i}(n,t)`$. The conditions Eqs. (22) suggest that we should take the matrices $`F_{N+2k}`$ diagonal, and the matrices $`F_{N+2k1}`$ off-diagonal in agreement with the first of Eqs. (29).
So will $`F_N^1`$ be diagonal from Eq. (23) we set
$$F_N^1F_{N+2k1}=\left(\begin{array}{cc}0& y_k\\ \stackrel{~}{y}_k& 0\end{array}\right),F_N^1F_{N+2k}=\left(\begin{array}{cc}x_k& 0\\ 0& \stackrel{~}{x}_k\end{array}\right)$$
(30)
in which $`y_k`$, $`\stackrel{~}{y}_k`$, $`x_k`$, $`\stackrel{~}{x}_k`$ are (so far) arbitrary independent complex numbers.
Then the conditions for the zeros $`\zeta _\nu `$ Eqs. (24) mean that we can instead solve
$$(1,0)+\underset{k=1}{\overset{N}{}}z_k\sigma _\nu ^{2k}=0,\nu =1,..,N;$$
(31)
in which the $`z_k`$ are row-vectors $`z_k=(x_k,y_k)`$, and the matrices $`\sigma _\nu =S_\nu \mathrm{\Lambda }_\nu S_\nu ^1`$ in which $`\mathrm{\Lambda }_\nu =\text{diag}(\zeta _\nu ,1/\zeta _\nu ^{})`$; the matrices $`S_\nu `$ are defined as
$$S_\nu =\left(\begin{array}{cc}\mathrm{\Phi }_1(\zeta _\nu )& \mathrm{\Phi }_2^{}(\zeta _\nu )\\ \frac{1}{\zeta _\nu }\mathrm{\Phi }_2(\zeta _\nu )& \zeta _\nu ^{}\mathrm{\Phi }_1^{}(\zeta _\nu )\end{array}\right)$$
(32)
and are determined from Eqs.(24) using the seed solution Eq.(18). In this way the $`N`$-soliton solution of Eqs. (2) is put in the form
$$\beta _n=\frac{i}{\sqrt{2\gamma }}y_N^{}x_N,s_{n1}=\frac{1}{\sqrt{2\gamma }}\frac{𝒬^{}}{𝒬}\frac{y_1^{}}{x_N}$$
(33)
$$q^{2(N_n+H_n)}=q^{2H}\frac{x_N(n+1)}{x_N(n)},$$
(34)
where $`x_N(n,t)`$ etc. is determined from Eqs. (31)
For the one-soliton case, $`N=1`$, we can choose the single point of the discrete spectrum $`\zeta _0=e^{\gamma _0+i\alpha _0}`$ (say) and $`\gamma _0<0`$. We then find the formulae
$$\beta _n(t)=i\sqrt{\frac{2}{\gamma }}\mathrm{sinh}(2\gamma _0)\frac{\mathrm{exp}i(\varphi (n,t)\alpha _0)}{\mathrm{cosh}(\psi (n,t)\gamma _0)},$$
(35)
$$s_{n1}(t)=\sqrt{\frac{2}{\gamma }}\mathrm{sinh}(2\gamma _0)\frac{\mathrm{exp}i(\varphi (n,t)+\alpha _0)}{\mathrm{cosh}(\psi (n,t)+\gamma _0)}$$
(36)
$$q^{2(N_n+H_n)}=q^{2H}\frac{1\mathrm{tanh}(\psi (n,t)\gamma _0)\mathrm{tanh}\vartheta _0}{1\mathrm{tanh}(\psi (n,t)+\gamma _0)\mathrm{tanh}\vartheta _0}.$$
(37)
Here
$$\varphi (n,t)=t\mathrm{cosh}(2\gamma _0)\mathrm{sin}(2\alpha _0)n\varrho _0+\varphi _0,$$
(38)
$$\psi (n,t)=t\mathrm{sinh}(2\gamma _0)\mathrm{cos}(2\alpha _0)n\vartheta _0+\psi _0,$$
(39)
$`\vartheta _0`$ $`=`$ $`{\displaystyle \frac{1}{2}}\mathrm{ln}{\displaystyle \frac{\text{sinh}^2(\gamma _0H\gamma )+\text{sin}^2\alpha _0}{\text{sinh}^2(\gamma _0+H\gamma )+\text{sin}^2\alpha _0}}+2\gamma _0`$ (40)
$`\varrho _0`$ $`=`$ $`\text{arg}{\displaystyle \frac{\mathrm{sinh}(\gamma _0H\gamma +i\alpha _0)}{\mathrm{sinh}(\gamma _0+H\gamma +i\alpha _0)}}+2\alpha _0,`$ (41)
$`\varphi _0`$ and $`\psi _0`$ are arbitrary real constants.
The formulae for various multisoliton solutions for the lattice system Eqs. (2) are too complicated to be presented in detail here. Since for the lattice these solutions depend explicitly on the deformation parameter $`q=e^\gamma `$, these $`N`$-soliton solutions ($`N=1,2,\mathrm{}`$) are naturally thought of as $`q`$-deformed solitons. In continuum limit Eq. (3), $`q1`$ and $`\gamma 0`$ and $`q`$-deformation disappear.
## 3 Conclusions and discussion
As was mentioned above in the case of the real (imaginary) dynamical variables and in the sharp line limit the MB system Eqs. (1) is equivalent to the Sine-Gordon equation. The same procedure is applicable to the LMB system which means that in the case of the reduction to the real (imaginary) dynamical variables the LMB system is in fact a new version of the lattice Sine-Gordon equation. The dressing procedure described in this paper can be extended to this case delivering the whole variety of solutions of the (lattice) S-G equation (solitons, breathers etc.). In so far as Eqs. (35)-(41) form a $`q`$-deformed soliton we can use this result to gain insight into the quantum case. One objective of the investigation of the quantum MB system must be to find out, the precise nature of, and to calculate, the ’quantum soliton’ solutions. In Refs. , we introduced and solved exactly through the quantum inverse method (up to the solutions of the Bethe equations) a quantum version of the LMB system Eqs. (2). Since this model provides a natural and exactly solvable lattice regularization of the continuous limit quantum envelope MB (or SIT) system (and recall that the quantum Sine-Gordon can be embedded in this quantum MB) it is very useful for the construction of the evolution operator and for investigating the quantum dynamics of these continuous models which have direct physical meaning. It is known from a number of quantum models that a ’string solution’ of the Bethe equations for the quantum model corresponds, in the limit of a large number of collective excitations $`M`$, to the soliton solution of the classical counterpart of the exactly solvable quantum system. A plausible conjecture which we will justify elsewhere is that the soliton solution Eq. (35) for the ’electric field’ is given by the matrix element $`lim_M\mathrm{}0|C(\lambda _1)C(\lambda _2)\mathrm{}C(\lambda _M)\beta _n^{}B(\lambda _1)B(\lambda _1)\mathrm{}B(\lambda _{M1})|0`$, where $`B(\lambda )`$ is a creation operator for a quasiparticle and $`C(\lambda )=B^{}(\lambda )`$ is an annihilation operator. The rapidities $`\{\lambda _l\}_{l=1}^M`$ are roots of the Bethe equations $`e^{2i\gamma M\lambda _n}\frac{\text{sin}^M\gamma (\lambda _liS)}{\text{sin}^M\gamma (\lambda _l+iS)}=_{j=1}^N\frac{\mathrm{sin}\gamma (\lambda _l\lambda _ji)}{\mathrm{sin}\gamma (\lambda _l\lambda _ji)}`$.The operator $`\beta _n^{}`$ is the electric field operator which satisfies the $`q`$-deformed $`q`$-boson algebra analogous to the algebra Eq. (7). In Ref. it was shown that the creation operator $`B(\lambda )`$ plays the role of the quantum counterpart of a Blaschke multiplier which builds the classical soliton solution . The dressing operator $`F(\zeta )`$ Eq. (21) up to certain modifications has the same meaning. This indicates very well how the experience obtained in the analysis of the $`c`$-number system reported in this paper can be used in understanding of the quantum case. In practice this experience helps us to trace out the formation of the classical optical soliton from a large number of quantum collective excitations, a physical problem of considerable interest.
## Acknowledgements
One of us (RKB) is grateful to Dr. P.J. Caudrey for helpful discussions of the soliton solutions reported in this paper, GV was partly supported by a Russian Federation research grant RFBR No 98-01-01063 |
warning/0001/quant-ph0001014.html | ar5iv | text | # Separability and Fourier representations of density matrices
## I Introduction
One of the predictions of quantum mechanics is that spatially separated components of a system can be entangled. The consequent prediction of non-classical correlations among the separated components of a quantum system has led to critiques of the foundations of quantum mechanics, as in the famous Einstein, Podolsky, Rosen paper , and to experiments that have confirmed the predicted non-classical correlations, as in . Interest in entangled systems has been heightened by proposed applications in quantum computation, for example , and in quantum communication, as exemplified most dramatically by teleportation . As a result there have been many publications which have examined various aspects of entanglement, its measurement, and its use in quantum communication such as references to mention only a few recent papers.
In this paper we shall be interested in the separability properties of quantum systems in states defined on finite dimensional Hilbert spaces $`H=H_1\mathrm{}H_n`$, where the $`H_k`$ denote the Hilbert spaces of the subsystems. A state specified by a density matrix $`\rho `$ is said to be completely separable on $`H`$ if it is a convex combination of tensor products:
$$\rho =\underset{a}{}p\left(a\right)\rho ^{\left(1\right)}\left(a\right)\mathrm{}\rho ^{\left(n\right)}\left(a\right),$$
(1)
where $`\rho ^{\left(k\right)}\left(a\right)`$ is a density matrix on $`H_k`$. Since the same $`\rho `$ can have different convex representations, it has proven difficult to determine generally applicable operational conditions for separability, and determining such conditions is one of the motivations for this paper. It is also possible to have different types of separability by allowing sets of the subsystems to be entangled, cf. , and one can describe a lattice of levels of separability. The theory we develop here applies to all of these various definitions of separability.
A necessary condition for separability is that the partial transpose $`\rho ^{T_r}`$ of a state $`\rho `$ should be a state . If we represent $`\rho `$ as a matrix, this means that if $`\rho =\left(\rho _{j_1\mathrm{}j_n,k_1\mathrm{}k_n,}\right)`$ then (taking $`r=1`$)
$$\rho ^{T_1}=\left(\rho _{_{k_1j_2\mathrm{}j_n,j_1k_2\mathrm{}k_n,}}\right)$$
(2)
is also a density matrix. It is easy to confirm that if a density matrix is separable, its partial transposes are also separable, but it has been shown that the converse is true only in the $`22`$ and $`23`$ cases. In the proof of this last result , a necessary and sufficient criterion for separability was established, but there seems to be no operational way of using this criterion as a general tool. Other studies of separability, such as those in have found operationally useful necessary conditions and sufficient conditions for classes of densities or for special cases, but no general sufficient conditions with a breadth of applicability analogous to that of the Peres condition.
Broadly speaking, necessary conditions tend to be described in the computational basis while sufficient conditions for $`2`$-level systems tend to be described in terms of the Pauli spin basis. That observation motivated the derivation of a change-of-basis formula in which facilitates the strategy of checking whether necessary conditions derived in the computational basis are sufficient by using the (real) Pauli spin basis. This approach leads to general sufficient conditions for full separability which essentially give the condition in as a corollary and also leads to necessary and sufficient conditions for full separability of a parametrized family of $`n`$-qubit densities which all satisfy the Peres condition. The difficulty with extending this approach to $`d`$-level systems is that the generally accepted definition of spin matrices as generators of rotations does not capture the computationally useful features of the Pauli matrices when $`d3.`$ One of the basic purposes of this paper is to propose a general definition of $`d`$-level spin matrices which possess many of those computational properties.
The Pauli matrices are special in that they are both Hermitian and unitary, and together with the identity matrix $`\sigma _0`$ they form a basis of the set of $`2\times 2`$ matrices. Our strategy is to generalize the role of the Pauli matrices as a basis of unitary matrices at the expense of Hermiticity. We show the applicability of these proposed $`d`$-level spin matrices and in the appendix examine the $`d=3`$ case in some detail, identifying properties analogous to those of the Pauli matrices. We also define directly a general characterization of certain classes of trace one projections of $`d`$–level systems. We use those projections to establish necessary and sufficient conditions for full separability of generalized Werner densities composed of any number $`n`$ of $`d`$–dimensional subsystems for any prime $`d`$. In addition, we establish a general sufficient condition for full or partial separability of densities of any dimension. Analogous results for full separability were obtained in for $`d=2`$ by essentially the same methodology.
Other authors have used a different set of operators in the $`d=3`$ case and some separability results were obtained recently in . Our proposed class is different, and we show that stronger separability results can be obtained using these matrices and the strategy developed in .
## II A Necessary Condition
As mentioned above, the Peres partial transpose condition is a general necessary condition for separability . In a weaker but useful condition was derived using the Cauchy-Schwarz inequality and has the following application. Suppose $`j=j_1\mathrm{}j_n`$ and $`k=k_1\mathrm{}k_n`$ differ in each component: $`j_rk_r`$. Let $`u`$ and $`v`$ be indices with $`u_rv_r`$ and $`\{u_r,v_r\}=`$ $`\{j_r,k_r\}.`$ Then for fully separable states $`\rho `$
$$\left(\sqrt{\rho _{j,j}}\sqrt{\rho _{k,k}}\right)\left|\rho _{u,v}\right|,$$
(3)
where $`\rho `$ is written as a matrix in the computational basis defined by the tensor products of $`|j_ik_i|,1in`$. As an application, consider the following generalization of the Werner density matrix on the $`N=d^n`$ dimensional Hilbert space $`H^{\left[N\right]}`$:
$`W^{\left[N\right]}\left(s\right)={\displaystyle \frac{1s}{d^n}}I+s\tau `$
where $`I`$ is the identity and $`\tau `$ is the projection defined by the state
$$|\psi ^{\left[N\right]}=\frac{1}{\sqrt{d}}\left(|0\mathrm{}0+|1\mathrm{}1+\mathrm{}+|(d1)\mathrm{}(d1)\right).$$
(4)
(In the sequel we let $`\stackrel{~}{k}`$ denote the repeated index $`k\mathrm{}k.`$) In the computational basis $`W_{j,j}^{\left[N\right]}\left(s\right)`$ equals $`\left(\frac{1s}{d^n}+\frac{s}{d}\right)`$ when $`j`$ is in $`\{\stackrel{~}{k}:0k<d\}`$ and equals $`\frac{1s}{d^n}`$ otherwise. The only non-zero off-diagonal elements are $`W_{\stackrel{~}{j},\stackrel{~}{k}}^{\left[N\right]}\left(s\right)=`$ $`\frac{s}{d}`$. Choosing $`j`$ and $`k`$ appropriately in (3), we have the necessary condition $`1s\left(1+d^{n1}\right)`$. To show that this condition is also sufficient, we will use the spin representation to prove $`W^{\left[N\right]}\left(s\right)`$ is fully separable when $`d`$ is prime and $`s=\left(1+d^{n1}\right)^1`$. In order to do that, however, we first need to define the spin representation.
## III Computational and Spin Bases
Let $`H^{\left[N\right]}`$ denote an $`N`$-dimensional Hilbert space where $`N=d_1\times d_2\times \mathrm{}\times d_b`$. In this section we define different bases for $`N\times N`$ matrices on $`H^{\left[N\right]}`$ based on different representations of $`H^{\left[N\right]}`$ as a tensor product space, and the discussion is purely mathematical. In the applications we will be concerned with a specific representation $`H^{\left[N\right]}=_{a=1}^bH^{\left[d_a\right]}`$ and with the corresponding separability properties of densities on $`H^{\left[N\right]}.`$ The bases used will depend on the order of the tensor product as will the representation of a density matrix as $`\rho =\rho _1\mathrm{}\rho _b`$, the tensor product of densities $`\rho _k`$ on the $`H^{\left[d_k\right]}`$. For example, we might want to examine separability of a density matrix on $`H^{\left[90\right]}=H^{\left[6\right]}H^{\left[15\right]}`$ using matrices consistent with that tensor product. In a subsequent application one might want $`H^{\left[6\right]}`$ to represent the tensor product of spin $`1/2`$ and spin $`1`$ particles, i.e. a tensor product of $`H^{\left[2\right]}`$ and $`H^{\left[3\right]}`$, and the order of the sub-tensor product shouldn’t affect the theory. We confirm that assertion by showing that permuting the order of a tensor product corresponds to a conjugation operation and thus that the theory is generally applicable with only notational changes for particular applications.
Lemma 1 Let $`N=d_1\times d_2\times \mathrm{}\times d_b`$ and suppose $`M`$ is an $`N\times N`$ matrix with $`M=C^{\left(1\right)}\mathrm{}C^{\left(b\right)}`$, where the $`C^{\left(k\right)}`$ are $`d_k\times d_k`$ matrices. Let $`\sigma `$ denote a permutation of $`\{1,\mathrm{},b\}`$ and let $`M_\sigma =C^{\left(\sigma \left(1\right)\right)}\mathrm{}C^{\left(\sigma \left(b\right)\right)}`$. Then for all such $`C^{\left(k\right)}`$’s there is a permutation matrix $`Q_\sigma `$ such that $`Q_\sigma M_\sigma Q_\sigma ^1=M`$.
Proof: If $`M(j,k)=C^{\left(1\right)}(j_1,k_1)\mathrm{}C^{\left(b\right)}(j_b,k_b)`$, the index $`j`$ corresponds to the ordered $`b`$-tuple $`(j_1,\mathrm{},j_b)`$ and $`j`$ is uniquely defined by $`j=j_1\left(d_2\times \mathrm{}\times d_b\right)+j_2\left(d_3\times \mathrm{}\times d_b\right)+\mathrm{}+j_b`$ and similarly for $`k`$. Let $`j_\sigma `$ correspond to $`(j_{\sigma \left(1\right)},\mathrm{},j_{\sigma \left(b\right)})`$ and define the permutation matrix $`Q_\sigma `$ by $`Q_\sigma (j,s)=\delta (j_\sigma ,s)`$. Then $`Q_\sigma ^1=Q_\sigma ^t`$ and
$`Q_\sigma M_\sigma Q_\sigma ^1(j,k)=M_\sigma (j_\sigma ,k_\sigma )=C^{\left(\sigma \left(1\right)\right)}(j_{\sigma \left(1\right)},k_{\sigma \left(1\right)})\mathrm{}C^{\left(\sigma \left(b\right)\right)}(j_{\sigma \left(b\right)},k_{\sigma \left(b\right)})=M(j,k),`$
completing the proof.
The motivation for this work comes from the identification of the $`2\times 2`$ Hadamard matrix as a key tool in working with $`2`$-level systems. Specifically, suppose $`\rho =\frac{1}{2}\left(\sigma _0+\sigma _m\right)`$, where $`\sigma _m=m_j\sigma _j`$ and $`\sigma _1=\sigma _x`$, $`\sigma _2=\sigma _y`$, and $`\sigma _3=\sigma _z`$ are the usual Pauli matrices. Then the coefficients in the spin basis are related to those in the computational basis by the $`2\times 2`$ Hadamard matrix:
$$\left(\begin{array}{cc}1& m_1\\ m_3& im_2\end{array}\right)=\left(\begin{array}{cc}1& 1\\ 1& 1\end{array}\right)\left(\begin{array}{cc}\rho _{00}& \rho _{01}\\ \rho _{11}& \rho _{10}\end{array}\right).$$
(5)
The matrices in the two bases are connected in a similar fashion:
$$\left(\begin{array}{cc}\sigma _0& \sigma _1\\ \sigma _3& i\sigma _2\end{array}\right)=\left(\begin{array}{cc}1& 1\\ 1& 1\end{array}\right)\left(\begin{array}{cc}E_{0,0}& E_{0,1}\\ E_{1,1}& E_{1,0}\end{array}\right)\text{,}$$
(6)
where the matrices $`E_{j,k}=|jk|`$ define the computational basis. Note that a systematic application of these relationships requires both the use of the real Pauli matrices and a reindexing of both the Pauli matrices and the computational basis matrices to conform to the matrix notation.
The Hadamard matrix is the $`2\times 2`$ Fourier transform, and we extend the idea of (6) to $`d\times d`$ matrices. The adjusted basis $`A=\{A_{j,k},0j,k<d\}`$ is the set of $`d\times d`$ matrices defined by $`A_{j,k}=E_{j,j+k}`$, where $`+`$ denotes addition modulo $`d`$, and we define the “spin” matrices $`S=\{S_{j,k},\text{ }0j,k<d\}`$ using the analogue of (6) and the finite Fourier transform. Thus $`\left(S\right)F\left(A\right)`$ where $`F(j,k)=\mathrm{exp}\left(2\pi ijk/d\right)=\eta ^{jk}`$ with $`\eta =\mathrm{exp}\left(2\pi i/d\right)`$. (We will make the dependence on $`d`$ explicit below.) In detail
$$S_{j,k}=\underset{r=0}{\overset{d1}{}}F(j,r)A_{r,k}$$
(7)
is a sum of products of scalars times matrices. Since $`F`$ is invertible, it follows that $`S`$ is also a basis for the $`d\times d`$ matrices. Note that (6) is a special case of (7) with $`d=2`$ and $`\eta =1`$.
To illustrate these ideas, it is useful to write out the results for $`d=3`$ in detail. Then $`\eta =\mathrm{exp}\left(2\pi i/3\right)`$ and
$`S_{00}=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right)`$ $`S_{01}=\left(\begin{array}{ccc}0& 1& 0\\ 0& 0& 1\\ 1& 0& 0\end{array}\right)`$ $`S_{02}=\left(\begin{array}{ccc}0& 0& 1\\ 1& 0& 0\\ 0& 1& 0\end{array}\right)`$
$`S_{10}=\left(\begin{array}{ccc}1& 0& 0\\ 0& \eta & 0\\ 0& 0& \eta ^2\end{array}\right)`$ $`S_{11}=\left(\begin{array}{ccc}0& 1& 0\\ 0& 0& \eta \\ \eta ^2& 0& 0\end{array}\right)`$ $`S_{12}=\left(\begin{array}{ccc}0& 0& 1\\ \eta & 0& 0\\ 0& \eta ^2& 0\end{array}\right)`$
$`S_{20}=\left(\begin{array}{ccc}1& 0& 0\\ 0& \eta ^2& 0\\ 0& 0& \eta \end{array}\right)`$ $`S_{21}=\left(\begin{array}{ccc}0& 1& 0\\ 0& 0& \eta ^2\\ \eta & 0& 0\end{array}\right)`$ $`S_{22}=\left(\begin{array}{ccc}0& 0& 1\\ \eta ^2& 0& 0\\ 0& \eta & 0\end{array}\right)`$
The spin matrices $`S`$ not only form a basis for $`d\times d`$ matrices, but share many other properties with the real Pauli matrices, which we record next. We should note that the matrices $`S_{j,k}`$ were also defined in an earlier work by Fivel on Hamiltonians on discrete spaces, and many of the properties listed below were first established there.
Proposition 1 Fix $`d2`$ and let $`S`$ denote the corresponding set of spin matrices.
$`(i)`$ $`S`$ is an orthogonal basis of unitary matrices with respect to the trace inner product.
$`(ii)`$ If $`d`$ is odd, each matrix in $`S`$ is in $`SU(d)`$, while if $`d`$ is even, $`S_{j,k}`$ is in $`SU(d)`$ if and only if $`j+k`$ is even.
$`(iii)`$ $`S_{j,k}=\left(S_{1,0}\right)^j\left(S_{0,1}\right)^k`$, $`\left(S_{j,k}\right)^{}=\eta ^{jk}S_{dj,dk}=(S_{j,k})^{d1}`$ and $`[S_{j,k},S_{r,s}]=\left(\eta ^{kr}\eta ^{js}\right)S_{j+r,k+s}`$ using addition $`modd.`$
$`(iv)`$ $`tr\left(S_{j,k}\right)=0`$ for all $`(j,k)(0,0).`$
Proof: The key observation, noted in , is that the matrices are generated by $`S_{1,0}`$ and $`S_{0,1}`$:$`S_{j,k}=\left(S_{1,0}\right)^j\left(S_{0,1}\right)^k`$ with $`S_{0,1}S_{1,0}=\eta S_{1,1}`$. All of the remaining assertions, including orthogonality, follow from those relations and from easy computations. A useful consequence of the manipulations is
$$(S_{j,k})^m=\eta ^{(jk)m(m1)/2}S_{mj,mk}\text{.}$$
(11)
Unlike the Pauli matrices, these spin matrices need not be Hermitian; for example, when $`d=3`$ only the identity matrix is Hermitian. Thus, when computing the coefficients of a density matrix in these bases, as we do next, the Hermitian conjugation notation has to be retained. Note that the very last assertion in Corollary 1 corresponds to the usual inequality relating the $`L_2`$ magnitude of a Fourier transform and the $`L_2`$ magnitiude of the original function.
Corollary 1 $`(i)`$ The matrix elements of a $`d\times d`$ density matrix $`\rho `$ in the different bases are related by $`\left(s\right)=F^{}\left(a\right)`$.
$`(ii)`$ $`s_{0,0}=1`$, $`s_{dj,dk}=\eta ^{jk}s_{j,k}^{}`$ and $`\frac{1}{d}(Fs)_{j,0}=\rho _{j,j}0`$.
$`\left(iii\right)`$ $`_{j,k}\left|s_{j,k}\right|^2=d\left|\rho _{j,k}\right|^2`$ and $`\sqrt{_{j,k}\left|s_{j,k}\right|^2}\sqrt{_{j,k}\left|\rho _{j,k}\right|^2}1/\sqrt{d}.`$
Proof: We expand an arbitrary density matrix in the two bases:
$`\rho ={\displaystyle \underset{j,k}{}}a_{j,k}A_{j,k}={\displaystyle \frac{1}{d}}{\displaystyle \underset{j,k}{}}s_{j,k}S_{j,k}`$
where $`a_{j,k}=Tr\left(A_{j,k}^{}\rho \right)`$ gives $`a_{j,k}=\rho _{j,j+k}`$, using addition $`modd,`$ and $`s_{j,k}=Tr\left(S_{j,k}^{}\rho \right)`$. Then from (7), $`s_{j,k}=_{r=0}^{d1}F^{}(j,r)a_{r,k}`$, which proves $`\left(i\right)`$. Note that we have to include a complex conjugation in the formula which is unnecessary in the $`d=2`$ case, since the Hadamard matrix has real entries. The assertions in $`\left(ii\right)`$ follow from the definitions and from the fact that $`\rho `$ is Hermitian with trace equal to one. Finally the relations in $`\left(iii\right)`$ follow from $`Tr\left(\left(s\right)^{}\left(s\right)\right)=Tr\left(\left(a\right)^{}\left(F^{}\right)^{}\left(F^{}\right)\left(a\right)\right)=dTr\left(\left(a\right)^{}\left(a\right)\right)=dTr\left(\rho ^2\right)`$, and from $`Tr\left(\rho ^2\right)=_k\lambda _k^2_k1/d^2=1/d`$, where the $`\lambda _k`$ are the non-negative eigenvalues of the density $`\rho .`$
Now let $`N=d_1\times d_2\times \mathrm{}\times d_b`$ with $`d_i2`$ and with the order of multiplication fixed throughout the discussion. We use the underlying and fixed tensor product representation of $`H^{\left[N\right]}`$ to define the sets of computational and adjusted bases $`E^{\left[N\right]}`$ and $`A^{\left[N\right]}`$ for $`N\times N`$ matrices as
$`E_{j,k}^{\left[N\right]}=E_{j_1,k_1}^{\left(1\right)}\mathrm{}E_{j_b,k_b}^{\left(b\right)}\text{and }A_{j,k}^{\left[N\right]}=A_{j_1,k_1}^{\left(1\right)}\mathrm{}A_{j_b,k_b}^{\left(b\right)},`$
where $`j`$ and $`k`$ correspond to their $`b`$-tuples and the superscripts in parentheses identify the corresponding $`d_i`$. It follows that $`A_{j,k}^{\left[N\right]}=E_{j,jk}^{\left[N\right]}`$ where the addition of the indices is defined by
$$jk(j_1+k_1modd_1,\mathrm{},j_b+k_bmodd_b).$$
(12)
The corresponding set of spin matrices $`S^{\left[N\right]}`$ is then defined by $`\left(S^{\left[N\right]}\right)=F^{\left[N\right]}\left(A^{\left[N\right]}\right)`$ or
$`S_{j,k}^{\left[N\right]}={\displaystyle \underset{r=0}{\overset{N1}{}}}F^{\left[N\right]}(j,r)A_{r,k}^{\left[N\right]}`$
where $`F^{\left[N\right]}=F^{\left(1\right)}\mathrm{}F^{\left(b\right)}`$ is the usual tensor product of the Fourier transforms $`F^{\left(k\right)}`$ which depend on $`d_k`$. Since we will be taking powers of the $`\eta `$’s, we will use subscripts to denote the dependency of $`\eta `$ on $`d_k`$: $`\eta _k=\mathrm{exp}\left(2\pi i/d_k\right)`$. It is easy to show that an equivalent definition of $`S^{\left[N\right]}`$ is given by
$$S_{j,k}^{\left[N\right]}=\underset{i=1}{\overset{b}{}}\left(F^{\left(i\right)}A^{\left(i\right)}\right)_{j_i,k_i}\text{.}$$
(13)
Linearity again implies that if $`\rho ^{\left[N\right]}`$ is a density matrix on the $`N\times N`$ Hilbert space $`H^{\left[N\right]}`$ with
$`\rho ^{\left[N\right]}={\displaystyle \underset{j,k}{}}a_{j,k}^{\left[N\right]}A_{j,k}^{\left[N\right]}={\displaystyle \frac{1}{N}}{\displaystyle \underset{j,k}{}}s_{j,k}^{\left[N\right]}S_{j,k}^{\left[N\right]},`$
then
$$\left(s^{\left[N\right]}\right)=F^{\left[N\right]}\left(a^{\left[N\right]}\right)$$
(14)
and $`a_{j,k}^{\left[N\right]}=\rho _{j.jk}^{\left[N\right]}`$. Thus we have two different representations for a density matrix $`\rho ^{\left[N\right]}`$, and both of them depend on the underlying tensor product representation of $`H^{\left[N\right]}`$.
## IV Sigma Variations
A fully separable density matrix can be represented as a convex combination of tensor products of pure states or trace one projections, and we need to represent such $`d\times d`$ projections in a systematic fashion in the spin basis. (All projections in this paper are trace one projections.) In the Appendix we show how all trace one projections for $`d=3`$ can be represented in a form completely analogous to the $`d=2`$ case, but for our immediate purposes we only need to characterize a subclass. The motivation is given by writing the particular $`d=2`$ projections $`\frac{1}{2}\left(\sigma _0\pm \sigma _k\right)`$ as $`P_k(r)=_{m=0}^1\left(\left(1\right)^r\sigma _k\right)^m`$, where $`r=0`$ or $`r=1`$. Then $`P_k(r)`$ is the average of the cyclic subgroup generated by $`(1)^r\sigma _k`$, and since $`\left((1)^r\sigma _k\right)^2=\sigma _0`$, the key property $`P_k(r)P_k(r)=P_k(r)`$ reduces to an exercise in group theory. The generalization of this idea to arbitrary $`d`$ is immediate, and we first treat the case when $`d`$ is prime.
Proposition 2 Let $`d=p2`$ be prime. Let $`u=(j,k)(0,0)`$ denote the index of a spin matrix $`S_u`$, and let $`r`$ be an integer. Then if $`p>2,`$ the matrix
$$P_u(r)\frac{1}{p}\underset{m=0}{\overset{p1}{}}\left(\eta ^rS_u\right)^m$$
(15)
is a projection with unit trace. The assertion is also valid for $`p=2`$ provided $`iS_u`$ is used in lieu of $`S_u`$ throughout when $`u=(1,1).`$
Proof: A matrix $`P`$ is a pure state or a trace one projection if it is Hermitian, has trace $`1,`$ and $`P^2=P`$. First, $`\left(\eta ^rS_u\right)^m`$ is proportional to $`S_{mj,mk}`$; consequently, it cannot be proportional to $`S_{0,0}`$ for $`0<m<p`$. Therefore, only $`S_{0,0}`$ contributes to the trace of $`P(u,r)`$, confirming the trace condition. Using (11) it follows that $`\left(\left(\eta ^rS_u\right)^m\right)^{}=\left(\eta ^rS_u\right)^{pm}`$ and that when $`p`$ is odd $`\left(\eta ^rS_u\right)^m\left(\eta ^rS_u\right)^{pm}=\left(S_u\right)^p=\left(\eta ^{jk}\right)^{\frac{p\left(p1\right)}{2}}S_{0,0}=S_{0,0}`$. Thus $`P(u,r)`$ is Hermitian. The verification that $`P(u,r)^2=P(u,r)`$ follows from an easy computation. The assertion that $`\left(\eta ^{jk}\right)^{\frac{p\left(p1\right)}{2}}=1`$ fails for prime $`p`$ only when $`p=2`$ and $`j=k=1`$. Thus, the reintroduction of $`i`$ and of $`\sigma _y=iS_{1,1}`$ is required to complete the proof.
As an example of the notation, it is easy to check that if $`k=0`$, then $`P_{j,0}(r)`$ is one of the diagonal projections $`E_{i,i}`$. Other projections are less sparse, however. For example, when $`d=3`$ and $`k0`$, $`P_{j,k}(r)`$ has no zero entries in the computational basis representation.
In the preceding proof, we exploited the fact that for $`d`$ an odd prime the powers of each matrix $`S_u,u(0,0),`$ form a cyclic subgroup of order $`d.`$ When $`d`$ is not prime we can get analogous results using a similar proof, but there are restrictions on the indices that arise since the coefficient of the identity matrix in (11) when $`m=d`$ need not be unity. In Proposition 2 this led to the introduction of the factor $`i=\mathrm{exp}\left(\pi i/2\right)`$ when $`d=2`$, and that modification is a special case of a more general situation.
Proposition 3 Suppose $`d`$ is composite. Let $`u=(j,k)`$ be $`(0,1)`$, $`(1,0)`$ or else an index such that $`j0`$ and $`k0`$ have no common factors. Suppose $`d`$ is odd or $`jk`$ is even. Then if $`r`$ is an integer, $`P_u(r)=\frac{1}{d}\underset{m=0}{\overset{d1}{}}(\eta ^rS_u)^m`$ is a projection with unit trace. If $`d`$ is even and $`jk`$ is odd, then $`P_u(r)=\frac{1}{d}_{m=0}^{d1}\left(\alpha \eta ^rS_u\right)^m`$ is a projection with unit trace, where $`\alpha =e^{\pi i/d}`$.
Proof: Suppose $`\left(\eta ^rS_u\right)^m`$ or $`\left(\alpha \eta ^rS_u\right)^m`$ is proportional to $`S_{0,0}`$ for $`0<m<d`$, so that $`mj=rd`$ and $`mk=sd`$ for some integers $`r`$ and $`s`$. Since $`j`$ and $`k`$ are relatively prime, there are integers $`a`$ and $`b`$ such that $`aj+bk=1`$ , and it follows that $`m=ard+bsd=\left(ar+bs\right)d`$, contradicting $`m<d`$. Thus $`P_u(r)`$ has trace one. Using (11) when $`m=d`$, we find that the coefficient of $`S_{0,0}`$ is one in the first case, while in the second case the extra factor of $`\alpha ^d=\left(1\right)`$ is necessary to make the overall coefficient equal to one. In both cases it follows from that key result as in Proposition 2 that $`P_u^2(r)=P_u(r)`$ and that $`P_u(r)`$ is Hermitian, completing the proof.
An important relationship between these subgroup projections and the generating spin matrix follows from the definitions.
Corollary 2 For any integer $`t0`$ and any $`d2`$,
$$(\eta ^rS_u)^t=\underset{m=0}{\overset{d1}{}}\eta ^{mt}P_u(m+r)\text{,}$$
(16)
subject to the usual caveat about $`\alpha `$. In particular, $`S_{0,0}=_{m=0}^{d1}P_u(m+r)`$
Proof: $`_{m=0}^{d1}\eta ^{mt}P_u(m+r)=\frac{1}{d}_k\left(\eta ^rS_u\right)^k_m\eta ^{mt}\eta ^{mk}=\left(\eta ^rS_u\right)^t`$, as required.
Next consider a Hilbert space that is the direct product of $`b`$ Hilbert spaces with dimensions $`d_1,\mathrm{},d_b.`$ Projections in the constituent $`d_i`$ dimensional spaces also define projections in tensor product spaces, and the proof of the following is immediate. As before, we let the superscript $`k`$ denote the dependence on $`d_k`$.
Corollary 3 Let $`N=d_1\times d_2\times \mathrm{}\times d_b`$ and let $`H^{\left[N\right]}=_{a=1}^bH^{\left[d_a\right]}`$ be an $`N`$ dimensional Hilbert space. Let $`u`$ denote a $`b`$-dimensional vector of index pairs $`u_i=(j_i,k_i)`$ where $`0j_i,k_id_i1`$, and let $`r=(r_1,\mathrm{},r_b)`$ where the $`r_i`$ are integers. Then if the $`P_{u_k}^{\left(k\right)}(r_k)`$ are trace one projections on $`H^{[d_k]}`$,
$`P_u(r)={\displaystyle \underset{k=1}{\overset{b}{}}}P_{u_k}^{(k)}(r_k)`$
is a trace one projection on $`H^{[N]}`$, provided $`\alpha S_u`$ is used in place of $`S_u`$ when $`d`$ is even and $`u=(j,k)`$ with $`jk`$ odd. Furthermore, if $`\eta \left(r\right)_{k=1}^b\eta _k^{r_k}`$ and $`t`$ is a non-negative integer,
$$\left(\eta \left(r\right)S_u^{[N]}\right)^t=\underset{l_1=0}{\overset{d_11}{}}\mathrm{}\underset{l_b=0}{\overset{d_b1}{}}\underset{k=1}{\overset{b}{}}\eta _k^{l_kt}P_{u_k}^{\left(k\right)}(l_k+r_k)\text{,}$$
(17)
and in particular $`S_{0,0}^{[N]}=\underset{l_1=0}{\overset{d_11}{}}\mathrm{}\underset{l_b=0}{\overset{d_b1}{}}_{k=1}^bP_{u_k}^{(k)}(l_k+r_k)`$.
In order to show separability results for Werner densities, we need to identify a special class of fully separable density matrices in the tensor product space $`H^{\left[N\right]}\left(d\right)`$ of $`n`$ $`d`$-dimensional Hilbert spaces, where $`N=d^n`$. This approach is motivated by results in and is our final variation on the Pauli $`\sigma `$-matrices.
Proposition 4 Let $`d2`$ and let $`u^{\left(n\right)}=(u_1,\mathrm{},u_n)`$ and $`r^{\left(n\right)}=(r_1,\mathrm{},r_n)`$ denote vectors of indices and values as defined in the preceding propositions. Then provided $`\alpha S_u`$ is used in place of $`S_u`$ when $`d`$ is even and $`u=(j,k)`$ with $`jk`$ odd,
$`\rho (u^{\left(n\right)},r^{\left(n\right)})={\displaystyle \frac{1}{d^n}}\left(\left(S_{0,0}\mathrm{}S_{0,0}\right)+{\displaystyle \underset{m=1}{\overset{d1}{}}}\left(\left(\eta ^{r_1}S_{u_1}\right)^m\mathrm{}\left(\eta ^{r_n}S_{u_n}\right)^m\right)\right)`$
is a fully separable density matrix on $`H^{\left[N\right]}\left(d\right)`$.
Proof: The assertion is true for $`n=1`$ and suppose it holds for $`n`$. Let $`u^{(n+1)}`$ and $`r^{(n+1)}`$ be given index and parameter vectors. Since we require only the $`n^{}th`$ and $`(n+1)^{}st`$ indices in the proof, we leave the other indices fixed and implicit and let $`\rho (u_n,r_n)`$ denote $`\rho (u^{\left(n\right)},r^{\left(n\right)})`$. By the induction hypothesis
$`{\displaystyle \frac{1}{d}}{\displaystyle \underset{s=0}{\overset{d1}{}}}\rho (u_n,r_n+s)P_{u_{n+1}}^{\left(n+1\right)}(r_{n+1}s)`$
is fully separable. Multiplying out and collecting terms produces expressions of the form
$`[(\eta ^{r_1}S_{u_1})^{m_1}\mathrm{}(\eta ^{r_n}S_{u_n})^{m_1}](\eta ^{r_{n+1}}S_{u_{n+1}})^{m_2}{\displaystyle \frac{1}{d}}{\displaystyle \underset{s=0}{\overset{d1}{}}}\eta ^{s(m_1m_2)}.`$
By the same analysis used earlier, terms with $`m_1=m_2`$ have an overall coefficient of $`1`$ while all other terms have coefficient $`0`$, and that completes the proof of the induction step.
## V Applications
We now have the tools to prove a sufficient condition for full and partial separability which extend the results in . This is a general sufficient condition for full or partial separability, and the results in and for $`N=2^n`$ and $`N=3^n`$ respectively showing the existence of a neighborhood of the weighted identity in which every density is fully separable follow as corollaries. As usual $`H^{\left[N\right]}`$ will denote an $`N`$ dimensional Hilbert space which can be written as a tensor product: $`H^{\left[N\right]}`$ $`=H^{\left[d_1\right]}\mathrm{}H^{\left[d_b\right]}`$, where the $`H^{\left[d_k\right]}`$ are $`d_k`$-dimensional spaces and $`N=d_1\times d_2\times \mathrm{}\times d_b`$. We define $`D(d_1,\mathrm{},d_b)`$ and refer to $`H^{\left[d_1\right]}\mathrm{}H^{\left[d_b\right]}`$ as the $`D`$ tensor product version of $`H^{\left[N\right]}`$. Since $`H^{\left[N\right]}`$ may be represented as a tensor product space in different ways, the kind of separability to be discussed depends on the representation. For example, if $`N=3^n`$ and $`H^{[N]}`$ is represented as the tensor product of $`n`$ three-dimensional spaces, we are discussing full separability. If subsets of the trits are taken together and represented in $`3^k`$-dimensional spaces, we are discussing the corresponding partial separability. By virtue of Lemma 1, we know that the fundamental mathematics involved doesn’t depend on the order in which the tensor products are taken or which trits are grouped together.
In expressing the condition of the theorem, we use the $`D`$ spin coefficients to introduce an $`L_1`$ norm on the space of $`N\times N`$ densities, and we will refer to that hereafter as the $`D`$ spin norm and to the related separability as $`D`$ separability.
Theorem 1 Let $`H^{\left[N\right]}`$ denote an $`N`$-dimensional Hilbert space with $`N=d_1\times d_2\times \mathrm{}\times d_b`$. Suppose $`H^{\left[N\right]}=H^{\left[d_1\right]}\mathrm{}H^{\left[d_b\right]}`$, where the $`H^{\left[d_k\right]}`$ are $`d_k`$-dimensional Hilbert spaces. If $`\rho `$ is a density matrix on $`H^{\left[N\right]}`$, then $`\rho `$ is $`D(d_1,\mathrm{},d_b)`$ separable provided
$$\rho _{1,D}\underset{(j,k)(0,0)}{}\left|s_{j,k}^{\left[N\right]}\right|1,$$
(18)
where $`\rho `$ has the spin representation $`\frac{1}{N}_{j,k}s_{j,k}^{\left[N\right]}S_{j,k}^{\left[N\right]}`$ defined in term of the $`D`$ tensor product: $`S_{j,k}^{\left[N\right]}=_{i=1}^bS_{j_i,k_i}^{\left(i\right)}`$. It follows that in the set of density matrices on $`H^{[N]}`$ there is a neighborhood relative to $`D`$ of the random state $`\frac{1}{N}S_{0,0}^{[N]}`$ in which every density matrix is $`D`$ separable.
Proof: If $`d_i`$ is prime or $`j_i`$ and $`k_i`$ are relatively prime, the factor $`S_{j_i,k_i}^{\left(i\right)}`$ can be written as a weighted sum of projections as in Corollary 2. If $`d_i`$ is composite and the indices $`j_i`$ and $`k_i`$ are not relatively prime, then up to a factor of $`\eta _i^{t_i}`$, $`S_{j_i,k_i}^{\left(i\right)}`$ can be written as $`\left(\eta _i^rS_u^{\left(i\right)}\right)^s`$ for some $`u=(\overline{j}_i,\overline{k}_i)`$ with $`\overline{j}_i`$ and $`\overline{k}_i`$ relatively prime, and thus $`S_{j_i,k_i}^{\left(i\right)}`$ can also be written as a weighted sum of projections. Now since $`\rho `$ is a density, either $`S_{j,k}^{\left[N\right]}`$ is Hermitian and thus $`s_{j,k}^{\left[N\right]}`$ is real, or $`S_{j,k}^{\left[N\right]}`$ appears in a pair $`s_{j,k}^{\left[N\right]}S_{j,k}^{\left[N\right]}+s_{j,k}^{\left[N\right]}\left(S_{j,k}^{\left[N\right]}\right)^{}`$. In the second case we use (17) in Corollary 3 and the preceding comments to collect the various factors of $`\eta _i^{t_i}`$ together and obtain
$`s_{j,k}^{[N]}S_{j,k}^{[N]}+s_{j,k}^{[N]}\left(S_{j,k}^{[N]}\right)^{}`$ $`=`$ $`{\displaystyle \underset{l_1=0}{\overset{d_11}{}}}\mathrm{}{\displaystyle \underset{l_b=0}{\overset{d_b1}{}}}{\displaystyle \underset{k=1}{\overset{b}{}}}P_{u_k}^{(k)}(l_k+r_k)\left\{\beta _{j,k}s_{j,k}^{[N]}\eta ^{}\left(l\right)+\beta _{j,k}^{}s_{j,k}^{[N]}\eta \left(l\right)\right\}`$
$`=`$ $`\left|s_{j,k}^{\left[N\right]}\right|{\displaystyle \underset{l_1=0}{\overset{d_11}{}}}\mathrm{}{\displaystyle \underset{l_b=0}{\overset{d_b1}{}}}{\displaystyle \underset{k=1}{\overset{b}{}}}P_{u_k}^{(k)}(l_k+r_k)\left\{\mathrm{exp}\left(i\theta _{j,k}\right)\eta ^{}\left(l\right)+\mathrm{exp}\left(i\theta _{j,k}\right)\eta \left(l\right)\right\}`$
where $`\theta _{j,k}`$ denotes the phase of $`\beta _{j,k}s_{j,k}^{[N]}`$ and $`l`$ denotes the $`b`$-vector with components $`l_k`$. The caveat that $`\alpha _iS_u`$ is in the projections $`P_{u_k}^{(i)}(l_k+r_k)`$ in lieu of $`S_u`$ when $`d_i`$ is even and $`u=(j_i,k_i)`$ with $`j_ik_i`$ odd applies throughout the proof and will not be explicitly cited. Since $`\alpha _i`$ has magnitude $`1`$, only the phase factor will be affected. Using the last assertion in Corollary 3, we can write $`\left|s_{j,k}^{\left[N\right]}\right|S_{0,0}^{\left[N\right]}+\frac{1}{2}\left(s_{j,k}^{\left[N\right]}S_{j,k}^{\left[N\right]}+s_{j,k}^{\left[N\right]}\left(S_{j,k}^{\left[N\right]}\right)^{}\right)`$ as
$`\left|s_{j,k}^{\left[N\right]}\right|{\displaystyle \underset{l_1=0}{\overset{d_11}{}}}\mathrm{}{\displaystyle \underset{l_b=0}{\overset{d_b1}{}}}{\displaystyle \underset{k=1}{\overset{b}{}}}P_{u_k}(l_k+r_k)\left\{1+\mathrm{cos}\left(\theta _{j,k}\mathrm{arg}\left(\eta \left(l\right)\right)\right)\right\}.`$
Since the expression in brackets is non-negative, the right-hand side is a non-negative multiple of a $`D`$-separable density. In the case when $`S_{j,k}^{\left[N\right]}`$ is Hermitian we derive the same expression with the same conclusion. It follows that $`\rho `$ can be written as a convex combination of fully separable densities plus the residual term $`\left(1_{(j,k)(0,0)}\left|s_{j,k}^{\left[N\right]}\right|\right)\frac{1}{N}S_{0,0}^{\left[N\right]}`$. The hypothesis guarantees that the coefficient of $`\frac{1}{N}S_{0,0}^{\left[N\right]}`$ is non-negative, and that completes the proof of $`D`$-separability.
As another application of the machinery, we can prove for prime $`p`$ that the necessary condition $`s\left(1+p^{n1}\right)^1`$ is sufficient for full separability of the generalized Werner density matrix $`W^{\left[N\right]}\left(s\right)=\frac{1s}{N}`$ $`I+s\tau `$. We have $`N=p^n`$, $`I`$ is the identity, $`\tau `$ is the projection defined by the state $`|\psi ^{\left[N\right]}=\frac{1}{\sqrt{p}}_{k=0}^{p1}|\stackrel{~}{k}`$ and $`\stackrel{~}{k}`$ denotes the $`n`$-long repeated index $`k\mathrm{}k`$. Given this special structure we find
$`W^{\left[N\right]}\left(s\right)={\displaystyle \frac{1s}{p^n}}I+{\displaystyle \frac{s}{p}}{\displaystyle \underset{j=0}{\overset{p1}{}}}{\displaystyle \underset{k=0}{\overset{p1}{}}}|\stackrel{~}{j}\stackrel{~}{k}|={\displaystyle \frac{1s}{p^n}}I+{\displaystyle \frac{s}{p}}{\displaystyle \underset{j=0}{\overset{p1}{}}}{\displaystyle \underset{k=0}{\overset{p1}{}}}A_{\stackrel{~}{j},\stackrel{~}{k}}^{\left[N\right]}`$
where we have used the modular vector addition defined in (12). Computing the spin coefficients gives $`s_{0,0}=1`$, $`s_{j,m}=0`$ if $`m`$ is not a $`\stackrel{~}{k}`$ with $`0k<p`$, and otherwise
$`s_{j,\stackrel{~}{k}}={\displaystyle \underset{r}{}}F^{}(j,r){\displaystyle \frac{s}{p}}\delta (r,Ind)`$
where $`Ind=\{\stackrel{~}{r}:0r<p\}`$. Using the dot product of the index vectors $`jr=j_kr_kmodp,`$
$`s_{j,\stackrel{~}{k}}={\displaystyle \underset{r}{}}\mathrm{exp}\left({\displaystyle \frac{2\pi i}{p}}\left(jr\right)\right)a_{r,\stackrel{~}{k}}={\displaystyle \frac{s}{p}}\left(1+{\displaystyle \underset{r=1}{\overset{p1}{}}}\mathrm{exp}\left({\displaystyle \frac{2\pi i}{p}}\left(j\stackrel{~}{r}\right)\right)\right).`$
Let $`Ind(p,n)=\{j:_{r=0}^{N1}j_r=0modp\}.`$ Then it’s easy to check that $`s_{j,\stackrel{~}{k}}=s`$ if and only if $`j`$ is in $`Ind(p,n)`$ and that there are exactly $`p^{n1}`$ such indices. All other $`s_{j,\stackrel{~}{k}}`$ equal zero, and we can write $`W^{\left[N\right]}\left(s\right)`$ in the spin basis as
$$W^{\left[N\right]}\left(s\right)=\frac{1s}{p^n}S_{0,0}^{\left[N\right]}+s\underset{jInd(p,n)}{}\underset{k=0}{\overset{p1}{}}S_{j,\stackrel{~}{k}}^{\left[N\right]}.$$
(19)
Theorem 2 Let $`p`$ be prime and $`N=p^n`$. Then the generalized Werner density matrix $`W^{\left[N\right]}\left(s\right)`$ is fully separable on $`H^{\left[n\right]}\left(p\right)`$ if and only if $`s\left(1+p^{n1}\right)^1`$.
Proof: As shown above, necessarily $`s\left(1+p^{n1}\right)^1`$. Checking the preceding derivation, note that
$`{\displaystyle \frac{1}{p}}{\displaystyle \underset{j=0}{\overset{p1}{}}}E_{\stackrel{~}{j},\stackrel{~}{j}}={\displaystyle \frac{1}{p}}{\displaystyle \underset{j=0}{\overset{p1}{}}}A_{\stackrel{~}{j},\stackrel{~}{0}}^{\left[N\right]}={\displaystyle \underset{jInd(p,n)}{}}S_{j,0}^{\left[N\right]}`$
is a sum of fully separable projections. Taking $`s=\left(1+p^{n1}\right)^1`$ we can write $`W^{\left[n\right]}\left(s\right)`$ as
$`W^{\left[n\right]}\left(s\right)={\displaystyle \frac{1}{1+p^{n1}}}\left[{\displaystyle \frac{1}{p}}{\displaystyle \underset{j=0}{\overset{p1}{}}}E_{\stackrel{~}{j},\stackrel{~}{j}}+{\displaystyle \underset{jInd(p,n)}{}}{\displaystyle \frac{1}{p^n}}\left(S_{\stackrel{~}{0},\stackrel{~}{0}}+{\displaystyle \underset{k=1}{\overset{p1}{}}}S_{j,\stackrel{~}{k}}\right)\right]\text{.}`$
For each $`k0`$, $`Ind(p,n)`$ is mapped in a one-to-one manner onto itself by $`jkj`$ where $`\left(kj\right)_r=kj_rmodp.`$ Thus
$$W^{\left[n\right]}\left(s\right)=\frac{1}{1+p^{n1}}\left[\frac{1}{p}\underset{j=0}{\overset{p1}{}}E_{\stackrel{~}{j},\stackrel{~}{j}}+\underset{jInd(p,n)}{}\left(\frac{1}{p^n}\left(S_{\stackrel{~}{0},\stackrel{~}{0}}+\underset{k=1}{\overset{p1}{}}S_{kj,\stackrel{~}{k}}\right)\right)\right]\text{.}$$
(20)
But since
$`\left(S_{j_1,1}\right)^k\mathrm{}\left(S_{j_n,1}\right)^k=\eta ^{k{\scriptscriptstyle j_i}}S_{kj,\stackrel{~}{k}}=S_{kj,\stackrel{~}{k}}`$
for $`j`$ in $`Ind(p,n)`$, each $`j`$-sum in (20) is fully separable by Proposition 4, completing the proof.
It follows for the Werner densities that at the extreme value $`s=\left(1+p^{n1}\right)^1`$, $`_{(j,k)(0,0)}\left|s_{j,k}^{\left[N\right]}\right|=p\frac{1p^n}{1+p^{\left(n1\right)}}`$, where the coefficients are based on the decomposition $`D=(p,\mathrm{},p).`$ When $`p=n=2`$, that value is $`1`$, showing that the global bound of Theorem 1 is attained. However, for larger $`n`$ and prime $`p2`$ the condition $`\rho _{1,D}1`$ is too strong for that class, and the special structure of the Werner densities allowed a more refined analysis of $`D=(p,\mathrm{},p)`$-separability.
It was shown in the qubit case in that for each $`n`$ and given $`ϵ>0`$, there exists a $`D=(2,\mathrm{},2)`$-inseparable density on $`H^{\left[2^n\right]}`$ which has $`\rho _{1,D}<1+ϵ`$. Thus, for each fixed $`n`$ the sufficient condition of Theorem 1 is the best possible for full separability of qubits. We conjecture that the same is true in general: given any separability vector $`D`$ and $`ϵ>0`$ there exists a $`D`$-inseparable density $`\rho `$ with $`\rho _{1,D}<1+ϵ`$.
## ACKNOWLEDGMENTS
A. O. Pittenger gratefully acknowledges the hospitality of the Centre for Quantum Computation at Oxford University and support from UMBC and the National Security Agency. M. H. Rubin wishes to thank the Office of Naval Research and the National Security Agency for support of this work.
## VI Appendix
By emphasizing selected properties of projections for $`d=2`$, we can obtain a representation of all (trace one) projections in spin notation for $`d>2`$. We concentrate on $`d=3`$. To motivate the approach, recall from (5) that when $`d=2`$, $`m_3=\left(1\right)\rho _{0,0}+\left(1\right)\rho _{1,1}`$, so that this particular spin coordinate is a convex combination of $`\left(+1\right)`$ and $`\left(1\right)`$, another way of stating the well-known correspondence between $`m_3`$, the coefficient of $`\sigma _z`$, and the diagonal of $`\rho `$. If $`\rho `$ is also a projection, then in the computational coordinates, $`\rho _{j,k}=b_jb_k\mathrm{exp}\left(i\left(\phi _j\phi _k\right)\right)`$, so that fixing $`m_3`$ fixes $`\rho _{0,0}=b_0^2`$ and $`\rho _{1,1}=b_1^2`$, and only the phase factor $`\theta =`$ $`\phi _0\phi _1`$ is unspecified. Using the change of basis formula, the two remaining spin coefficients of a projection with prescribed $`m_3`$ are thus given in terms of the parameter $`\theta `$ by
$`\left(\begin{array}{c}m_1\\ im_2\end{array}\right)=\left(\begin{array}{c}s_{0,1}\\ s_{1,1}\end{array}\right)=\left(\begin{array}{cc}1& 1\\ 1& 1\end{array}\right)\left(\begin{array}{c}b_0b_1e^{i\theta }\\ b_0b_1e^{i\theta }\end{array}\right)\text{,}`$
where $`0\theta <2\pi `$. If we let $`t_k`$ denote the value of $`s_{k,1}`$ when $`\theta =0`$, we can rewrite the preceding equation as
$`\left(\begin{array}{c}s_{0,1}\\ s_{1,1}\end{array}\right)=={\displaystyle \frac{1}{2}}\left(\begin{array}{cc}1& 1\\ 1& 1\end{array}\right)\left(\begin{array}{cc}e^{i\theta }& 0\\ 0& e^{i\theta }\end{array}\right)\left(\begin{array}{cc}1& 1\\ 1& 1\end{array}\right)\left(\begin{array}{c}t_0\\ t_1\end{array}\right)\text{.}`$
Making the obvious definitions, this gives $`\stackrel{}{s}=M_2\left(\theta \right)\stackrel{}{t}`$, and we also find that
$`M_2\left(\theta \right)=\left(\begin{array}{cc}\mathrm{cos}\left(\theta \right)& i\mathrm{sin}\left(\theta \right)\\ i\mathrm{sin}\left(\theta \right)& \mathrm{cos}\left(\theta \right)\end{array}\right)=\mathrm{cos}\left(\theta \right)\sigma _0+i\mathrm{sin}\left(\theta \right)\sigma _x.`$
The geometry of this result is that if $`1<m_3<1`$, then the remaining spin coefficients in the projections associated with $`m_3`$ can be identified with the range of a one parameter family of invertible mappings $`\left\{M_2\left(\theta \right)\right\}`$ acting on $`\stackrel{}{t}`$ and are represented by the intersection of the surface of the Bloch sphere with a horizontal plane at height $`m_3`$.
The same pattern of results holds for $`d=3`$. Since $`s_{2,0}=s_{1,0}^{}`$, the diagonal of a given $`\rho `$ is in one-to-one correspondence with $`s_{1,0}`$ via the equation $`s_{1,0}=\rho _{0,0}\left(1\right)+\rho _{1,1}\left(\eta ^2\right)+\rho _{2,2}\left(\eta \right)`$. That is, $`s_{1,0}`$ is a convex combination of the vertices of an equilateral triangle in the complex plane and thus uniquely corresponds to the weights of the vertices, weights which are the entries of the diagonal of $`\rho `$. For larger values of $`d`$, the geometry is more complicated. For example if $`d=4,`$ the diagonal of a given $`\rho `$ corresponds to two spin coefficients: $`1s_{2,0}+1`$ and $`s_{1,0}`$ which is restricted to a rectangle in the complex plane with vertices $`\pm \left(1+s_{2,0}\right)/2\pm i\left(1s_{2,0}\right)/2`$. In general the diagonal of a density matrix $`\rho `$ corresponds to $`d/2`$ spin coefficients $`s_{j,0}`$, $`j0`$, when $`d`$ is even and $`\left(d1\right)/2`$ spin coefficients when $`d`$ is odd.
Once $`s_{1,0}`$ is fixed in the $`d=3`$ case, there are three complex parameters remaining to be specified: $`s_{0,1},s_{1,1},`$ and $`s_{2,1}`$, since the other four spin coefficients are forced by the restriction $`s_{3j,3k}=\eta ^{jk}s_{j,k}^{}`$. If $`\rho `$ is a projection, $`\left|s_{j,k}\right|^2=3\left|\rho _{j,k}\right|^2=3`$, and thus $`\left|s_{2,0}\right|^2+\left|s_{k,1}\right|^2=1`$, tempting one to look for an analogue of the Bloch sphere to represent all densities. However, the normalization arising from $`tr\left(\rho ^2\right)=1`$ is only a necessary condition on the parameters, and examples show it’s not sufficient. (See also .) Instead we follow the $`d=2`$ paradigm and describe trace one projections associated with a fixed $`s_{1,0}`$. If $`\rho `$ is such a projection, then in the computational coordinates $`\rho =|uu|`$, where $`|u`$ denotes a normalized three vector with $`u_k=b_ke^{i\phi _k}`$ and $`\left|b_k\right|^2=1`$. Fixing $`s_{1,0}`$ fixes the $`b_k`$’s, and it follows from Corollary 1 and the structure of $`\rho `$ that
$`\left(\begin{array}{c}s_{0,1}\\ s_{1,1}\\ s_{2,1}\end{array}\right)=\left(\begin{array}{ccc}1& 1& 1\\ 1& \eta ^2& \eta \\ 1& \eta & \eta ^2\end{array}\right)\left(\begin{array}{ccc}e^{i\theta _0}& 0& 0\\ 0& e^{i\theta _1}& 0\\ 0& 0& e^{i\theta _2}\end{array}\right)\left(\begin{array}{c}b_0b_1\\ b_1b_2\\ b_2b_0\end{array}\right)`$
where $`\theta _k=\phi _k\phi _{k+1}`$ with addition modulo $`2\pi `$ and with the normalization $`\theta _k=0mod\mathrm{\hspace{0.17em}2}\pi .`$ Again letting $`t_k`$ denote the value of $`s_{k,1}`$ when the $`\theta _k`$’s are chosen to be zero, this time we obtain a two parameter family of projections associated with a given value of $`s_{1,0}`$. Letting $`\stackrel{}{s}`$ denote the column vector of parameters, $`\stackrel{}{t}`$ the column vector with components $`t_k`$, and $`\theta `$ the $`3`$vector of phase parameters, we have $`\stackrel{}{s}=M_3\left(\theta \right)\stackrel{}{t}`$, where
$$M_3\left(\theta \right)=\frac{1}{3}\left(\begin{array}{ccc}1& 1& 1\\ 1& \eta ^2& \eta \\ 1& \eta & \eta ^2\end{array}\right)\left(\begin{array}{ccc}e^{i\theta _0}& 0& 0\\ 0& e^{i\theta _1}& 0\\ 0& 0& e^{i\theta _2}\end{array}\right)\left(\begin{array}{ccc}1& 1& 1\\ 1& \eta & \eta ^2\\ 1& \eta ^2& \eta \end{array}\right)=\underset{k=0}{\overset{2}{}}f(k,\theta )S_{0,k}\text{.}$$
(21)
If $`\theta +\varphi `$ is defined as component-wise addition, then it is easy to check that $`\left\{M_3\left(\theta \right)\right\}`$ also defines an Abelian group of invertible mappings,
$`M_3\left(\theta \right)M_3\left(\varphi \right)=M_3\left(\theta +\varphi \right)\text{,}`$
giving the functional equation $`_kf(k,\theta )f(jk,\varphi )=f(j,\theta +\varphi )`$ in analogy with the corresponding result when $`d=2.`$ We have thus established a correspondence between all trace one projection matrices with given diagonal and the range of a two parameter family of mappings acting on $`\stackrel{}{t}`$. (We are endebted to Rasmus Hansen for bringing to our attention , which contains an analysis of the geometry of the convex space of $`d=3`$ densities. The pre-tranform characterization of the projections associated with a given diagonal is similar to the results derived here.)
In the $`d=2`$ case the choices of $`m_z=\pm 1`$ produce special cases of projections, and the same is true when $`d=3`$. If $`s_{1,0}`$ is one of the extreme points $`1`$, $`\eta `$, or $`\eta ^2`$, then two of the $`b_k`$’s equal zero and all of the $`s_{k,1}`$’s equal zero. It follows that for $`r=0`$, $`1`$, and $`2`$, $`\frac{1}{3}\left[S_{0,0}+\eta ^rS_{1,0}+\left(\eta ^rS_{1,0}\right)^{}\right]`$ is a trace one projection, and those are the three subgroup projections $`P_{1,0}(r).`$ A degeneracy which has no analogue in the $`d=2`$ case occurs when $`s_{1,0}`$ lies between two extreme points on an edge. Then exactly one of the $`b_k`$’s equals zero, and there is a one parameter family of projections associated with $`s_{1,0}`$. The most interesting cases occur when $`s_{1,0}`$ lies in the interior of the equilateral triangle. In particular when $`s_{1,0}=0`$, the $`b_k`$’s are equal to $`1/\sqrt{3}`$, and by choosing the components of $`\theta `$ appropriately from $`\{0,2\pi /3,4\pi /3\}`$ we find the remaining subgroup projections $`P_u(r).`$ Thus, our entire analysis of separability in the $`d=3`$ case uses only the projections associated with the origin and with the vertices of the equilateral triangle. |
warning/0001/quant-ph0001052.html | ar5iv | text | # Canonical Quantum Teleportation
(April 26, 1999; Revised July 13, 1999)
## Abstract
Canonically conjugated observables such as position-momentum and phase-number are found to play a 3-fold role in the drama of the quantum teleportation. Firstly, the common eigenstate of two commuting canonical observables like phase-difference and number-sum provides the quantum channel between two systems. Secondly, a similar pair of canonical observables from another two systems is measured in the Bell operator measurements. Finally, two translations generated by the canonically conjugated observables of a single system constitute the local unitary operation to recover the unknown state. In addition, the necessary and sufficient condition is presented for a reliable quantum teleportation of finite-level systems.
The quantum teleportation bent , a disembodied transmission of quantum state, has been demonstrated in several experiments both for finite-level systems exp and continuous variables vaid1 ; brn1 ; fur . Along with the resulting discussions brnd ; vaid2 about its experimental realization, many other aspects such as general schemes bent2 ; tor ; stig and some applications chiral of the quantum teleportation have also been investigated. All these investigations so far emphasize mainly on the states of the systems. In this Letter we shall show the fundamental roles played by the canonically conjugated ($`c.c.`$) observables in the drama of the quantum teleportation in order to reveal the physical contents of its basic ingredients.
Generally speaking, the quantum teleportation consists of three basic steps: (i) To prepare two systems in an Einstein-Podolsky-Rosen (EPR) entangled state or a Bell state and send them to two different places to establish a quantum channel; (ii) At one place, to perform the so-called joint Bell operator measurements with respect to one system involved in the EPR entanglement and a third system at an unknown state to be transferred; (iii) At the other place, to perform necessary unitary operations to the other system involved in the EPR entanglement according to the outcomes of Bell operator measurements. By this means the unknown state is transferred from one place to another.
In the case of continuous variables, three similar systems 1,2 and 3 are considered, which are described by canonical observables $`\widehat{x}_a,\widehat{p}_a`$ $`(a=1,2,3)`$ satisfying canonical commutation rules
$$[\widehat{x}_a,\widehat{p}_b]=i\delta _{ab},(a,b=1,2,3).$$
(1)
System 1 and 2 are prepared in a common eigenstate of the position-difference $`\widehat{x}_1\widehat{x}_2`$ and the momentum-sum $`\widehat{p}_1+\widehat{p}_2`$ corresponding to eigenvalues $`x_{12}`$ and $`p_{12}`$ brn2 :
$$|x_{12};p_{12}=e^{i\widehat{p}_1\widehat{x}_2}|x_{12}_1|p_{12}_2,$$
(2)
where $`|x_{12}_1`$ is an eigenstate of $`\widehat{x}_1`$ with eigenvalue $`x_{12}`$ and $`|p_{12}_2`$ is an eigenstate of $`\widehat{p}_2`$ with eigenvalue $`p_{12}`$. And system 3 is in an unknown state $`|\psi _3`$ to be teleported to the first system.
Then a kind of Bell operator measurement measuring the position-difference $`\widehat{x}_2\widehat{x}_3`$ and the momentum-sum $`\widehat{p}_2+\widehat{p}_3`$ is performed on systems 2 and 3. This measurement projects systems 2 and 3 to one of the common eigenstates $`|x_{23};p_{23}`$ of $`\widehat{p}_2+\widehat{p}_3`$ and $`\widehat{x}_2\widehat{x}_3`$ with $`x_{23}`$ and $`p_{23}`$ taking values on the real line uniformly. Accordingly, system 1 is transformed into state $`𝒪_c^{}|\psi _1`$ where
$$𝒪_c=e^{ip_{23}x_{12}}e^{ip_{13}\widehat{x}_1}e^{ix_{13}\widehat{p}_1}$$
(3)
with $`p_{13}=p_{12}p_{23}`$, $`x_{13}=x_{12}+x_{23}`$. At this central stage, the measured observables are exactly two commuting canonical observables: momentum-sum and position-difference. According to the outcomes $`x_{23}`$, $`p_{23}`$ of the measurement and values $`x_{12},p_{12}`$ known from the state preparation, one is able to perform unitary operation $`𝒪_c`$ to system 1. And system 1 is then at the unknown state though no one knows what the unknown state is.
Since $`p_{12}`$ and $`p_{23}`$ are the momentum-sums of corresponding systems, $`p_{13}=p_{12}p_{23}`$ is naturally the momentum-difference between systems 1 and 3. Similarly $`x_{13}=x_{12}+x_{23}`$ can be viewed as the position-difference between systems 1 and 3. The unitary operation $`𝒪_c`$, being made up of two successive translations up to a phase factor, has therefore a natural physical meaning: it compensates the position-difference and momentum-difference between systems 1 and 3. This obvious fact was already noticed in Ref. vaid1 where the teleportation of continuous variables was first proposed.
We see clearly that the $`c.c.`$ observables, position and momentum in this case, play a 3-fold role in the drama of the quantum teleportation of continuous variables. Firstly the common eigenstate of two commuting canonical observables, e.g. the position-difference and the momentum-sum, provides the quantum channel between two systems. Secondly the same commuting canonical pair of another two systems is measured in the Bell operator measurement. Finally the $`c.c.`$ observables of a single system generate two translations, which make up of the unitary operation to recover the unknown state. So the quantum teleportation deserves the name canonical quantum teleportation.
Given one pair of $`c.c.`$ observables one may design one possible canonical quantum teleportation with exactly those three steps. Notice that in the procedure of quantum teleportation the real position and momentum cannot be used because localization of the particle is required. In fact in the recent experimental realization of the quantum teleportation of continuous variables brn1 , a pair of $`c.c.`$ observables of the photon field, phase quadrature and number quadrature, have been used.
At the very first look, in the case of the finite-level systems those three steps of the quantum teleportation seem to be three unrelated procedures: Bell states preparation, Bell operator measurements brna or nonlocal measurements vaid1 and special unitary operations, whose physical meanings need clarifying. We shall then demonstrate that there is also a pair of $`c.c.`$ observables that plays the same 3-fold role for finite-level systems. As it turns out, one observable is the number operator and the other one is the phase operator of a finite-level system.
For an infinite-level system as simple as a quantum harmonic oscillator, a Hermitian phase operator does not exist drc ; lynch ; index0 . After a series of efforts to solve this problem newton ; bant ; luis0 ; pd it was clear recently that the quantum phase of a harmonic oscillator can only be described by means of the phase-difference between two systems with a rational-number-type of spectrum and the quantized phase-difference obeys a quantum addition rule pd ; index . Among the early approaches to this dilemma, the truncated Hilbert space approach proposed by Pegg and Barnett bant describes in de facto the phase variable of a finite-level system instead of a harmonic oscillator with infinite many energy levels. This approach was also investigated in some details by others vds ; luis1 .
For an $`s`$-level system $`A`$, the number operator $`𝒩_A`$ has spectrum $`Z_s=\{0,1,\mathrm{},s1\}`$ and its eigenstates $`|n_A`$ with $`nZ_s`$ span the Hilbert space of the system. In this Hilbert space, taking the phase window as $`[0,2\pi )`$, one can define the exponential phase operator as
$$e^{in𝒫_A}=\underset{mZ_s}{}|m+n_Am|,nZ_s.$$
(4)
Here the state $`|ks+n_A`$ is identified with the state $`|n_A`$ whenever $`k`$ is an integer. This identification seems to be trivial enough for a single system, but it is crucial for the combination of number operators from different systems. The so-defined exponential operator is obvious unitary which leads to a Hermitian phase operator $`𝒫_𝒜`$ with spectrum $`\mathrm{\Xi }_s=\{2m\pi /s|m=0,1,\mathrm{},s1\}`$ and eigenstates
$$|\theta _A=\frac{1}{\sqrt{s}}\underset{nZ_s}{}e^{in\theta }|n_A,\theta \mathrm{\Xi }_s.$$
(5)
The motivation to define a Hermitian phase operator is, analogous to the well-known canonical position and momentum, to find the $`c.c.`$ partner for the number operator. However, the canonical relationship between the quantum phase and number cannot be explicitly manifested through their commutator. The quantum phase and number have a very complicated commutator bant due to the fact that the phase variable has a curved configure space because of its the periodicity, which is also the origin of the rational-number-type of spectrum of quantized phase difference pd . Only when the unitary operations instead of Hermitian observables are considered, dose the canonical relationship between the phase and number manifest itself manif . As shown explicitly in Eq.(2) and Eq.(3) it is also the operations represented by unitary operators instead of the observables represented by Hermitian operators that plays the main roles in the case of continuous variables.
As is well known, the unitary operations generated by position and momentum, which represent the translations in the momentum and configuration spaces respectively, satisfy the Weyl form of commutation relation
$$e^{ix\widehat{p}}e^{ip\widehat{x}}e^{ix\widehat{p}}e^{ip\widehat{x}}=e^{ixp}.$$
(6)
This kind of relation indicates also the canonical relationship, even more intrinsically than the commutator. This is because the exponential phase and number operators satisfy also a similar relation
$$e^{i\theta 𝒩_A}e^{in𝒫_A}e^{i\theta 𝒩_A}e^{in𝒫_A}=e^{in\theta }.$$
(7)
In this sense the quantum phase and number operator are $`c.c.`$ observables. The exponential phase-difference and number-difference operators of two quantum harmonic oscillators satisfy also this kind of relation which yield another pair of $`c.c.`$ observables manif .
As relation Eq.(6) indicates that the operator $`e^{ix\widehat{p}}`$ represents a translation by $`x`$ in the configuration space, so the relation Eq.(7) ensures that the exponential phase operator $`e^{in𝒫_A}`$ represents also a translation by $`n`$ (modular $`s`$) of the number. Similarly, the exponential number operator $`e^{i\theta 𝒩_A}`$ represents a translation by $`\theta `$ (modular $`2\pi `$) of the quantum phase. These are exactly the physical contents of these two unitary operations.
The quantum phase and phase differences were found to observe a quantum addition rule pd , which assures another quantum phase or phase difference with the same kind of spectrum. The quantum addition of phase operators $`𝒫_A`$ and $`𝒫_B`$ of two $`s`$-level systems $`A`$ and $`B`$, since they are commuting, is simply $`𝒫_A\dot{}𝒫_B𝒫_A𝒫_B`$ modular $`2\pi `$. Similarly, to preserve the spectrum of the number operator, the quantum number-sum can be defined as $`𝒩_A\dot{+}𝒩_B𝒩_A+𝒩_B`$ modular $`s`$. Because the quantum phase-difference and number-sum are commuting, they possess common eigenstates
$$|\theta _{AB};n_{AB}=e^{i𝒩_A𝒫_B}|\theta _{AB}_A|n_{AB}_B,$$
(8)
where $`|\theta _{AB}_A`$ is the eigenstate of $`𝒫_A`$ with eigenvalue $`\theta _{AB}\mathrm{\Xi }_s`$ and $`|n_{AB}_B`$ is the eigenstate of $`𝒩_B`$ with eigenvalue $`n_{AB}Z_s`$. They form a complete and orthonormal basis of systems $`A`$ and $`B`$. These two observables are measurable in the framework of nonlocal measurements vaid1 .
Now that a complete analogue between the well-known $`c.c.`$ observables, position and momentum, and the less obviously c.c. observables, quantum phase and number, has been established, we can formulate the quantum teleportation of finite-level systems in the same canonical manner. As a quantum channel of the quantum teleportation of finite-level systems, systems $`A`$ and $`B`$ are prepared in a common eigenstate $`|\theta _{AB};n_{AB}`$ of their quantum phase-difference and number-sum.
Suppose that another $`s`$-level system $`C`$ is in an unknown state $`|\varphi _C`$ which will be teleported to the system $`A`$. To this end we perform a joint measurement of the quantum phase-difference $`𝒫_B\dot{}𝒫_C`$ and the number-sum $`𝒩_B\dot{+}𝒩_C`$ of the systems $`B`$ and $`C`$. With probability $`1/s^2`$, the total state of the whole system $`|\mathrm{\Phi }=|\theta _{AB};n_{AB}|\varphi _C`$ is projected to state $`𝒪_s^{}|\varphi _A\theta _{BC};n_{BC}|\mathrm{\Phi }`$ where
$$𝒪_s=e^{in_{BC}\theta _{AB}}e^{in_{AC}𝒫_A}e^{i\theta _{AC}𝒩_A}$$
(9)
with $`\theta _{AC}=\theta _{AB}+\theta _{BC}`$ and $`n_{AC}=n_{AB}n_{BC}`$ after the measurement. The number-sum $`n_{BC}`$ takes value in $`Z_s`$ and the phase-difference $`\theta _{BC}`$ takes values in $`\mathrm{\Xi }_s`$ with equal probability, which label the $`s^2`$ outcomes of the measurements.
After knowing these phase-differences $`\theta _{AB}`$, $`\theta _{BC}`$ and number-sums $`n_{AB}`$, $`n_{BC}`$, one can perform a unitary transformation $`𝒪_s`$ to system $`A`$ so that the unknown state of system $`C`$ appears at the other end of the quantum channel. We note that operation $`𝒪_s`$ is made up of an exponential phase operator and an exponential number operator up to a phase factor. From the discussions above we know that these two operations represent a phase translation by values $`\theta _{AC}`$ and a number translation by values $`n_{AC}`$. Because $`\theta _{AC}`$ can be regarded as the phase-difference and $`n_{AC}`$ as the number-difference between systems $`A`$ and $`C`$, the meaning of these two unitary operations become now clear: before the unknown state can be recovered the phase-difference and number-difference between systems $`A`$ and $`C`$ must be compensated.
Consider the simple case of 2-level systems, where we identify state $`|0`$ with $`|`$ and state $`|1`$ with $`|`$. As the quantum channel we prepare systems $`A`$ and $`B`$ in the state as in Eq.(8) with $`\theta _{AB}=\pi ,n_{AB}=1`$. Four possible outcomes of the Bell operator measurements on systems $`B`$ and $`C`$ are labeled by phase-difference $`\theta _{BC}=0,\pi `$ and number-sum $`n_{BC}=0,1`$. We can see that four corresponding unitary operations $`𝒪_2`$ in Eq.(9) applied to system $`A`$ are exactly the same as those in Ref. bent .
Canonical transformations, which preserve the canonical commutators among observables as in Eq.(1) or relations such as Eq.(7) of corresponding unitary operations, can be performed to $`c.c.`$ observables. Some canonical transformations can result in some new forms of quantum teleportations. The simplest case is to make a canonical transformation only to system $`B`$, for example, $`𝒫_B𝒫_B`$ and $`𝒩_B𝒩_B`$, which results a quantum teleportation as follows. The quantum channel is a common eigenstate of $`𝒫_A\dot{+}𝒫_B`$ and $`𝒩_A\dot{}𝒩_B`$, e.g. state
$$|\mathrm{\Psi }_{AB}=\frac{1}{\sqrt{s}}\underset{mZ_s}{}|m_A|m_B$$
(10)
corresponding to zero number-difference and zero phase-sum. The observables measured in the second step are $`𝒫_B\dot{+}𝒫_C`$ and $`𝒩_B\dot{}𝒩_C`$. And the final operation Eq.(9) to recover the unknown state remains unchanged. This scheme is exactly the original teleportation of systems with more than 2 levels discussed in Ref. bent . One notes that when $`s=2`$ the quantum phase-difference and number-sum are identical with quantum phase-sum and number-difference respectively, therefore these two teleportation schemes are identical in the case of $`s=2`$.
Now we try to take a general pure state of systems $`A`$ and $`B`$ as our quantum channel. Any normalized state can be expressed as $`T|\mathrm{\Psi }_{AB}`$ where operator $`T`$ acts only on system $`A`$ with $`\mathrm{Tr}(T^{}T)=s`$. Then we perform a general Bell operator measurement on systems $`B`$ and $`C`$. This is equivalent to projection to some orthonormal basis of systems $`B`$ and $`C`$
$$|k;l=\frac{1}{\sqrt{s}}\underset{mZ_s}{}|m_B𝒪_{kl}|m_C,$$
(11)
where $`s^2`$ operators $`𝒪_{kl}`$ act only on a single system and satisfy the following normalization conditions
$$\mathrm{Tr}\left(𝒪_{kl}𝒪_{k^{}l^{}}^{}\right)=s\delta _{kk^{}}\delta _{ll^{}},k,k^{},l,l^{}Z_s.$$
(12)
Numbers $`k,l`$ label all possible outcomes of the measurements. Given outcomes $`k,l`$ of the measurements, appearing with equal probability, system $`A`$ is found to be in state
$$s^2k;l|T|\mathrm{\Psi }_{AB}|\varphi _C=T𝒪_{kl}^{}|\varphi _A,$$
(13)
where operator $`𝒪_{kl}`$ is now acting on system $`A`$. The only requirement for a reliable quantum teleportation is therefore to have $`T𝒪_{kl}^{}`$ unitary, which infers that $`T`$ must be reversible. From Eq.(12) one obtains Tr$`((T^{}T)^1)=s`$, which is compatible with $`\mathrm{Tr}(T^{}T)=s`$ iff $`T`$ is unitary. Therefore to have $`T𝒪_{kl}^{}`$ unitary is equivalent to have all the operators $`T`$ and $`𝒪_{kl}`$ unitary. This is the necessary and sufficient condition for a reliable quantum teleportation. In other words the quantum channel must be a maximum entangled state and the measurements must be projections to maximum entangled states. And the recovering operation at the final stage is simply $`𝒪_{kl}T^{}`$ depending on the outcomes of the measurements.
As one wishes, from orthonormal bases $`|k;l`$ one can construct two commuting canonical observables like phase-difference and number-sum, whose common eigenstates are exactly these bases. As a result, the measured observables in the second step of the quantum teleportation may be different from the observables determines the quantum channel. For example, the quantum channel may be provided by the common eigenstate of the quantum phase-sum and number-difference and the quantum phase-difference and number-sum are the Bell operators. By this means one can also teleport an unknown state from one place to another. The general scheme discussed in Ref.stig is included here as a special example.
The continuous variables case can be analyzed similarly. Let us fix our measurements at the second step to the projections to states $`|x_{23},p_{23}`$. All the pure states that can be used as quantum channel should have form $`_{n=0}^{\mathrm{}}D^{}|n_1|n_2`$ where $`D`$ is an arbitrary unitary operator acting on system 1 only. The operation at the final stage is $`M^{}D`$ where $`M`$ is a unitary operator acting on system 1 with elements $`m|M|n=x_{23},p_{23}|m,n`$ where $`|m,n=|m_1|n_2`$ denote the number state bases with $`m,n`$ going from zero to infinity. When the elements of $`D`$ are taken as $`m|D|n=x_{12},p_{12}|m,n`$, the teleportation of continuous variables discussed at the beginning is regained. This discrete formulation of the quantum channel upto a normalization constant was noticed in Ref. enk .
When one consider three quantum harmonic oscillators, although the quantized phase-differences between each two of them are well defined, it is impossible to perform a quantum teleportation using the quantized phase difference and number-sum. This is because the exponential phase operator of a single oscillator, which ought to be employed to compensate a number-difference at the final stage of the quantum teleportation, dose not exist.
In conclusion, the quantum teleportation is characterized by $`c.c.`$ observables completely: The quantum channel is provided by the common eigenstate of two commuting canonical observables, the Bell operator measurement measures a similar pair of canonical observables and the recovering operation consists of two translations generated by the $`c.c.`$ observables. By applying suitable canonical transformations to the $`c.c.`$ observables, one can design new schemes of quantum teleportation. The necessary and sufficient condition for a reliable quantum teleportation of finite systems is to have a maximum entangled state as quantum channel and the Bell operator measurements are projections to maximum entangled states. The nonexistence of certain $`c.c.`$ observables makes the quantum teleportation using these variables impossible. All these investigations concern the ideal quantum teleportation. In the real experiments where non-ideal elements must be considered, it becomes ambiguous how to characterize quantum teleportation. In this aspect some efforts have been made lam . The attention to the roles played by the $`c.c.`$ observables in the drama quantum teleportation may help to establish such kinds of criteria both for the continuous and discrete variables.
One of the authors (S. Yu) gratefully acknowledges the support of K. C. Wong Education Foundation for postdoctors, Hong Kong. This work is also partially supported by the NSF of China. |
warning/0001/hep-ph0001096.html | ar5iv | text | # 1 ℎ(𝑥) extracted from CCFR’97 data
## Abstract
The summary of the results of our next-to-next-to-leading fits of the Tevatron experimental data for $`xF_3`$ structure function of the $`\nu N`$ deep-inelastic scattering is given. The special attention is paid to the extraction of twist-4 contributions and demonstration of the interplay between these effects and higher order perturbative QCD corrections. The factorization and renormalization scale uncertainties of the results obtained are analysed.
DTP/00/02
January 2000
The interplay between perturbative QCD and power corrections: the description of scaling or automodelling limit violation in deep-inelastic scattering
A.L. Kataev<sup>a,</sup><sup>1</sup><sup>1</sup>1Supported by UK Royal Society and in part by the Russian Foundation of Basic Research, Grant N 99-01-00091, G. Parente<sup>b,</sup><sup>2</sup><sup>2</sup>2Supported by CICYT (AEN96-1773) and Xunta de Galicia (Xuga-20602B98) A.V. Sidorov<sup>c,</sup><sup>3</sup><sup>3</sup>3Supported in part by the Russian Foundation of Basic Research, Grant N 99-01-00091
(a) Centre for Particle Theory of the University of Durham,
DH1 3LE, Durham, United Kingdom and
Institute for Nuclear Research of the Academy of Sciences of Russia,
117312 Moscow, Russian Federation<sup>4</sup><sup>4</sup>4Permanent address
(b) Department of Particle Physics, University of Santiago de Compostela,
15706 Santiago de Compostela, Spain
(c) Bogolyubov Laboratory of Theoretical Physics, Joint Institute for Nuclear Research, 141980 Dubna, Russian Federation
Contributed to Bogolyubov Conference
“Problems of Theoretical and Mathematical Physics”
Moscow-Dubna-Kyiv, 27 September - 6 October 1999
1. The study of deep-inelastic scattering (DIS) processes has a rather long and inspiring history. One of the first realizations that the analysis of $`\nu N`$ DIS could play an important role in investigations of the properties of the nucleon came in Ref.. The fundamental concept of scaling of DIS structure functions (SFs) has lead to many subsequent investigations. Other important stages in the development of both theoretical and experimental studies of various characteristics of DIS processes in this productive period were reviewed in detail recently . In particular, it was stressed that after the experimental confirmation of scaling and indications of the existence of point-like constituents of the nucleon, the more rigorous theoretical explanation of the behaviour of DIS form factors came onto the agenda. A series of works by N. N. Bogolyubov and coauthors , were devoted to the development of the new method, which made it possible to analyse the asymptotics of the form factors of $`eN`$ DIS using the Jost-Lehmann-Dyson integral representation, and explain the property of scaling (or as called it by the authors of Ref. “automodelling”) behaviour of the corresponding SFs in the framework of general principles of local quantum field theory .
We now know that this property is true only in the asymptotic regime and that it is violated within the framework of QCD (see e.g. the extensive discussions in a number of books on the subject ). Indeed, the theory of QCD predicts that scaling or automodelling behaviour of SFs is violated by the logarithmically decreasing perturbative QCD contributions to the leading twist operators. However, in the intermediate and low $`Q^2`$ regime the higher twist operators, which give rise to scaling violations of the form $`1/Q^2`$, $`1/Q^4`$, etc., might also be important . Indeed, the NLO DGLAP fits of the BCDMS data of DIS of charged leptons on nucleons and reanalysed SLAC $`eN`$ data resulted in the detection of the signals from the twist-4 contributions.
During the last few years there has been considerable progress in modeling these effects with the help of the infrared renormalon (IRR) approach (for the review see Ref.) and the dispersive method (see also Ref.). Using these methods the authors of Ref. explained the behaviour of the twist-4 contributions to the $`F_2`$ SF observed in Ref. and constructed a model for the similar power-suppressed corrections to $`xF_3`$ SF. In view of this it became important to check the predictions of Ref. and to study the possibility of extracting higher-twist contributions from the new more precise experimental data for $`\nu N`$ DIS, obtained by the CCFR collaboration at Fermilab Tevatron , and also to exploit the considerable progress in calculations of the perturbative QCD corrections to characteristics of DIS, achieved in the last decade.
Indeed, the analytic expressions for the next-to-next-to-leading order (NNLO) perturbative QCD corrections to the coefficient functions of SFs $`F_2`$ and $`xF_3`$ are now known. Moreover, the expressions for the NNLO corrections to the anomalous dimensions of non-singlet (NS) even Mellin moments of $`F_2`$ SF with $`n=2,4,6,8,10`$ and for the N<sup>3</sup>LO corrections to the coefficient functions of these moments are also available . In this report we will summarize the results of the series of the works of Refs.-, devoted to the analysis of the CCFR data at NNLO, which has the aim to determine the NNLO value of the QCD coupling constant $`\alpha _s(M_Z)`$ and to extract the effects of the twist-4 contributions to SF $`xF_3`$ . In particular, we will concentrate on the discussion of the factorization and renormalization scale uncertainties of the results obtained.
2. Our analysis of Refs.\- is based on reconstruction of the SF $`xF_3`$ from its Mellin moments $`M_n(Q^2)=_0^1x^{n1}F_3(x,Q^2)𝑑x`$ using the Jacobi polynomials method, proposed in Ref. and further developed in the works of Ref.. Within this framework one has:
$$xF_3(x,Q^2)=x^\alpha (1x)^\beta \underset{n=0}{\overset{N_{max}}{}}\mathrm{\Theta }_n^{\alpha ,\beta }(x)\underset{j=0}{\overset{n}{}}c_j^{(n)}(\alpha ,\beta )M_{j+2}(Q^2)$$
(1)
where $`\mathrm{\Theta }_n^{\alpha ,\beta }`$ are the Jacobi polynomials, $`c_j^{(n)}(\alpha ,\beta )`$ are combinatorial coefficients given in terms of Euler $`\mathrm{\Gamma }`$-functions of the $`\alpha `$ and $`\beta `$ weight parameters. In view of the reasons discussed in Ref., they were fixed to 0.7 and 3 respectively. The QCD evolution of the moments is defined by the solution of the corresponding renormalization group equation:
$$\frac{M_n(Q^2)}{M_n(Q_0^2)}=exp\left[_{A_s(Q_0^2)}^{A_s(Q^2)}\frac{\gamma _{NS}^{(n)}(x)}{\beta (x)}𝑑x\right]\frac{C_{NS}^{(n)}(A_s(Q^2))}{C_{NS}^{(n)}(A_s(Q_0^2)}$$
(2)
The QCD running coupling constant enters this equation through $`A_s(Q^2)=\alpha _s(Q^2)/(4\pi )`$ and is defined as the expansion in terms of inverse powers of $`ln(Q^2/\mathrm{\Lambda }_{\overline{MS}}^{(4)2})`$. For the initial scale $`Q_0^2`$, from which the evolution is started, the moments in Eq.(2) were parametrized as $`M_n(Q_0^2)=_0^1x^{n2}A(Q_0^2)x^{b(Q_0^2)}(1x)^{c(Q_0^2)}(1+\gamma (Q_0^2)x)𝑑x`$. In the process of our analysis we took into account both target mass corrections and twist-4 contributions. The latter were modeled using the IRR approach as $`M_n^{IRR}=C(n)M_n(Q^2)A_2^{^{}}/Q^2`$ and by adding into the r.h.s. of Eq.(1) the term $`h(x)/Q^2`$ with $`h(x)`$ considered as a free parameter for each $`x`$-bin of the experimental data.
For arbitrary factorization and renormalization scales the NNLO expression for the NS Mellin moments reads:
$$M_n(Q^2)(A_s(Q^2k_F))^a\times \overline{AD}(n,A_s(Q^2k_F))\times C_{NS}^{(n)}(A_s(Q^2k_R))$$
(3)
where $`a=\gamma _{NS}^{(0)}/(2\beta _0)`$, $`\overline{AD}=1+\left[p(n)+ak_1^F\right]A_s(Q^2k_F)+\left[q(n)+p(n)(a+1)k_1^F+(\beta _1/\beta _0)ak_1^F+a(a+1)(k_1^F)^2/2\right]A_s^2(Q^2k_F)`$ and $`C_{NS}^{(n)}=1+C^{(1)}(n)A_s(Q^2k_R)+\left[C^{(2)}(n)+C^{(1)}(n)k_1^R\right]A_s^2(Q^2k_R)`$. Here $`\gamma _{NS}^{(0)}`$, $`\beta _0`$ and $`\beta _1`$ are the scheme-independent coefficients of the anomalous dimension function $`\gamma _{NS}(x)`$ and QCD $`\beta `$-function $`\beta (x)`$, $`p(n)`$ and $`q(n)`$-terms are expressed through the NLO and NNLO coefficients of $`\gamma _{NS}(x)`$ and $`\beta (x)`$ via equations, given in Refs.. Within the $`\overline{MS}`$-like schemes the factorization and renormalization scale ambiguities are parameterized by the terms $`k_1^F=\beta _0ln(k_F)`$ and $`k_1^R=\beta _0ln(k_R)`$, where $`k_F`$ ($`k_R`$) is the ratio of the factorization (renormalization) scale and the scale of the $`\overline{MS}`$-scheme. Following the analysis of Ref. we take $`k_R=k_F=k`$, fixing identically the factorization scale and the renormalization scale. We performed our fits for the case of $`k=1`$ (namely, in the pure $`\overline{MS}`$-scheme) and then determine the scale uncertainties of $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$, the twist-4 parameter $`A_2^{^{}}`$ and the $`x`$-shape of $`h(x)`$ by choosing $`k=1/4`$ and $`k=4`$ and repeating the fits for these two cases.
3. In the process of our analysis of CCFR’97 data we applied the same kinematic cuts as in Ref., namely $`Q^2>5GeV^2`$, $`x<0.7`$ and $`W^2>10GeV^2`$. We started the QCD evolution from the initial scale $`Q_0^2=20GeV^2`$, which we consider as more appropriate from the point of view of stability of the NLO and NNLO results for $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$ due to variation of the initial scale . In order to estimate the uncertainties of the NNLO results, we also performed the N<sup>3</sup>LO fits with the help of the expanded Padé approximations technique (for the detailed discussions see Ref.). The results are presented in Table 1.
| | $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$ (MeV) | $`A_2^{^{}}`$ (GeV<sup>2</sup>) | $`\chi ^2`$/points |
| --- | --- | --- | --- |
| LO | 264$`\pm `$37 | | 113.1/86 |
| | 433$`\pm `$53 | -0.33$`\pm `$0.06 | 83.1/86 |
| | 331$`\pm `$162 | h(x) in Fig.1 | 66.3/86 |
| NLO | 339$`\pm `$42 | | 87.6/86 |
| | 369$`\pm `$39 | -0.12$`\pm `$0.06 | 82.3/86 |
| | 440$`\pm `$183 | h(x) in Fig.1 | 65.8/86 |
| NNLO | 326$`\pm `$35 | | 77.0/86 |
| | 327$`\pm `$35 | -0.01$`\pm `$0.05 | 76.9/86 |
| | 372$`\pm `$133 | h(x) in Fig.1 | 65.0/86 |
| N<sup>3</sup>LO | 332$`\pm `$28 | | 76.9/86 |
| Pade | 333$`\pm `$27 | -0.04$`\pm `$0.05 | 76.3/86 |
| | 371$`\pm `$127 | h(x) in Fig.1 | 64.8/86 |
Table 1. The results of the fits of CCFR’97 data with the cut $`Q^2>5GeV^2`$.
At NLO the value for $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$ is in good agreement with the NLO result $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}=337\pm 28MeV`$, obtained by the CCFR collaboration with the help of DGLAP NLO analysis of both $`F_2`$ and $`xF_3`$ SFs data in the case when HT-corrections were neglected . The obtained NLO value of the IRR-model parameter $`A_2^{^{}}`$ is in agreement with the estimates of Ref. and of Ref. especially. However, at NNLO a significant decrease of the magnitude of the parameter $`A_2^{^{}}`$ is observed. In view of this the results for $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$ obtained at the NNLO without HT corrections and with IRR-model of twist-4 term almost coincide. A similar tendency was observed in the process of the N<sup>3</sup>LO Padé fits. To study this feature in more detail we extracted the $`x`$-shape of the model-independent function $`h(x)`$ (see Fig.1) and analysed the factorization/renormalization scale uncertainties of the outcomes of our fits . The corresponding results are presented in Table 2 where $`\mathrm{\Delta }_k`$ is defined as $`\mathrm{\Delta }_k=\mathrm{\Lambda }_{\overline{MS}}^{(4)}(k)\mathrm{\Lambda }_{\overline{MS}}^{(4)}(k=1)`$. The related $`x`$-shapes of $`h(x)`$ are presented in Fig.2.
| Order | $`k`$ | $`\mathrm{\Delta }_k`$ (MeV) | $`A_2^{^{}}`$ (GeV<sup>2</sup>) | $`\chi ^2`$/points |
| --- | --- | --- | --- | --- |
| NLO | 4 | 116 | | 99.1/86 |
| | 4 | 213 | -0.22$`\pm `$0.006 | 84.2/86 |
| | 1/4 | -61 | | 80.4/86 |
| | 1/4 | -99 | +0.02$`\pm `$0.005 | 80.2/86 |
| NNLO | 4 | 35 | | 83.5/86 |
| | 4 | 66 | -0.11$`\pm `$0.06 | 83.5/86 |
| | 1/4 | -51 | | 87.3/86 |
| | 1/4 | -45 | +0.09$`\pm `$0.05 | 84.5/86 |
Table 2. The results of NLO and NNLO fits of CCFR’97 data for different values of factorization/renormalization scales.
4. We will concentrate first on discussing the presented behaviour of the twist-4 parameter $`h(x)`$ of $`xF_3`$ SF, presented in Figs.1,2. In the case of $`k=1`$, namely in the pure $`\overline{MS}`$-scheme, $`x`$-shape of $`h(x)`$ obtained from the LO and NLO analysis of Refs. is in agreement with the IRR-model predictions of Ref.. Note also that the combination of the quark counting rules with the results of Ref. predict the following $`x`$-form of $`h(x)`$: $`h(x)A_2^{^{}}(1x)^2`$. Taking into account the negative values of $`A_2^{^{}}`$, obtained in the process of our LO and NLO fits (see Table 1), we conclude that the related behaviour of $`h(x)`$ is in qualitative agreement with these predictions.
At the NNLO the situation is more intriguing. Indeed, though a certain indication of the twist-4 term survives even at this level, the NNLO part of Fig.1 demonstrates that its extracted $`x`$-shape starts to deviate both from the IRR prediction of Ref. and from the quark-parton model picture, mentioned above. Notice also that within the statistical error bars the NNLO value of $`A_2^{^{}}`$ is indistinguishable from zero. These conclusions are confirmed by the studies of the factorization/renormalization scale dependence of the NLO and NNLO outcomes of the fits .
Indeed, it is known that the variation of the related scales is simulating in part the effects of the higher-order perturbative QCD corrections. In view of this the NLO (NNLO) results, obtained in the case of $`k=1/4`$ (see Table 2 and Fig.2 in particular), are almost identical to the NNLO (Padé motivated N<sup>3</sup>LO) extractions of $`h(x)`$ and of the IRR model parameter $`A_2^{^{}}`$ from the fits with $`k=1`$ (see Fig.1 and Table 1). Thus, we conclude, that as the result of analysis of the CCFR’97 data the NNLO and beyond we observe the minimization of the twist-4 contributions to $`xF_3`$ SF. This feature is related to the interplay between NNLO perturbative QCD and twist-4 $`1/Q^2`$ corrections. The recent studies of the scale-dependence of the NLO DGLAP extraction of the twist-4 terms from different recent DIS experimental data are supporting the foundations of Refs.. This means that the higher-twist parameters cannot be defined independently of the effects of perturbation theory and that the NNLO corrections can mimick the contributions of higher twists provided the experimental data is not precise enough for the clear separation of the nonperturbative from perturbative effects. Thus, it is highly desirable to have new experimental data for $`xF_3`$ SF, which are more precise than the ones given by the CCFR collaboration.
In conclusion we present also the NLO and NNLO values of $`\alpha _s(M_Z)`$, obtained by us in Ref. from the fits of CCFR’97 data for $`xF_3`$ SF with twist-4 terms modeled through the IRR approach :
$`NLO\alpha _s(M_Z)`$ $`=`$ $`0.120\pm 0.003(stat)\pm 0.005(syst)_{0.007}^{+0.009}`$ (4)
$`NNLO\alpha _s(M_Z)`$ $`=`$ $`0.118\pm 0.003(stat)\pm 0.005(syst)\pm 0.003`$
The systematical uncertainties in these results are determined by the systematical uncertainties of the CCFR’97 data and the theoretical errors are fixed from the numbers for $`\mathrm{\Delta }_k`$ (see Table 2), which reflect the factorization/renormalization scale uncertainties of the values of $`\mathrm{\Lambda }_{\overline{MS}}^{(4)}`$. The incorporation into the $`\overline{MS}`$-matching formula of the proposals of Ref. for estimates of the ambiguities due to smooth transition to the world with $`f=5`$ numbers of active flavours was also taken into account. The theoretical uncertainties presented are in agreement with the ones, obtained in Ref., while the NNLO value of $`\alpha _s(M_Z)`$ is in agreement with another NNLO result $`\alpha _s(M_Z)=0.1172\pm 0.0024`$, which was obtained from the analysis of SLAC, BCDMS, E665 and HERA data for $`F_2`$ SF with the help of the Bernstein polynomial technique . It might be of interest to verify the theoretical errors of these two available phenomenological NNLO analysis using different variants of fixing scheme-dependence ambiguities. The first steps towards the analysis of this problem are already made .
Acknowledgements. We are grateful to C.J. Maxwell for careful reading of the manuscript. |
warning/0001/quant-ph0001027.html | ar5iv | text | # New nonlinear coherent states and some of their nonclassical properties
## 1 Introduction
The importance of coherent states of various Lie algebras in different branches of physics(in particular quantum optics) hardly needs to be emphasized. For example the standard coherent state of the harmonic oscillator corresponds to the Heisenberg-Weyl algebra. Similarly coherent states corresponding to Lie algebras like Su(1,1),Su(2) can also be constructed and they have also found numerous applications in quantum optics . In this connection it may be mentioned that coherent states are usually constructed using any of the following three procedures : (1) Displacement operator technique (2) Annihilation operator eigenstates (3) Minimum uncertainty states. However,these three approaches are generally nonequivalent and only in the case of standard harmonic oscillator coherent states obtained using any of the three approaches are equivalent.
On the other hand nonlinear coherent states or the f-coherent states are coherent states corresponding to nonlinear algebras rather than Lie algebras. Nonlinear algebras are distinct from Lie algebras and have recently been used to analyse a number of quantum mechanical systems . However nonlinear coherent states are not merely mathematical objects but they are useful too. Recently it has been shown that nonlinear coherent states are useful in the description of the motion of a trapped ion and various nonclassical properties of such states have also been studied . We note that in refs and nonlinear coherent states have been defined as the right eigenstate of a generalised annihilation operator $`A`$ (which emerges from the Hamiltonian describing the dynamics). This is because in the case of nonlinear algebras the commutator $`[A,A^{}]`$ is not a constant or a linear function of the generators of the algebra but nonlinear in the generators. As a consequence it is difficult to obtain an explicit form of nonlinear coherent state constructed via the displacement operator technique.
In the present paper our aim is to construct nonlinear coherent states using an operator which is similar to the displacement operator. This approach has been used previously to construct nonlinear coherent states in the context of isospectral Hamiltonians and deformed algebras . Subsequently we shall examine various nonclassical properties like quadrature as well as amplitude squared squeezing,sub-Poissonian behaviour etc. of the nonlinear coherent states so obtained. The organisation of the paper is as follows: in section 2 we shall describe the construction of nonlinear coherent states; in section 3 we shall shall study nonclassical properties of the nonlinear coherent state; finally section 4 is devoted to a conclusion.
## 2 Construction of nonlinear coherent states using a displacement type operator
To begin with we note that the generalised annihilation (creation) operator associated with nonlinear coherent states are given by
$$A=af(N),A^{}=f(N)a^{},N=a^{}a$$
(1)
where $`f(x)`$ is a reasonably well behaved real function and $`a^{}(a)`$ is the harmonic oscillator creation(annihilation) operator. It can be easily verified that $`A^{},A`$ and $`N`$ satisfy the following nonlinear algebra:
$$[N,A]=A,[N,A^{}]=A^{},[A,A^{}]=(N+1)f^2(N+1)Nf^2(N)$$
(2)
Clearly the nature of the nonlinear algebra depends on the choice of the nonlinearity function $`f(N)`$. If however $`f(N)=1`$ then the nonlinear algebra in (2) reduces to the Heisenberg algebra.
Nonlinear coherent states $`|\alpha >`$ are then defined as right eigenstates of the generalised annihilation operator $`A`$ :
$$A|\alpha >=\alpha |\alpha >$$
(3)
where $`\alpha `$ is an arbitrary complex number. From (3) we can now obtain an explicit form of the nonlinear coherent state in a number state representation:
$$|\alpha >=C\underset{n=0}{\overset{\mathrm{}}{}}\alpha ^nd_n|n>$$
(4)
where the coefficients $`d_n`$’s are given by
$$d_0=1,d_n=[\sqrt{n}!f(n)!]^1,f(n)!f(1)\mathrm{}f(n)$$
(5)
and the normalisation constant C can be obtained from the condition $`<\alpha |\alpha >=1`$ and is given by
$$C^2=[\underset{n=0}{\overset{\mathrm{}}{}}d_n^2|\alpha |^{2n}]^1$$
(6)
We now turn to the construction of a new type of nonlinear coherent state. From the relation (2) we find that the r.h.s of the commutator $`[A,A^{}]`$ is neither a constant nor linear in the generators but is a nonlinear function of the generator $`N`$. As a result BCH disentangling theorem can not be applied and one can not use the displacement operator $`exp(\alpha A\alpha ^{}A^{})`$ to construct coherent states (see ref for some recent results concerning the applicability of BCH disentangling theorem to nonlinear algebras).
We now proceed to determine an operator $`B^{}`$ which is conjugate of the operator $`A`$. In other words $`A`$ and $`B^{}`$ satisfy the commutation relation
$$[A,B^{}]=1$$
(7)
while their hermitian conjugates $`A^{}`$ and $`B`$ satisfy the dual algebra
$$[B,A^{}]=1$$
(8)
Then from (1),(7) and (8) it follows that
$$B=a\frac{1}{f(N)},B^{}=\frac{1}{f(N)}a^{}$$
(9)
Let us now consider the following operators ($`\beta `$ being an arbitrary complex number):
$$\begin{array}{ccc}D(\beta )\hfill & =& exp(\beta A^{}\beta ^{}B)\hfill \\ D_1(\beta )\hfill & =& exp(\beta B^{}\beta ^{}A)\hfill \end{array}$$
(10)
and note that for any two operators $`X`$ and $`Y`$ satisfying the relation $`[X,Y]=1`$ the BCH theorem yields
$$exp(\beta X\beta ^{}Y)=exp(\frac{\beta \beta ^{}}{2})exp(\beta X)exp(\beta ^{}Y)$$
(11)
We now define nonlinear coherent states corresponding to the algebra (7) as $`D(\alpha )|0>`$ while those corresponding to the dual algebra (8) as $`D_1(\beta )|0>`$. Let us first consider the second case. Applying $`D_1(\beta )`$ on $`|0>`$ we obtain on using (11)
$$|\beta >_1=D_1(\beta )|0>=c_1\underset{n=0}{\overset{\mathrm{}}{}}\frac{\beta ^n}{\sqrt{n}!f(n)!}|n>$$
(12)
where $`c_1`$ is a normalisation constant. Comparing (4) and (12) it is seen that the coherent state $`|\beta >_1`$ is the same as the nonlinear coherent state $`|\alpha >`$. Thus the nonlinear coherent state $`|\alpha >`$ can not only be obtained as an annihilation operator eigenstate but it can also be obtained by the application of a displacement type operator.
Let us now turn to the first possibility. Applying the operator $`D(\beta )`$ on $`|0>`$ we get
$$|\beta >=D(\beta )|0>=c\underset{n=0}{\overset{\mathrm{}}{}}\frac{\beta ^nf(n)!}{\sqrt{n}!}|n>$$
(13)
where as before $`c`$ is a normalisation constant and we have used (11) to obtain (13). The normalisation constant $`c`$ can be determined from the condition $`<\beta |\beta >=1`$ and we get
$$c^2=[\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^n(f(n)!)^2}{n!}]^1$$
(14)
where $`f(0)!0`$. The superposition state obtained in (13) is the new nonlinear coherent state and this is distinct from the nonlinear coherent state defined in (4)(provided of course we use the same nonlinearity function $`f(n)`$ in both the cases). In the next section we shall study various properties of the nonlinear coherent state (13).
## 3 Non classical properties of nonlinear coherent state $`|\beta >`$
In this section we shall examine squeezing and antibunching properties of the new nonlinear coherent state $`|\beta >`$. However before we proceed any further it is necessary to specify the nonlinearity function $`f(n)`$. It is clear that for every choice of $`f(n)`$ we shall have a different nonlinear coherent states. In the present case we choose the following nonlinearity function which is useful in the description of the motion of a trapped ion :
$$f(n)=L_n^1(\eta ^2)[(n+1)L_n^0(\eta ^2)]^1$$
(15)
where $`L_n^m(x)`$ are generalised Laguerre polynomials and $`\eta `$ is known as the Lamb-Dicke parameter. Clearly $`f(n)=1`$ when $`\eta =0`$ and in this case nonlinear coherent states become the standard coherent states. However when $`\eta 0`$ nonlinearity starts developing with the degree of nonlinearity depending on the magnitude of $`\eta `$ .
3.1 Quadrature Squeezing
Here we shall study quadrature squeezing of the new nonlinear coherent state $`|\beta >`$. In order to do so let us consider the following hermitian quadrature operators:
$$X_1=\frac{(a+a^{})}{2},Y_1=\frac{(aa^{})}{2i}$$
(16)
Then $`X_1`$ and $`Y_1`$ satisfy the following uncertainty relation:
$$<\mathrm{\Delta }X_1^2><\mathrm{\Delta }Y_1^2>\frac{1}{16}$$
(17)
where $`<\mathrm{\Delta }X^2>=<X^2><X>^2`$. From (17) it follows that a state is squeezed if any of the following conditions hold:
$$\begin{array}{ccc}<\mathrm{\Delta }X_1^2>\hfill & <& \frac{1}{4}\hfill \\ & & \\ or\hfill & & \\ <\mathrm{\Delta }Y_1^2>\hfill & <& \frac{1}{4}\hfill \end{array}$$
(18)
Now using (13) and (16) the squeezing conditions in (LABEL:ineq) can be reduced to the following forms:
$$\begin{array}{ccc}F_1\hfill & =& <a^^2>+<a^{}a>2<a^{}>^2=\beta ^2I_2+I_32\beta ^2I_1^2<0\hfill \\ & & \\ or\hfill & & \\ G_1\hfill & =& <a^{}a><a^{}>^2=I_3\beta ^2I_2<0\hfill \end{array}$$
(19)
where $`\beta `$ is taken to be real and $`I_i,i=1,2,3`$ are infinite series whose explicit forms are given in the appendix.
We now evaluate the inequalities in (LABEL:ineq1) and the results are presented in fig 1. In fig 1 we have plotted graphs of $`F_1`$ and $`G_1`$ against $`\beta `$ for fixed $`\eta `$. From fig 1 it is seen that while the curve of $`F_1`$ is greater than zero that of $`G_1`$ is less than zero for a wide range of $`\beta `$. Thus one of the inequalities in (LABEL:ineq1) is satisfied. This implies that the nonlinear coherent state exhibits quadrature squeezing. We would like to mention here that we have examined the inequalities in (LABEL:ineq) for a wide range of values of $`\beta `$ and $`\eta `$ and obtained the same qualitative behaviour as in fig 1.
3.2 Amplitude squared squeezing
In order to examine whether or not the nonlinear coherent state exhibits amplitude squared squeezing we introduce the following hermitian operators:
$$X_2=\frac{(a^2+a^^2)}{2},Y_2=\frac{(a^2a^^2)}{2i}$$
(20)
Then $`X_2`$ and $`Y_2`$ obey the uncertainty relation
$$<\mathrm{\Delta }X_2^2><\mathrm{\Delta }Y_2^2>\frac{1}{4}|<[X_2,Y_2]>|^2$$
(21)
From (21) it follows that the nonlinear coherent state will exhibit amplitude squared squeezing if
$$\begin{array}{ccc}<\mathrm{\Delta }X_2^2>\hfill & <& \frac{1}{2}|<[X_2,Y_2]>|\hfill \\ & & \\ or\hfill & & \\ <\mathrm{\Delta }Y_2^2>\hfill & <& \frac{1}{2}|<[X_2,Y_2]>|\hfill \end{array}$$
(22)
Now proceeding as before the conditions (22) for amplitude squared squeezing become
$$\begin{array}{ccc}F_2\hfill & =& <a^^4>+<a^^2a^2><a^^2>=\beta ^4I_4+I_5I_2^2<0\hfill \\ & & \\ or\hfill & & \\ G_2\hfill & =& <a^^4><a^^2a^2>=I_5\beta ^4I_4<0\hfill \end{array}$$
(23)
where $`I_4`$ and $`I_5`$ are infinite series whose exact forms are given in the appendix. To examine the inequalities in (23) we plot $`F_2`$ and $`G_2`$ against $`\beta `$ for $`\eta `$ fixed. From figure 2 we find that $`F_2<0`$ for a certain range of $`\beta `$ and subsequently it becomes positive. On the other hand $`G_2`$ is always positive. This implies that the nonlinear coherent state $`|\beta >`$ exhibits amplitude squared squeezing in the $`X_2`$ component. As in the case of quadrature squeezing we have examined the inequalities in (23) for different values of $`\beta `$ and $`\eta `$ and it has been found that although squeezing can be increased (or decreased) by changing the parameter values the basic qualitative features remain the same as in fig 2.
3.3 Sub-Poissonian Behaviour
To examine sub-Poissonian behaviour we consider the second order correlation function $`g^2(0)`$ defined by
$$g^2(0)=\frac{<a^^2a^2>}{<a^{}a>^2}=\frac{I_4}{I_3^2}$$
(24)
Then the state exhibits super-Poissonian/Poissonian/sub-Poissonian behaviour according to
$$g^2(0)\begin{array}{c}>\\ =\\ <\end{array}1$$
(25)
We now plot $`g^2(0)`$ against $`\beta `$ keeping $`\eta `$ constant. From fig 3 it is seen that $`g^2(0)<1`$ for the range of $`\beta `$ considered. This implies that the nonlinear coherent state $`|\beta >`$ exhibits sub-Poissonian behaviour. It may be noted that $`g^(2)(0)`$ has an increasing trend and so for sufficiently large values of $`\beta `$ it may show super-Poissonian behaviour.
## 4 Conclusion
In this paper we have used a displacement type operator to construct a new class of nonlinear coherent states which are distinct from those which are annihilation operator eigenstates . It has been shown that this nonlinear coherent state exhibits interesting nonclassical properties like squeezing and sub-Poissonian behaviour. We feel it would be interesting to examine other properties e.g.,quantum interference ,phase properties etc of the nonlinear coherent state $`|\beta >`$.
Appendix
In order to examine squeezing and antibunching we need to evaluate several expectation values like $`<a^{}>,<a^^2>,<a^{}a>`$ etc. These expectation values are given in terms of the following series:
$$\beta ^1<a>=\beta ^^1<a^{}>=I_1=c^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^nf(n)!f(n+1)!}{n!}$$
$`(A1)`$
$$\beta ^2<a^2>=\beta ^^2<a^^2>=I_2=c^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^nf(n)!f(n+2)!}{n!}$$
$`(A2)`$
$$<a^{}a>=I_3=c^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^{n+1}[f(n+1)!]^2}{n!}$$
$`(A3)`$
$$<a^^2a^2>=I_4=c^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^n[f(n+2)!]^2}{n!}$$
$`(A4)`$
$$\beta ^4<a^4>=\beta ^^4<a^^4>=I_5=c^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(\beta ^{}\beta )^nf(n)!f(n+4)!}{n!}$$
$`(A5)`$ |
warning/0001/quant-ph0001036.html | ar5iv | text | # Antibunching effect of the radiation field in a microcavity with a mirror undergoing heavily damping oscillation
## Abstract
The interaction between the radiation field in a microcavity with a mirror undergoing damping oscillation is investigated. Under the heavily damping cases, the mirror variables are adiabatically eliminated. The the stationary conditions of the system are discussed. The small fluctuation approximation around steady values is applied to analysis the antibunching effect of the cavity field. The antibunching condition is given under two limit cases.
1. Introduction
There has been interest in the investigation of the non-classical behavior of the light field, for example, the squeezed state of the electromagnetic field . A single mode or a multimode electromagnetic field is squeezed when the fluctuation of its one quadrature component is reduced to below the standard quantum limit. But the fluctuation of its conjugate partner will be enlarged in terms of the uncertainty relation. The squeezed electromagnetic field could be applied to the optical communication and ultrasensitive gravitational wave detection.
Recently, it has been shown that a cavity with free oscillating mirror might be employed as a model for squeezing. It was firstly proposed by Stenholm without taking into account the effect of the fluctuation on the oscillating mirror . Since then, this model is generalized to the coupling of the system to the external world . But up to now, another non-classical behavior of the radiation field in a microcavity with a movable mirror , such as antibunching effect, isn’t still studied.
In this paper, we will discuss the antibunching effect of the radiation field in a microcavity with a mirror undergoing heavily damping oscillation. In section 2, we firstly give a quantizing Hamiltonian including external dissipative effect. The master equation of the reduced density matrix for the system is given. We convert the master equation into a c-number equation (the Fokker-Planck equation) by using the two-mode positive $`P`$ representation. Then the stochastic equation corresponding to the Fokker-Planck equation is obtained. In section 3, we discuss the stability of the system under the good cavity limit and linearize the stochastic equation around stable values. In section 4, antibunching effect is investigated and antibunching condition of the system is given. Finally we give the conclusion of this paper.
2. Hamiltonian and Master equation
We firstly consider the interaction between a single mode cavity field and a movable mirror. Based on reference , we have a effective quantizing Hamiltonian
$$H_0=\mathrm{}\omega _ca^+a+\mathrm{}\omega _mb^+b\mathrm{}ga^+a(b^++b)+i\mathrm{}(E(t)a^+E^{}(t)a)$$
(1)
where $`E(t)`$ is proportional to the amplitude of the external driving field of the cavity mode. $`E^{}(t)`$ is complex conjugate of $`E(t)`$. $`\omega _c`$ is the cavity frequency. $`a`$($`a^+`$) are annihilate (create) operators of the cavity field. $`\omega _m`$ is a frequency of the mirror. $`b`$($`b^+`$) are annihilate (create) operators of the mirror. $`g=\frac{\omega _c}{L}(\frac{\mathrm{}}{2m\omega _m})`$. $`m`$ is a mass of the mirror and $`L`$ is the equilibrium cavity length.
If the cavity is bad, and the mirror is damped by the circumstance when it is moving, then the Hamiltonian (1) need correct to add the dassipative effect of the circumstance. So we have
$$H=H_0+\mathrm{}a^+\mathrm{\Gamma }_1+\mathrm{}a\mathrm{\Gamma }_1^++\mathrm{}b^+\mathrm{\Gamma }_2+\mathrm{}b\mathrm{\Gamma }_2^+$$
(2)
where $`\mathrm{\Gamma }_1(\mathrm{\Gamma }_1^+)`$ is the reservoir operators of the cavity. We also model the damping effect of the mirror as the result of the interaction between the mirror and the many harmonic oscillator. They satisfy the Markovian correlation function (where for simplicity, we only consider the zero temperature case):
$`<\mathrm{\Gamma }_i(t)\mathrm{\Gamma }_i^+(t^{})>`$ $`=`$ $`\gamma _i\delta (tt^{})`$ (3)
$`<\mathrm{\Gamma }_i^+(t)\mathrm{\Gamma }_i(t^{})>`$ $`=`$ $`0`$ (4)
with $`i=1`$, $`2`$. $`\gamma _1`$ is the decay rate of the cavity field. $`\gamma _2`$ is the damping coefficient of the motion mirror. The master equation of the reduced density matrix for the system is
$`{\displaystyle \frac{\rho }{t}}`$ $`=`$ $`{\displaystyle \frac{1}{i\mathrm{}}}[H_0,\rho ]+\gamma _1(2a\rho a^+\rho a^+aa^+a\rho )`$ (5)
$`+`$ $`\gamma _2(2b\rho b^+\rho b^+bb^+b\rho )`$ (6)
We could convert the operator eq.(4) into a c-number equation (the Fokker-Planck Equation) by using the two-mode positive $`P`$ representation . This representation ensures that the $`P`$ function exists as a well-behaved distribution which is singular when the usual Glauber-Sudarshan $`P`$ representation was applied. That is:
$`{\displaystyle \frac{P}{t}}`$ $`=`$ $`[{\displaystyle \frac{}{\alpha _1}}(i\omega _c\alpha _1ig\alpha _1\alpha _2^+ig\alpha _1\alpha _2+\gamma _1\alpha _1E(t))`$ (7)
$`+`$ $`{\displaystyle \frac{}{\alpha _1^+}}(i\omega _c\alpha _1^++ig\alpha _1^+\alpha _2^++ig\alpha _1^+\alpha _2+\gamma _1\alpha _1^+E^{}(t))`$ (8)
$`+`$ $`{\displaystyle \frac{}{\alpha _2}}(i\omega _m\alpha _2ig\alpha _1^+\alpha _1+\gamma _2\alpha _2)`$ (9)
$`+`$ $`{\displaystyle \frac{}{\alpha _2^+}}(i\omega _m\alpha _2^++ig\alpha _1^+\alpha _1+\gamma _2\alpha _2^+)`$ (10)
$`+`$ $`{\displaystyle \frac{}{\alpha _1}}{\displaystyle \frac{}{\alpha _2}}ig\alpha _1{\displaystyle \frac{}{\alpha _1^+}}{\displaystyle \frac{}{\alpha _2^+}}ig\alpha _1^+]P`$ (11)
where $`\alpha _1`$ and $`\alpha _1^+`$, and $`\alpha _2`$ and $`\alpha _2^+`$ are no longer complex conjugate of each other, instead they are independent complex variables. In terms of the Ito ruler, the Fokker-Planck eq.(5) is equivalent to the following set of the stochastic equations.
$`{\displaystyle \frac{\alpha _1}{t}}`$ $`=`$ $`i\omega _c\alpha _1+ig\alpha _1\alpha _2^++ig\alpha _1\alpha _2\gamma _1\alpha _1+E(t)+\mathrm{\Gamma }_{\alpha _1}`$ (13)
$`{\displaystyle \frac{\alpha _1^+}{t}}`$ $`=`$ $`i\omega _c\alpha _1^+ig\alpha _1^+\alpha _2^+ig\alpha _1^+\alpha _2\gamma _1\alpha _1^++E^{}(t)+\mathrm{\Gamma }_{\alpha _1^+}`$ (14)
$`{\displaystyle \frac{\alpha _2}{t}}`$ $`=`$ $`i\omega _m\alpha _2+ig\alpha _1^+\alpha _1\gamma _2\alpha _2+\mathrm{\Gamma }_{\alpha _2}`$ (15)
$`{\displaystyle \frac{\alpha _2^+}{t}}`$ $`=`$ $`i\omega _m\alpha _2^+ig\alpha _1^+\alpha _1\gamma _2\alpha _2^++\mathrm{\Gamma }_{\alpha _2^+}`$ (16)
where $`\mathrm{\Gamma }_{\alpha _i}`$ and $`\mathrm{\Gamma }_{\alpha _i^+}`$ are Gaussian random variables with zero mean. Their correlation functions satisfy:
$`<\mathrm{\Gamma }_{\alpha _1}(t)\mathrm{\Gamma }_{\alpha _2}(t^{})>`$ $`=`$ $`<\mathrm{\Gamma }_{\alpha _2}(t)\mathrm{\Gamma }_{\alpha _1}(t^{})>=ig\alpha _1\delta (tt^{})`$ (18)
$`<\mathrm{\Gamma }_{\alpha _1^+}(t)\mathrm{\Gamma }_{\alpha _2^+}(t^{})>`$ $`=`$ $`<\mathrm{\Gamma }_{\alpha _2^+}(t)\mathrm{\Gamma }_{\alpha _1^+}(t^{})>=ig\alpha _1^+\delta (tt^{})`$ (19)
$`<\mathrm{\Gamma }_{\alpha _l}(t)\mathrm{\Gamma }_{\alpha _m^+}(t^{})>`$ $`=`$ $`<\mathrm{\Gamma }_{\alpha _m^+}(t)\mathrm{\Gamma }_{\alpha _l}(t^{})>=0`$ (20)
with $`l=1`$ or $`2`$ and $`m=1`$ or $`2`$. The above correlation functions could be obtained from the matrix elements of the diffusion matrix of eq.(5) by using Ito ruler.
3. Adiabatic elimination of the mirror variable and stability analysis
In the good-cavity limit, that is, the damping coefficient of the mirror is much larger than the decay rate of the cavity field ($`\gamma _2\gamma _1`$). ˚This means that the decay time of the cavity field is much longer than the decay time of the mirror. So the mirror variables could be adiabatically eliminated from eqs.(6a-6b) . Firstly we find a steady-state of the mirror variables . In the case of the steady state, $`\dot{\alpha }_2=0`$ and $`\dot{\alpha }_2^+=0`$. So we have from eqs.(6c-6d)
$`i\omega _m\alpha _2+ig\alpha _1^+\alpha _1\gamma _2\alpha _2+\mathrm{\Gamma }_{\alpha _2}`$ $`=`$ $`0`$ (22)
$`i\omega _m\alpha _2^+ig\alpha _1^+\alpha _1\gamma _2\alpha _2^++\mathrm{\Gamma }_{\alpha _2^+}`$ $`=`$ $`0`$ (23)
From eqs.(8a-8b), we obtain:
$`\alpha _2`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{\alpha _2}+ig\alpha _1\alpha _1^+}{\gamma _2+i\omega _m}}`$ (25)
$`\alpha _2^+`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{\alpha _2^+}ig\alpha _1\alpha _1^+}{\gamma _2i\omega _m}}`$ (26)
We substitute eqs.(9a-9b) into eqs.(6a-6b) and eliminate the mirror variables. Then we have:
$`{\displaystyle \frac{\alpha _1}{t}}`$ $`=`$ $`i\omega _c\alpha _1+iG(\omega _m)\alpha _1^+\alpha _1^2\gamma _1\alpha _1+\mathrm{\Gamma }`$ (28)
$`{\displaystyle \frac{\alpha _1^+}{t}}`$ $`=`$ $`i\omega _c\alpha _1^+iG(\omega _m)\alpha _1^{+2}\alpha _1\gamma _1\alpha _1^++\mathrm{\Gamma }^+`$ (29)
with $`G(\omega _m)=\frac{2g^2\omega _m}{\gamma _2+\omega _m^2}`$ and
$`\mathrm{\Gamma }`$ $`=`$ $`{\displaystyle \frac{ig\alpha _1}{\gamma _2+i\omega _m}}\mathrm{\Gamma }_{\alpha _2}+{\displaystyle \frac{ig\alpha _1}{\gamma _2i\omega _m}}\mathrm{\Gamma }_{\alpha _2^+}+\mathrm{\Gamma }_{\alpha _1}`$ (31)
$`\mathrm{\Gamma }^+`$ $`=`$ $`{\displaystyle \frac{ig\alpha _1^+}{\gamma _2+i\omega _m}}\mathrm{\Gamma }_{\alpha _2}{\displaystyle \frac{ig\alpha _1^+}{\gamma _2i\omega _m}}\mathrm{\Gamma }_{\alpha _2^+}+\mathrm{\Gamma }_{\alpha _1^+}`$ (32)
Using eqs.(7a-7c) and eqs.(11a-11b), we calculate the correlation function
$`<\mathrm{\Gamma }(t)\mathrm{\Gamma }^+(t^{})>`$ $`=`$ $`<\mathrm{\Gamma }^+(t)\mathrm{\Gamma }^+(t)>={\displaystyle \frac{2\gamma _2g^2n}{\gamma _2^2+\omega _m^2}}\delta (tt^{})`$ (34)
$`<\mathrm{\Gamma }^+(t)\mathrm{\Gamma }^+(t^{})>`$ $`=`$ $`{\displaystyle \frac{2g^2\alpha _1^{+2}}{\gamma _2i\omega _m}}\delta (tt^{})`$ (35)
$`<\mathrm{\Gamma }(t)\mathrm{\Gamma }(t^{})>`$ $`=`$ $`{\displaystyle \frac{2g^2\alpha _1^2}{\gamma _2+i\omega _m}}\delta (tt^{})`$ (36)
The eqs.(10a-10b) are difficult to be solved. In general, we are interested in the properties of the steady state. So we denote the steady values of $`\alpha _1`$ and $`\alpha _1^+`$ by $`\alpha _0`$ and $`\alpha _0^+`$ respectively. We assume that the system has a small approximation around the steady values, namely
$$\{\begin{array}{c}\alpha _1(t)=\alpha _0+\delta \alpha _1(t)\hfill \\ \alpha _1^+(t)=\alpha _0^++\delta \alpha _1^+(t)\hfill \end{array}$$
(37)
So non-linear eqs.(10a-10b) are simplified into the following linear equations around the steady values.
$`{\displaystyle \frac{}{t}}\left(\begin{array}{c}\delta \alpha _1\\ \delta \alpha _1^+\end{array}\right)`$ $`=`$ $`\left(\begin{array}{cc}i\omega _c\gamma _1+i2G(\omega _m)\alpha _0^+\alpha _0& iG(\omega _m)\alpha _0^2\\ iG(\omega _m)\alpha _0^{+2}& i\omega _c\gamma _1+i2G(\omega _m)\alpha _0^+\alpha _0\end{array}\right)\left(\begin{array}{c}\delta \alpha _1\\ \delta \alpha _1^+\end{array}\right)`$ (44)
$`+`$ $`\left(\begin{array}{cc}\frac{2g^2\alpha _0^2}{\gamma _2+i\omega _m}& \frac{2\gamma _2g^2n}{\gamma _2^2+\omega _m^2}\\ \frac{2\gamma _2g^2n}{\gamma _2^2+\omega _m^2}& \frac{2g^2\alpha _0^{+2}}{\gamma _2i\omega _m}\end{array}\right)^{\frac{1}{2}}\left(\begin{array}{c}\eta _1(t)\\ \eta _2(t)\end{array}\right)`$ (49)
where $`n=\alpha _0^+\alpha _0`$ and $`\eta _i(t)`$ statisfy delta correlation function:
$`<\eta _i(t)>`$ $`=`$ $`0`$ (51)
$`<\eta _i(t)\eta _j(t^{})>`$ $`=`$ $`\delta _{ij}\delta (tt^{})`$ (52)
We abbreviate eq.(8) as following:
$$\frac{}{t}\delta \stackrel{}{\alpha }(t)=A\stackrel{}{\alpha }(t)+D^{\frac{1}{2}}\stackrel{}{\eta }(t)$$
(53)
But one of the important feature of eq.(14) is whether the steady solutions are stable, that is, when $`\alpha _1`$ and $`\alpha _1^+`$ somewhat deviate from $`\alpha _0`$ and $`\alpha _0^+`$, whether they will still return to steady values. So we neglect the fluctuation forces of eq.(16) and have:
$$\frac{}{t}\delta \stackrel{}{\alpha }(t)=A\stackrel{}{\alpha }(t)$$
(54)
In order to investigate the stationary of the system, we seek the solutions of the form $`e^{\lambda t}`$ of eq.(17). The eigenvalues $`\lambda `$ are determined by the equation
$$|A\lambda I|=0$$
(55)
where $`I`$ is a identity matrix. The solutions of the eq.(18) are
$$\lambda =\gamma _1\pm \sqrt{(\omega _c3G(\omega _m)n)(\omega _cG(\omega _m)n)}$$
(56)
This equation indicates if $`G(\omega _m)n\omega _c3G(\omega _m)n`$, the real part of the eigen-solution of the eq.(18) is positive. The system is stable. When $`\omega _cnG(\omega _m)`$ or $`\omega _c3nG(\omega _m)`$, only if $`\gamma _1\sqrt{(\omega _c3G(\omega _m)n)(\omega _cG(\omega _m)n)}`$ then the system is also stable. So only we chose the proper parameter, the system always may reach to stability. The small fluctuation approximation is appreciate.
4. Second-order correlation function
The intensity correlation is another quantity of the experimental interest besides the first order optical coherence. It’s truly photon correlation measurement. Theoretically, people have defined a second-order correlation function to investigate the joint photocount probability of detecting the arrival of a photon at one time and other ptoton at another time. For zero time delay, The second order correlation function is
$$g^{(2)}(0)=\frac{<a^+a^+aa>}{<a^+a>^2}$$
(57)
Under the positive $`P`$ representation we keep the second order terms of $`\frac{\delta \alpha }{\alpha }`$, then eq(20) becomes into:
$$g^{(2)}(0)==\frac{<\alpha ^{+2}\alpha ^2>}{<\alpha ^+\alpha >^2}=1+\frac{<\delta \alpha ^2>}{\alpha ^2}+\frac{<\delta \alpha ^{+2}>}{\alpha ^{+2}}+4\frac{<\delta \alpha ^+\delta \alpha >}{\alpha ^+\alpha }$$
(58)
In order to calculate the correlation functions $`<(\delta \alpha )^2>`$ et.al, we need calculate matrix:
$$C=\left(\begin{array}{cc}<\delta \alpha \delta \alpha >& <\delta \alpha \delta \alpha ^+>\\ <\delta \alpha ^+\delta \alpha >& <\delta \alpha ^+\delta \alpha ^+>\end{array}\right)$$
(59)
The matrix $`C`$ could be obtained by calculating of $`\delta \alpha _1`$ and $`\delta \alpha _1^+`$ from the linearized eq.(14). It is also could be given by :
$$C=\frac{D.Det(A)+(AITr(A))D(AITr(A))^T}{2Tr(A)Det(A)}.$$
(60)
after tedious calculation, we have:
$`C_{11}`$ $`=`$ $`C_{22}^{}={\displaystyle \frac{\alpha _1^2[2M\gamma _1^2+2NG(B+i\gamma _1)i2\gamma _1BMG^2n^2(M+M^{})]}{4\gamma _1(\gamma _1^2+B^2G^2n^2)}}`$ (62)
$`C_{12}`$ $`=`$ $`C_{21}={\displaystyle \frac{2N(\gamma _1^2+B^2)+i\gamma _1Gn^2(M^{}M)GBn^2(M+M^{})}{4\gamma _1(\gamma _1^2+B^2G^2n^2)}}`$ (63)
with
$$\begin{array}{ccc}M=\frac{2g^2}{\gamma _2+i\omega _m}& N=\frac{2\gamma _2g^2n}{\gamma _2^2+\omega _m^2}& B=\omega _c2Gn\end{array}$$
(64)
So the second order correlation function is:
$$g(0)=1+\frac{G[(\omega _c2nG)(2n\gamma _2\omega _cn^2\gamma _2G\gamma _1\omega _m)+n(\gamma _1^2\gamma _2+n^2G^2\gamma _22\gamma _1\omega _mGn)]}{n\gamma _1\omega _m[\gamma _1^2(\omega _c+5nG)(\omega _cnG)]}$$
(65)
From this equation, we see that the antibunching of the cavity field appear when the second term of the eq.(26) is a negative number. This condition could be satisfied after we chose some proper parameters. Now we consider two limit cases. In the case of the strong cavity field, we only keep the terms including $`n^3`$. The second term of eq.(26) is positive, no antibunching appears. But under case of the weak cavity field, namely when $`n0`$, the cavity field presents antibunching behavior.
5. Conclusion
The interaction between the radiation field in a microcavity with a movable mirror undergoing heavily damping oscillation is investigated. Under the heavily damping cases, the mirror variables are adiabatically eliminated. The stationary condition of the system is given. The small fluctuation approximation around steady values is applied to analysis the antibunching behavior of the cavity field. In the case of the strong cavity field, no antibunching appears. But under the case of the weak cavity field, cavity field presents the antibunching effect. |
warning/0001/cond-mat0001334.html | ar5iv | text | # Uptake of gases in bundles of carbon nanotubes
## I Introduction
The absorption of gases in nanopores is a subject of growing experimental and theoretical interest, stimulated by both fundamental scientific questions and the potential for many technologies . One of the most important questions to be addressed is whether or not a specific gas is significantly absorbed within carbon nanotubes; we will define the word “significant” in Eq. 3 below. While the answer depends in detail on the specific thermodynamic conditions of the coexisting vapor (pressure $`P`$ and temperature $`T`$), one expects that intuitive considerations based on size and energy scales ought to provide useful qualitative insights. For example, it has been demonstrated that gases whose condensed phases possess low surface tensions are strongly imbibed in these tubes . This important result can be understood from either the Kelvin equation or a comparison of competing interaction (adhesive vs. cohesive) energies. These considerations arise in the analogous problem of wetting transitions .
This paper addresses this basic question by employing a simple, but plausible, model of the interaction potential, from which we compute the adsorption as a function of $`P`$ and $`T`$. We assume that the adsorption potential can be derived from a sum of Lennard-Jones (LJ) two-body interactions between the host C atoms and the adsorbate. This pair potential has distance and energy parameters obtained with semiempirical combining rules from the LJ $`ϵ`$ and $`\sigma `$ parameters of the C atoms and the adsorbate :
$`\sigma _{g\mathrm{SS}C}`$ $`=`$ $`{\displaystyle \frac{\sigma _{gg}+\sigma _{\mathrm{SS}CC}}{2}}`$ (1)
$`ϵ_{g\mathrm{SS}C}`$ $`=`$ $`\sqrt{ϵ_{gg}ϵ_{\mathrm{SS}CC}}`$ (2)
where “g” and “C” refer to the gas and C atoms, respectively. Estimates of the gas parameters are given for some relevant systems in Table I , while for C atoms we use $`\sigma _{\mathrm{SS}CC}=3.4`$ Å and $`ϵ_{\mathrm{SS}CC}=28`$ K . These values are typical, but uncertain within a 15% range .
This paper’s approach is the following. We first choose a particular (somewhat arbitrary) criterion for calling the uptake “significant”. For example, Fig. 1(a) and most of our work employ the criterion
$$\rho ^{}=\rho \sigma _{g\mathrm{SS}C}=0.1$$
(3)
where $`\rho =N/L`$ is the one-dimensional (1d) density, with $`N`$ the number of adsorbed atoms and $`L`$ the length of the tube, and $`\rho ^{}`$ is the corresponding dimensionless density. For gases of interest here, this criterion corresponds to a mean 1d spacing of order 30 Å. This is a very low density. Although we do consider more stringent criteria elsewhere in this paper, the results do not differ qualitatively. The reason for the lack of sensitivity to the threshold is that once adsorption commences, it rises rapidly as a function of $`P`$ (until the crowding effect of repulsive forces slows the variation of coverage with $`P`$).
In the nanotube bundle geometry, adsorption can take place inside the tubes, in the interstitial channels, and on the outer surface of the nanotube bundle (Fig. 2). Typical length scales for the triangular lattice of nanotubes in the bundle are: lattice constant 17 Å, nanotube radius 6.9 Å, bundle diameter between 50 and 100 Å, and bundle length $`10100\mu m`$ . We will see that size is a critical variable determining uptake. Some key findings of this paper appear in Fig. 1(a), which shows the uptake at a very small ratio of $`P`$ to saturated vapor pressure ($`P_0`$). Small atoms and molecules (which typically have small values of $`ϵ_{gg}`$) are strongly adsorbed within both the nanotubes and the narrow interstitial channels (IC’s) between nanotubes. Larger particles, in contrast, do not “fit” within the IC’s but do imbibe within the tubes. Perhaps a surprising feature of Fig. 1(a) is that a hypothetical gas with a very large value of $`ϵ_{gg}`$ adsorbs in neither place. This occurs because the relative tendency (compared with bulk condensation) of a gas to be absorbed within the tube at a given undersaturation depends on the ratio of adhesive to cohesive energies. The geometric mean combining rule for $`ϵ_{g\mathrm{SS}C}`$ implies that this ratio varies as the inverse square root of $`ϵ_{gg}`$, so a large $`ϵ_{gg}`$ implies small uptake. This finding is qualitatively consistent with the empirical correlation between uptake and surface tension mentioned above. It also correlates with the physics determining wetting behavior of liquids for which the analogous comparison involves the same kind of interaction ratio .
This paper makes a number of simplifications in order to draw such general conclusions. Arguably the most drastic assumptions are that the nanotubes are infinite and perfect and that the nanotube bundles involve a unique species of tubes in a regular array (geometry unaffected by the adsorption).
The outline of this paper is the following. Section II describes our model of the interactions. Section III presents the statistical mechanical model used in the calculations. Section IV reports our results. Section V summarizes these and discusses open questions.
## II Adsorption potential
A basic assumption in our model is that the potential energy experienced by a molecule at position $`𝐫`$ can be evaluated by a summation of two-body interactions $`U(𝐱)`$ between the molecule and the carbon atoms comprising the tube:
$$V(𝐫)=\underset{i}{}U(𝐫𝐑_𝐢)$$
(4)
This assumption is made in the overwhelming majority of calculations of gas interactions with either graphite and carbon nanotubes. In the graphite case, many-body effects have been found to be $``$ 15% corrections to ab initio pair potential sums . Hence, the empirical pair potential should be regarded as an effective pair potential. One might expect somewhat smaller many body contributions in the nanotube case because the molecule is somewhat farther from the nearest carbon atom and because the effective coordination number is larger in the nanotube case than on graphite. In contrast, the argument in the IC case leads to the prediction of a larger many body effect than on graphite. These expectations, however, might not be correct because the many-body expansion involves geometry-dependent competing terms of opposite signs and because the two body energy for the IC is typically of much larger magnitude than on a flat surface.
Another key assumption made here is that the pair potential is isotropic and of LJ form: $`U(x)=4ϵ[(\sigma /x)^{12}(\sigma /x)^6]`$. There is ab initio and empirical evidence to the effect that anisotropy of the pair potential plays a role in adsorption potentials on graphite . Nevertheless, most studies of adsorption on that surface neglect such an effect and use a LJ pair potential similar to what we use here. The final assumption is the use of an azimuthally and longitudinally averaged potential. The potential at distance $`r`$ from the axis of the cylinder is then :
$$V(r;R)=3\pi \theta ϵ\sigma ^2\left[\frac{21}{32}\left(\frac{\sigma }{R}\right)^{10}f_{11}(x)M_{11}(x)\left(\frac{\sigma }{R}\right)^4f_5(x)M_5(x)\right]$$
(5)
where $`\theta `$ = 0.38 Å<sup>-2</sup> is the surface density of C atoms and $`R`$ is the radius of the cylinder. Here, $`x=r_</r_>`$, and $`r_{<(>)}`$ are the smaller (greater) of $`r`$ and $`R`$. The function $`f_n(x)`$ is defined as 1 for $`r<R`$ and $`(R/r)^n`$ for $`r>R`$, with $`n`$ a positive integer. Here we use the integrals
$$M_n(x)=_0^\pi 𝑑\phi \frac{1}{(1+x^22xcos\phi )^{n/2}}$$
(6)
We emphasize that each approximation introduces an error, but the qualitative trends ought to be reliable. At this time, the lack of high quality ab initio calculations would seem to warrant this kind of approach.
The IC potential is obtained by summing up the contribution from three nanotubes and azimuthally averaging the result. Figures 3-5 show contour plots in the $`\sigma _{gg}ϵ_{gg}`$ plane of the reduced minimum of the adsorption potential ($`V_{min}^{}V_{min}/ϵ_{gg}`$) for all of these sites. Inside both the tubes and in IC’s there is a threshold $`\sigma _{gg}`$ value above which the potential becomes repulsive, corresponding to gases which are too big to fit in these restricted geometries; these thresholds are $`\sigma _{gg}`$ 11.4 Å (tubes) and 3.4 Å (IC’s) for nanotubes of radius 6.9 Å studied here. Outside of the bundle, there are no such size constraints for the adsorbed atoms/molecules, as the adsorbate can always find a region in which the potential is attractive; at a fixed value of $`ϵ_{gg}`$ large systems yield larger $`|V_{min}^{}|`$ due to their larger coordination number of C atoms. In all three cases the most negative values of $`V_{min}^{}`$ occur for small values of $`ϵ_{gg}`$. In the tube and external surface cases, but not the IC case, the most negative values of $`V_{min}^{}`$ occur for large $`\sigma _{gg}`$ ($`9`$ Å).
## III Statistical Mechanics
Our interest is whether atoms are likely to go inside the tubes, in the interstitial channels, and on the outer surface of the nanotube bundle. This behavior is determined by the thermodynamic conditions ($`P`$, $`T`$) and microscopic parameters (especially $`\sigma _{gg}`$ relative to $`R`$). A key factor implicit here is the cohesive energy of the bulk phase of the adsorbate which determines a relevant pressure, i. e. saturated vapor pressure $`P_0`$. We construct a simple model for the low coverage regime of atoms inside nanotubes, neglecting the interactions between adsorbate atoms, while for atoms moving in the very confining IC’s any density can be considered because of the mathematical simplicity resulting from the 1d character of the system. We have discussed elsewhere the extreme quantum behavior of He at low T . In the present case we assume that classical statistical mechanics applies .
We now compute the chemical potential $`\mu `$ of the adsorbate. All of our calculations take the coexisting three-dimensional vapor to be an ideal gas, so that the chemical potential can be expressed in terms of pressure as $`\mu =\beta ^1ln(\beta P\lambda ^3)`$. Here $`\beta ^1=k_BT`$, and $`\lambda =(2\pi \mathrm{}^2\beta /m)^{1/2}`$ is the de Broglie thermal wavelength for particles of mass $`m`$. It is convenient to measure the chemical potential with respect to its value at saturation, $`\mu _0`$,
$$\mathrm{\Delta }\mu =\mu \mu _0=\beta ^1ln(P/P_0)$$
(7)
An analytical expression for $`P_0`$ is available from computer simulation data of the Lennard-Jones system’s liquid-vapor coexistence , $`lnP_0^{}=1.2629T^{}4.9095/T^{}0.15115/T^4`$, where $`P_0^{}=P_0\sigma _{gg}^3/ϵ_{gg}`$ and $`T^{}=k_BT/ϵ_{gg}`$ are reduced quantities.
Consider first the adsorption inside a single nanotube. The chemical potential of the ideal gas in an external potential can be expressed as a function of the number of adsorbed atoms, $`N`$, and temperature:
$$e^{\beta \mu }=\frac{N\lambda ^3}{_{\mathrm{SS}NT}𝑑𝐫exp(\beta V(r))}$$
(8)
where the integral is performed over the volume of the nanotube. This is an application of Henry’s law. Then, the chemical potential relative to its value at saturation assumes a simple form due to the cylindrical symmetry of the adsorption potential:
$$\mathrm{\Delta }\mu =\beta ^1ln\left(\frac{\rho }{2\pi \beta P_0_{\mathrm{SS}NT}𝑑rrexp(\beta V(r))}\right)$$
(9)
Atoms in the narrow IC’s are strongly confined to the vicinity of the axis so that a 1d model is applicable and solvable for all densities. As previously discussed in the case of very small nanotubes , the transverse motion may be treated independently of the longitudinal motion and the chemical potential in this case has the form:
$$\mu =\mu _{\mathrm{SS}}+\mu _{\mathrm{SS1}d}$$
(10)
where $`\mu _{\mathrm{SS}}`$ is the transverse contribution and $`\mu _{\mathrm{SS1}d}`$ is the chemical potential of a 1d gas. In general, $`\beta \mu _{\mathrm{SS}}=ln(_iexp(\beta ϵ_i))`$, where $`\{ϵ_i\}_{i=0,1,\mathrm{}}`$ is the transverse spectrum of individual atoms/molecules. At low T ($`\beta (ϵ_1ϵ_0)<<1`$), the ground state dominates the sum and $`\mu _{\mathrm{SS}}ϵ_0`$. The ground state energy can be determined very accurately using the WKB method , since the adsorption potential is well-approximated by a parabola in the vicinity of the IC axis . Results for the potential well depths of various gases are shown in Table I.
The 1d chemical potential is obtained by integrating the 1d Gibbs-Duhem relation
$$\frac{\mu _{\mathrm{SS1}d}}{P_{\mathrm{SS1}d}}=\frac{1}{\rho }$$
(11)
where $`P_{\mathrm{SS1}d}`$ is the 1d pressure. The particle density in the case of only nearest-neighbor interactions is given by the equation of state
$$\rho =\frac{_0^{\mathrm{}}𝑑zexp(\beta [u(z)+zP_{\mathrm{SS1}d}])}{_0^{\mathrm{}}𝑑zzexp(\beta [u(z)+zP_{\mathrm{SS1}d}])}$$
(12)
Here u(z) is the LJ potential describing the interactions between adsorbed atoms. The integration of eq. 11 leads to
$$\beta \mu _{\mathrm{SS1}d}=ln\left(\frac{\beta \lambda P_{1d,0}_0^{\mathrm{}}𝑑zexp(\beta [u(z)+zP_{1d,0}])}{_0^{\mathrm{}}𝑑zexp(\beta [u(z)+zP_{\mathrm{SS1}d}])}\right)$$
(13)
$`P_{1d,0}`$ is an initial low pressure chosen such that the ideal gas limit is reproduced. The density dependence of the 1d chemical potential is finally obtained by eliminating the 1d pressure between eqs. 11, 12 and 13.
As shown in Fig. 5, the external surface of the nanotube bundle also provides an attractive domain of adsorption. We have studied adsorption in the very attractive groove-like channel which runs parallel to the nanotube axes, as shown in Figure 2 . Then, a procedure similar to that employed in the case of IC’s is applicable to computing the coverage. The contribution of the longitudinal motion to the chemical potential is determined in the same fashion as in the case of IC’s. The ground state energy of the transverse motion can be estimated through a parabolic approximation for the adsorption potential at this site. Values of the ground state energy ($`E_0^{ext}`$) obtained in this fashion for the systems studied, as well as the well-depth of the adsorption potential ($`V_{min}^{ext}`$), are listed in Table I.
## IV Results
The lines obtained by setting the coverage equal to the threshold criterion can be seen in Figures 1(a)-(d). Figure 1(a) shows this behavior in the case of $`\rho ^{}=0.1`$, for $`\mathrm{\Delta }\mu ^{}=\mathrm{\Delta }\mu /ϵ_{gg}=10`$ and $`T^{}=1`$. As expected, small atoms or molecules (He, Ne, H<sub>2</sub>) fit easily inside both the tubes and IC’s, while large molecules do not fit in the narrow IC’s. Hypothetical (but nonexistent) atoms with $`\sigma _{gg}<2.5`$ Å are adsorbed in the IC’s only if their self-interaction energy ($`ϵ_{gg}`$) does not exceed a threshold value. The upper limit to the molecular size for adsorption inside the tubes can be seen in Fig. 6. Indeed, the experimental observation of C<sub>60</sub> molecules encapsulated in nanotubes is consistent with this expectation (as the point near $`\sigma _{gg}9`$ Å indicates).
Including the effect of interactions does not affect our results significantly. In the framework of the gas-surface virial expansion ,
$$Ne^{\beta \mu }Q_1[1+\rho \eta (T)]$$
(14)
where $`Q_1`$ is the single particle canonical partition function, and
$$\eta (T)=L\frac{_{NT}d𝐫_\mathrm{𝟏}d𝐫_\mathrm{𝟐}exp(\beta (V(r_1)+V(r_2))[e^{\beta u(|𝐫_\mathrm{𝟏}𝐫_\mathrm{𝟐}|)}1]}{[_{NT}𝑑𝐫exp(\beta V(r))]^2}$$
(15)
The net effect of the virial correction is at most a 0.1% change of $`\mathrm{\Delta }\mu `$; such a small magnitude is consistent with the expected behavior in the low pressure regime of interest here.
The evolution of the diagram as a function of the adsorption criterion can be seen in Fig. 1(b). As the threshold density decreases ($`\rho ^{}=0.05`$ here) more systems satisfy the uptake criterion. Fig. 1(c) shows a similar effect on the diagram of an increase in chemical potential, to $`\mathrm{\Delta }\mu ^{}=8`$. In both geometries, the altered criterion corresponds to more systems being allowed in the respective cavities. A different effect on the diagram occurs if the size of the nanotubes is changed, as shown in Fig. 1(d), under the thermodynamic conditions of Fig. 1(a). In the case of nanotubes with radius 8 Å more atoms enter interstitial channels because of the larger channel space, while fewer atoms go inside the tubes because the adhesive energy decreases. The trends seen in Fig. 1(a)-(d) are qualitatively consistent with the behavior of $`V_{min}^{}`$ presented in Figures 3 and 4.
In the Henry’s law regime of low coverage there is a convenient way to characterize the variation of uptake with geometry. We compute the ratio of particle occupations in the nanotubes and IC’s at the same $`P`$ and $`T`$:
$$\mathrm{\Gamma }(ϵ_{gg},\sigma _{gg})=\frac{\nu _{\mathrm{SS}NT}}{\nu _{\mathrm{SS}IC}}\frac{_{\mathrm{SS}NT}𝑑𝐫exp(\beta V)}{_{\mathrm{SS}IC}𝑑𝐫exp(\beta V)}$$
(16)
where $`\nu _{\mathrm{SS}NT(IC)}`$ are the number of nanotubes (IC’s) in the bundle and the integrations are over one region (assumed infinitely long). For an infinite array of nanotubes, $`\nu _{\mathrm{SS}NT}/\nu _{\mathrm{SS}IC}=1/2`$. The finiteness of the bundle changes the ratio; however, there is no qualitative effect on our conclusions unless the bundle is very small. This ratio depends on the two gas parameters, $`ϵ_{gg}`$ and $`\sigma _{gg}`$. In order to simplify the presentation, we fit the general trend of systems in Table I to an empirical equation:
$$ϵ_{gg}^{fit}a\sigma _{gg}+b$$
(17)
with values $`a=147`$ K/Å and $`b=376`$ K. We then consider a function of one variable
$$\mathrm{\Gamma }(\sigma _{gg})\mathrm{\Gamma }(ϵ_{gg}^{fit},\sigma _{gg})$$
(18)
This ratio function is presented in Figures 7 and 8. In Fig. 7 we consider a common value of $`T^{}=1`$, while in Fig.8, we consider a fixed $`T=77`$ K. The data in Fig. 7 shows, as expected, that large (small) molecules adsorb preferentially in the nanotubes (IC’s). Fig. 8 differs for small $`\sigma _{gg}`$ because at 77 K the entropic advantage of the tubes is manifested as a larger uptake there than is seen in Fig. 7 at the much lower T given by Eq. (17).
## V Conclusions
Model calculations were used to investigate adsorption in nanotube bundles. Simplifying assumptions were made, such as the pairwise summation of gas-surface interactions, the use of combining rules to determine energy and size parameters, and the continuum, rigid model of the carbon atoms of the tube. We studied mainly the regime of low coverage, where interactions between adsorbed atoms are omitted; in the IC case, this assumption was not needed, as a quasi-one dimensional approximation permits exact treatment of LJ interactions at finite coverages. The conclusions drawn are expected to be qualitatively accurate in general situations, and so they provide useful insight for experiments. The key result appears in Fig. 1, indicating which molecules go where under “typical” experimental conditions. More general behavior can be estimated from the reduced potential curves (Figures 3-5) we have presented.
Williams and Eklund have computed the H<sub>2</sub> adsorption on the bounding surface of bundles containing a finite number of tubes. In some cases, this contribution can be a significant fraction of the total adsorption. Adsorption isotherms of classical gases on the external surface of the bundle is the subject of our current investigations to be reported in the future.
We discuss the relevant experiments very briefly. Teizer et al. studied He uptake and found consistency with our calculations for one-dimensional motion and the computed binding energy within the interstitial channels. Kuznetsova et al. studied uptake of Xe and their data are consistent with our calculations of the uptake within the nanotubes.
Interestingly, a recent experimental study of adsorption of methane in nanotube bundles concluded that significant IC adsorption occurs. This conclusion was reached from the fact that nanotubes were capped and the measured binding energy of CH<sub>4</sub> determined (2570 K) was 76% larger than that on graphite (1460 K) , which compared favorably with previous estimates of the IC binding energy of H<sub>2</sub>, He and Ne . Our present calculations indicate, however, that the large size of CH<sub>4</sub> prevent it from populating the narrow IC’s significantly. In contrast, the external surface of the nanotube bundle is accessible and the binding energy in this case (Table I) is $``$ 20% larger than the one for graphite. A more realistic potential exhibits corrugation, which we have neglected here; its effect is to increase the binding energy , but we have not undertaken that calculation as yet so no definite comparison is possible.
## ACKNOWLEDGMENTS
We are grateful to Victor Bakaev, Moses Chan, Vincent Crespi, Peter Eklund, Karl Johnson, James Kurtz, Aldo Migone, Bill Steele and Keith Williams for useful discussions. This research was supported by the National Science Foundation, the Petroleum Research Fund of the American Chemical Society and the Army Research Office. One of the authors, S. C., would like to acknowledge the generous support of Fondazione Ing. Aldo Gini.
Figure Captions
FIG. 1(a)
FIG. 1(b)
FIG. 1(c)
FIG. 1(d)
FIG. 2
FIG. 3
FIG. 4
FIG. 5
FIG. 6
FIG. 7
FIG. 8 |
warning/0001/hep-th0001098.html | ar5iv | text | # Electromagnetic dipole radiation of oscillating D-branes
## 1 INTRODUCTION
Probably the best way to introduce D-branes is to recall that in non-Abelian gauge field theories there exist classical BPS solutions corresponding to new particles, extended objects, vortices and monopoles which are not present in the perturbative spectrum of the quantized theory . In these theories in addition to the point particle spectrum of perturbation theory consisting of a massless photon, massive vector bosons and Xiggs particles there also exist stable solutions with masses inversely proportional to the square of the coupling constant
$$(Mass)_{BPS}^2=4\pi M_W/e^2.$$
(1)
where $`M_W`$ is a vector boson mass. In the case of vortices it is energy per unit of length and should be associated with the string tension. These states cannot be seen in the weak coupling limit simply because the masses of these states become very large when $`e0`$, therefore in gauge theories we have mass hierarchy: the particle masses are of order $`O(e^0)`$ and the soliton masses are of order $`O(1/e^2)`$. These states are invisible in perturbation theory. One can probably think about these solutions as ” typhoons” of gauge bosons.
The second important message which bring to us these solutions is that extended objects in gauge theories carry a new type of charge, in the given case it is magnetic charge , which was not present in perturbative spectrum, as well
$$g=2\pi /e.$$
(2)
The magnetic charge satisfies the Dirac quantization condition
$$eg=2\pi nnZ,$$
(3)
which ties together the units of electric and magnetic charges . (Generally speaking magnetic charge is realized as a genuine solution without singularities when $`U(1)`$ is embedded into non-Abelian compact group .) With the existence of magnetic charges and also dyons – particles carrying electric and magnetic charges the theory may exhibit full electromagnetic duality . According to this conjecture, the strong coupling limit of the theory is equivalent to the weak coupling limit with ordinary particles and solitons exchanged.
In string theory one can also expect solutions which cannot be seen in the perturbation theory. In string theory we do not have the full set of equations which define the theory to all orders of coupling constant as it was in gauge theories, nevertheless such solutions have been found as solutions of low energy superstring effective action . The lagrangian can be represented in the form:
$`S={\displaystyle \frac{1}{16\pi G_N^{(10)}}}{\displaystyle d^{10}x\sqrt{g}}`$
$`\{e^{2\phi }[R+4(\phi )^2{\displaystyle \frac{1}{3}}H^2]{\displaystyle \frac{1}{3}}G^2\}`$ (4)
where $`\phi `$ is the dilaton field, $`H`$ \- the NS-NS gauge field and $`G`$ \- RR gauge field have essentially different order in dilaton coupling for RR fields. The corresponding equations have two types of solutions carrying NS-NS or RR charges .
The NS-NS soliton is determined by the balance between the first three terms and its tension (tension $``$ mass/volume ), scales as the action (4)
$$(Tension)_{NSNS}\frac{1}{g_s^2}$$
(5)
where $`e^\phi =g_s`$ is the string coupling constant. In NS-NS sector the electrically charged object is the fundamental string of the mass $`O(g_s^0)`$ and is accompanied by the above solitonic D5-brane carring the dual magnetic charge of the mass $`O(1/g_s^2)`$. This is very similar to what we had in gauge theories.
In RR sector the situation is different, all string states are neutral with respect to RR gauge field and in perturbative string spectrum there is no fundamental objects carrying elementary RR charges . In the RR sector the solution of the equations is determined by the balance between the first three and the fourth term in the effective action (4) and the tension of the solution scales as
$$(Tension)_{RR}\frac{1}{g_s}$$
(6)
so that both types of RR charges, electric p-brane and its dual magnetic 6-p brane, are solitons of the mass $`O(1/g_s)`$ which is smaller than that of the NS-NS solution (5). This behaviour, intermediate between fundamental string spectrum $`O(g_s^0)`$ and solitonic brane $`O(1/g_s^2)`$, was not known in field theory (see Figure 3).
The existence of dynamical extended objects in string theory which carry RR charges can be deduced also by considering string theory with mixed boundary conditions . Indeed one can consider type II closed superstring theory which has been completed by adding open strings with Neumann boundary conditions on $`p+1`$ directions and Dirichlet conditions on the remaining 9-p transverse directions ,
$`n^a_aX^\mu =0,\mu =0,\mathrm{}p,`$ (7)
$`X^\mu =0,\mu =p+1,\mathrm{}9.`$ (8)
Here the open string endpoints live on the extended object, D-brane, and in this indirect way defines a string soliton with $`p`$ spatial and one timelike coordinates. The D-branes are surfaces of dimension p on which open string can end. This is a consistent string theory if p is even in type IIA theory and is odd in type IIB theory . The identification with solitons can be seen if one compute the tension
$$\tau _p=\frac{1}{g_s(2\pi )^{\frac{p1}{2}}(2\pi \alpha ^{})^{\frac{p+1}{2}}}$$
(9)
This is the same coupling constant dependence as for the branes carrying RR charges (6)! The RR charge which carry D-brane is
$$\mu _p=\frac{1}{(2\pi )^{\frac{1}{2}}(4\pi ^2\alpha ^{})^{\frac{p3}{2}}}.$$
(10)
and satisfies the Dirac quantization condition
$$\mu _p\mu _{6p}=2\pi nnZ$$
(11)
with $`n=1`$. Thus RR gauge fields naturally couple to Dp-branes and identify Dp-branes as carrier of RR charge . The general remark is that the tension of the p-brane should have dimension $`mass^{p+1}`$ and it is indeed so in (9), but what is important is that it is a function of the fundamental string tension $`1/2\pi \alpha ^{}`$ (as well as $`1/g_s`$) and not just a new parameter of the theory . The D-branes are indeed the ”typhoons” of strings.
## 2 THE D-BRANE EXCITATIONS
The Dp-brane is a dynamical extended object which can fluctuate in shape and position and these fluctuations are described by the attached open strings. The boundary condition (8) allows open string ends to be at any place inside the p-brane with coordinates $`x^0,\mathrm{},x^p`$ and $`x^{p+1}=\mathrm{}=x^9=0`$. In this way the open string describes the excitations of the Dp-brane, and as it is well known the open string massless states are vector and spinor fields of ten-dimensional $`N=1`$ supersymmetric $`U(1)`$ gauge theory in ten dimensions. The massless field $`A_\mu (x^\nu ),\mu ,\nu =0,\mathrm{},p`$ propagates as gauge bosons on the p-brane worldvolume, while the other components of the vector potential $`A_4(x^\nu ),\mathrm{},A_9(x^\nu )`$ can be interpreted as oscillation in the position of the p-brane. Thus the theory on the p-brane worldvolume is a ten-dimensional supersymmetric $`U(1)`$ gauge theory dimensionally reduced to $`p+1`$ dimensions. The low energy effective action for these fields is the Dirac-Born-Infeld lagrangian . The key difference with previously known p-branes and supermembranes is the inclusion of a worldvolume electromagnetic field.
Callan and Maldacena showed that the Dirac-Born-Infeld action, can be used to build a configuration with a semi-infinite fundamental string ending on a 3-brane<sup>1</sup><sup>1</sup>1 We review the favorite case of the 3-brane first because of its non-singular behaviour in SUGRA, and secondly it is suggestive of our own world which is after all 3-dimensional., whereby the string is actually made out of the brane wrapped on $`S_2`$ (see also ). The relevant action can be obtained by computing a simple Born-Infeld determinant, dimensionally reduced from 10 dimensions
$$L=\frac{1}{(2\pi )^3g_s}d^4x\sqrt{1\stackrel{}{E}^2+(\stackrel{}{}x_9)^2},$$
(12)
where $`g_s`$ is the string coupling ($`\alpha ^{}=1`$).
The above mentioned theory contains 6 scalars $`x_4,\mathrm{}.,x_9,`$ which are essentially Kaluza-Klein remnants from the 10-dimensional $`N=1`$ electrodynamics after dimensional reduction to $`3+1`$ dimensions. As we explain these extra components of the e.m. field $`A_4,\mathrm{},A_9`$ describe the transverse deviations of the brane $`x_4,\mathrm{},x_9`$.
The solution, which satisfies the BPS conditions, is necessarily also a solution of the linear theory, the $`N=4`$ super-Electrodynamics ,
$$\stackrel{}{E}=\frac{c}{r^2}\stackrel{}{e_r},\stackrel{}{}x_9=\frac{c}{r^2}\stackrel{}{e_r},x_9=\frac{c}{r},$$
(13)
where $`c=2\pi ^2g_s`$ is the unit charge, and can be thought of as setting the distance scale. Here the scalar field represents the geometrical spike, and the electric field insures that the string carries uniform NS charge along it. The RR charge of the 3-brane
$$\mathrm{\Omega }^{MNL}=ϵ^{abc}_aX^M_bX^N_cX^L$$
(14)
cancels out on the string, or rather the tube behaves as a kind of RR dipole whose magnitude can be ignored when the tube becomes thin. It was demonstrated in that the infinite electrostatic energy of the point charge can be reinterpreted as being due to the infinite length of the attached string. The energy per unit length comes from the electric field and corresponds exactly to the fundamental string tension.
Polchinski, when he introduced D-branes as objects on which strings can end, required that the string has Dirichlet (fixed) boundary conditions for coordinates normal to the brane (8), and Neumann (free) boundary conditions for coordinate directions parallel to the brane (7) .
It was shown in that small fluctuations which are normal to both the string and the brane are mostly reflected back with a $`phaseshift\pi `$ which indeed corresponds to Dirichlet boundary condition (8). See also and for a supergravity treatment of this problem.
In the paper it was demonstrated that P-wave excitations which are coming down the string with a polarization along a direction parallel to the brane are almost completely reflected just as in the case of all-normal excitations, but the end of the string moves freely on the 3-brane, thus realizing Polchinski’s open string Neumann boundary condition (7) dynamically. As we will see a superposition of excitations of the scalar $`x_9`$ and of the electromagnetic field reproduces the required behaviour, e.g. reflection of the geometrical fluctuation with a $`phaseshift0`$ (Neumann boundary condition).
In it was also observed the electromagnetic dipole radiation which escapes to infinity from the place where the string is attached to the 3-brane. It was shown that in the low energy limit $`\omega 0`$, i.e. for wavelengths much larger than the string scale a small fraction $`\omega ^4`$ of the energy escapes to infinity in the form of electromagnetic dipole radiation. The physical interpretation is that a string attached to the 3-brane manifests itself as an electric charge, and waves on the string cause the end point of the string to freely oscillate and produce electromagnetic dipole radiation in the asymptotic outer region of the 3-brane. This result is also in a good agreement with the interpretation of the open string ending on D-brane (7), according to which the open string states describe the excitations of the D-brane. Indeed as we will see open string massless vector boson can propagate inside the D-brane and in our case this excitation is identified with electromagnetic dipole radiation. Thus not only in the static case, but also in a more general dynamical situation the above interpretation remains valid. This result provides additional support to the idea that the electron (quark) may be understood as the end of a fundamental string ending on a D-brane.
## 3 THE LAGRANGIAN AND THE EQUATIONS
Let us write out the full Lagrangian which contains both electric and magnetic fields, plus the scalar $`x_9\varphi `$
$`L={\displaystyle d^4x\sqrt{Det}}`$ (15)
$`whereDet=1+\stackrel{}{B}^2\stackrel{}{E}^2(\stackrel{}{E}\stackrel{}{B})^2`$
$`(_0\varphi )^2(1+\stackrel{}{B}^2)+(\stackrel{}{}\varphi )^2+(\stackrel{}{B}\stackrel{}{}\varphi )^2`$
$`(\stackrel{}{E}\times \stackrel{}{}\varphi )^2+2_0\varphi (\stackrel{}{B}[\stackrel{}{}\varphi \times \stackrel{}{E}])`$
We will proceed by adding a fluctuation to the background values (13) :
$$\stackrel{}{E}=\stackrel{}{E}_0+\delta \stackrel{}{E},\stackrel{}{B}=\delta \stackrel{}{B},\varphi =\varphi _0+\eta .$$
Then keeping only terms in the $`Det`$ which are linear and quadratic in the fluctuation we will get
$`\delta Det=\delta \stackrel{}{B}^2\delta \stackrel{}{E}^2(\stackrel{}{E_0}\delta \stackrel{}{B})^2(_0\eta )^2`$ (16)
$`+(\stackrel{}{}\eta )^2+(\delta \stackrel{}{B}\stackrel{}{}\varphi )^2(\stackrel{}{E_0}\times \stackrel{}{}\eta )^2(\delta \stackrel{}{E}\times \stackrel{}{}\varphi )^2`$
$`2(\stackrel{}{E_0}\times \stackrel{}{}\eta )(\delta \stackrel{}{E}\times \stackrel{}{}\varphi )2(\stackrel{}{E_0}\delta \stackrel{}{E})+2(\stackrel{}{}\varphi \stackrel{}{}\eta )`$
Note that one should keep the last two linear terms because they produce additional quadratic terms after taking the square root. These terms involve the longitudinal polarization of the e.m. field and cancel out at quadratic order. The resulting quadratic Lagrangian is
$`2L_q=\delta \stackrel{}{E}^2(1+(\stackrel{}{}\varphi )^2)\delta \stackrel{}{B}^2+(_0\eta )^2`$ (17)
$`(\stackrel{}{}\eta )^2(1\stackrel{}{E_0}^2)+\stackrel{}{E_0}^2(\stackrel{}{}\eta \delta \stackrel{}{E}).`$
Let us introduce the gauge potential for the fluctuation part of the e.m. field as $`(A_0,\stackrel{}{A})`$ and substitute the values of the background fields from (13)
$`2L_q=(_0\stackrel{}{A}\stackrel{}{}A_0)^2(1+{\displaystyle \frac{1}{r^4}})(\stackrel{}{}\times \stackrel{}{A})^2`$ (18)
$`+(_0\eta )^2(\stackrel{}{}\eta )^2(1{\displaystyle \frac{1}{r^4}})+{\displaystyle \frac{1}{r^4}}(_0\stackrel{}{A}\stackrel{}{}A_0)\stackrel{}{}\eta .`$
The equations that follow from this lagrangian contain dynamical equations for the vector potential and for the scalar field, and a separate equation which represents a constraint. These equations in the Lorenz gauge $`\stackrel{}{}\stackrel{}{A}=_0A_0`$ are
$`_0^2\stackrel{}{A}(1+{\displaystyle \frac{1}{r^4}})+\mathrm{\Delta }\stackrel{}{A}+{\displaystyle \frac{1}{r^4}}\stackrel{}{}_0(A_0+\eta )=0`$ (19)
$`_0^2A_0+\mathrm{\Delta }A_0+\stackrel{}{}{\displaystyle \frac{1}{r^4}}\stackrel{}{}(A_0+\eta )\stackrel{}{}{\displaystyle \frac{1}{r^4}}_0\stackrel{}{A}=0`$ (20)
$`_0^2\eta +\mathrm{\Delta }\eta \stackrel{}{}{\displaystyle \frac{1}{r^4}}\stackrel{}{}(A_0+\eta )+\stackrel{}{}{\displaystyle \frac{1}{r^4}}_0\stackrel{}{A}=0`$ (21)
Equation (20) is a constraint: the time derivative of the lhs is zero, as can be shown using the equation of motion (19).
Let us choose $`A_0=\eta `$. This condition can be viewed as (an attempt to) preserve the BPS relation which holds for the background: $`\stackrel{}{E}=\stackrel{}{}\varphi `$. Another point of view is that this fixes the general coordinate invariance which is inherent in the Born-Infeld lagrangian in such a way as to make the given perturbation to be normal to the surface. Of course transversality is insured automatically but this choice makes it explicite. The general treatment of this subject can be found in .
With this condition the equations (20) and (21) become the same, and the first equation is also simplified :
$`_0^2\stackrel{}{A}(1+{\displaystyle \frac{1}{r^4}})+\mathrm{\Delta }\stackrel{}{A}=0,`$ (22)
$`_0^2\eta +\mathrm{\Delta }\eta +\stackrel{}{}{\displaystyle \frac{1}{r^4}}_0\stackrel{}{A}=0.`$ (23)
This should be understood to imply that once we obtain a solution, $`A_0`$ is determined from $`\eta `$, but in addition we are now obliged to respect the gauge condition which goes over to $`\stackrel{}{}\stackrel{}{A}=_0\eta `$.
## 4 NEUMANN BOUNDARY CONDITIONS FROM BORN-INFELD DYNAMICS
We will seek a solution with definite energy (frequency $`w`$) in the following form: $`\stackrel{}{A}`$ should have only one component $`A_z`$, and $`\eta `$ be an $`l=1`$ spherical $`P`$-wave
$$A_z=\zeta (r)e^{i\omega t},\eta =\frac{z}{r}\psi (r)e^{i\omega t}$$
The geometrical meaning of such a choice for $`\eta `$ is explained in Figure 1, and the particular choice of $`z`$ dependence corresponds to the polarization of the oscillations along the $`z`$ direction of the brane. With this ansatz the equations become
$`\left(1+{\displaystyle \frac{c^2}{r^4}}\right)\omega ^2\zeta +{\displaystyle \frac{c^2}{r^2}}_r(r^2_r\zeta )=0,`$ (24)
$`{\displaystyle \frac{z}{r}}\omega ^2\psi +{\displaystyle \frac{z}{r}}{\displaystyle \frac{1}{r^2}}_r(r^2_r\psi )+{\displaystyle \frac{z}{r}}{\displaystyle \frac{2}{r^2}}\psi `$
$`i\omega _z\left({\displaystyle \frac{c^2}{r^4}}\zeta \right)=0,`$ (25)
with the gauge condition becoming $`_r\zeta =i\omega \psi `$. It can be seen again, that the second equation follows from the first by differentiation. This is because the former coincides with the constraint in our ansatz.
Therefore the problem is reduced to finding the solution of a single scalar equation, and determining the other fields through subsidiary conditions. The equation (24) itself surprisingly turned out to be the one familiar from for the transverse fluctuations. There it was solved by going over to a coordinate $`\xi `$ which measures the distance radially along the surface of the brane
$$\xi (r)=\omega \underset{\sqrt{c}}{\overset{r}{}}𝑑u\sqrt{1+\frac{c^2}{u^4}},$$
and a new wavefunction
$$\stackrel{~}{\zeta }=\zeta (c^2+r^4)^{1/4}.$$
This coordinate behaves as $`\xi r`$ in the outer region ($`r\mathrm{}`$) and $`\xi 1/r`$ on the string ($`r0`$). The exact symmetry of the equation $`rc/r`$ goes over to $`\xi \xi `$. The equation, when written in this coordinate becomes just the free wave equation, plus a narrow symmetric potential at $`\xi 0`$
$$\frac{d^2}{d\xi ^2}\stackrel{~}{\zeta }+\frac{5c^2/\omega ^2}{(r^2+c^2/r^2)^3}\stackrel{~}{\zeta }=\stackrel{~}{\zeta }.$$
(26)
The asymptotic wave functions can be constructed as plain waves in $`\xi `$,
$$\zeta (r)=(c^2+r^4)^{1/4}e^{\pm i\xi (r)},$$
or in the various limits:
$`r0\zeta {\displaystyle \frac{1}{\sqrt{c}}}e^{\pm i\xi (r)},`$
$`r\mathrm{}\zeta {\displaystyle \frac{1}{r}}e^{\pm i\xi (r)}.`$
These formulae give us the asymptotic wave function in the regions $`\xi \pm \mathrm{}`$, while around $`\xi =0(r=1)`$ there is a symmetric repulsive potential which drops very fast $`1/\xi ^6`$ on either side of the origin. The scattering is described by a single dimensionless parameter $`\omega \sqrt{c}`$, and in the limit of small $`\omega `$ and/or coupling $`c=2\pi ^2g_s`$ the potential becomes narrow and high, and can be replaced by a $`\delta `$-function with an equivalent area $`\frac{1}{\omega \sqrt{g_s}}`$ under the curve. Therefore the scattering matrix becomes almost independent of the exact form of the potential. We refer the reader to pp 18-20 of the thesis for the more detailed calculation of transmission amplitude (29). The end result is that most of the $`\zeta `$ amplitude is reflected back with a phase shift close to $`\pi `$.
In order to obtain $`\psi `$ (and $`\eta `$) we need to differentiate $`\zeta `$ with respect to $`r`$:
$`i\omega \psi ={\displaystyle \frac{1}{4}}{\displaystyle \frac{4r^3}{(c^2+r^4)^{5/4}}}e^{\pm i\xi (r)}`$
$`\pm {\displaystyle \frac{i\omega }{(c^2+r^4)^{1/4}}}\left(1+{\displaystyle \frac{c^2}{r^4}}\right)^{1/2}e^{\pm i\xi (r)}.`$ (27)
Again it is easy to obtain the simplified limiting form:
$`r0i\omega \psi \left({\displaystyle \frac{r^3}{c^{5/2}}}\pm {\displaystyle \frac{i\omega \sqrt{c}}{r^2}}\right)e^{\pm i\xi (r)}`$
$`r\mathrm{}i\omega \psi \left({\displaystyle \frac{1}{r^2}}\pm {\displaystyle \frac{i\omega }{r}}\right)e^{\pm i\xi (r)}`$
This brings about several consequences for $`\psi `$. Firstly, it causes $`\psi `$ to grow as $`1/r^2`$ when $`r0`$. This is the correct behaviour because when converted to displacement in the $`z`$ direction, it means constant amplitude :
$$\eta =\frac{z}{r}\psi =\delta \frac{c}{r}=\frac{z}{r}\frac{c}{r^2}\delta z\delta zconst.$$
(28)
Secondly, the $`i`$ that enters causes the superposition of the incoming and reflected waves to become a cosine from a sine, as is the case for $`\zeta `$ waves. This corresponds to a $`0`$ phase shift for the $`\eta `$ wave (Figure 2) and implies therefore Neumann boundary condition , that is it allows open string ends to be at any place inside the p-brane .
## 5 ELECTROMAGNETIC DIPOLE RADIATION OF OSCILLATING D- BRANE
Because of the $`\omega `$ factor in the gauge condition we need to be careful about normalizations, thus we shall choose to fix the amplitude of the $`\delta z`$ (or $`\eta `$) wave to be independent of frequency $`\omega `$. Then the magnitude of the electromagnetic field in the inner region becomes independent of $`\omega `$ as well, thus the amplitude of $`\zeta `$ wave at $`r0`$ is
$$A_z=\zeta _0e^{i\omega (t+c/r)}$$
Combined with the transmission factor
$$T=i\omega \sqrt{c}+O(\omega ^2),$$
(29)
this gives the following form of the dipole radiation field at infinity<sup>2</sup><sup>2</sup>2The unit of electric charge in our notation is $`2\pi ^2g_s=e^2`$. This is because of the scaling of the fields needed to get the U(1) action with $`1/e^2`$ in front of it.
$`A_z={\displaystyle \frac{T\zeta _0}{r}}e^{i\omega t}={\displaystyle \frac{i\omega \sqrt{2\pi ^2g_s}}{r}}\zeta _0e^{i\omega t}`$
$`={\displaystyle \frac{i\omega e\zeta _0}{r}}e^{i\omega t}={\displaystyle \frac{\dot{d}}{r}},`$ (30)
where
$$d=e\zeta _0e^{i\omega t}.$$
In order to make a comparison with Thompson formula
$$I=\frac{4}{3}\omega ^4e^2\zeta _{0}^{}{}_{}{}^{2}$$
for the total power emitted by an oscillating dipole, we note that the agreement is guaranteed if the exact coefficient behind $`i\omega \sqrt{c}`$ in (29) is equal to one. As it was shown in , the approximate computation of the transmission amplitude $`T`$ (29) gives $`Ti.92\omega \sqrt{c}`$. The later computation of this coefficient using the numerical solution of the equation (24) shows that it is indeed equal to one. To solve properly the equation on a computer it is necessary to introduce new variables
$$y=r\frac{c}{r},orr(y)=\frac{y+\sqrt{y^2+4c}}{2},$$
(31)
so that the equation (24) take the form
$$\omega ^2(c^2+r(y)^4)\zeta +(c+r(y)^2)_y(c+r(y)^2)_y\zeta =0$$
(32)
and boundary condition at $`r_0=\omega c/2\pi n`$, $`n\mathrm{}`$ the form
$`\zeta (y_0)={\displaystyle \frac{1}{\sqrt{c}}},\zeta ^{}(y_0)=i{\displaystyle \frac{(2\pi n)^2}{\omega c^{3/2}}},`$
$`aty_0={\displaystyle \frac{\omega c}{2\pi n}}{\displaystyle \frac{2\pi n}{\omega }}\mathrm{}.`$
Measuring the amplitudes $`y\zeta (y)\pm (y+\frac{\pi }{2\omega })\zeta (y+\frac{\pi }{2\omega })`$ at $`y\mathrm{}(r\mathrm{})`$ one can get that the coefficient behind $`i\omega \sqrt{c}`$ in (29) is equal to one. Thus the magnitude of emitted power is in fact exactly equal to the one given by Thompson formula in electrodynamics<sup>3</sup><sup>3</sup>3see also .
In conclusion, we need to analyze the outgoing scalar wave. This wave has both real and imaginary parts, the imaginary part is $`1/r^2`$ and drops faster than radiation. The real part does contribute to the radiation at spatial infinity, as can be shown from the integral of the energy density $`(_r\eta )^2d^3r\omega ^4/r^24\pi r^2𝑑r\omega ^4`$. This is not altogether surprising, as we are dealing with a supersymmetric theory where the different fields are tied together. In fact, it is easy to show that the fields still do asymptotically satisfy the BPS conditions at infinity, and $`1/4`$ of the SUSY is preserved. Thus the observer at spatial infinity will see both an electromagnetic dipole radiation field and a scalar wave. Interestingly, the direction dependences of the two conspire to produce a spherically symmetric distribution of the energy radiated.
I wish to thank the organizers of the conference QG99 in Villasimius and especially Vittorio de Alfaro for the invitation and for arranging interesting and stimulating meeting. I also thank Konstantin Savvidy for the enjoyable collaboration on the work reported here. The work was supported in part by the EEC Grant no. HPRN-CT-1999-00161 and ESF Grant ”Geometry and Disorder: From Membranes to Quantum Gravity”. |
warning/0001/cond-mat0001383.html | ar5iv | text | # Far-infrared excitations in a quantum antidot at finite magnetic fields
## I Introduction
The far-infrared (FIR) spectroscopy of laterally confined superlattices in the two dimensional electron gas (2DEG) made of holes surrounded by electrons, called antidots, has uncovered a very peculiar behavior of the collective spectrum of these systems. It has been found that it consists of a high frequency branch which starts either with a negative magnetic field $`(B)`$ dispersion in the case of high electronic surface density $`n_s`$, or with a rather flat $`B`$ dispersion at low $`n_s`$, that eventually converges to the cyclotron energy at high $`B`$’s, plus a low frequency branch which at high $`B`$’s corresponds to the usual edge magnetoplasmon, but that approaches the cyclotron frequency for small magnetic fields. Anticrossing of the modes appears as $`B`$ increases. It is worth to recall that the mentioned behavior of the high frequency branch is also a feature of the FIR spectrum of quantum rings.
A study employing circularly polarized radiation has shown that both edge and bulk magnetoplasmons in antidots exhibit the same circular polarization, in contradistinction with what is found in quantum dots. Interestingly, a very recent experiment has detected weak bulk modes anticyclotron-like polarized, whose existence had been predicted some time ago.
Another difference with dots appeared after a careful analysis of the edge magnetoplasmon: the frequency of this mode shows a conspicuous oscillation with $`B`$ that has maxima at fully occupied Landau levels (even filling factors $`\nu `$) in the case of dots, and at half-filled Landau levels (odd $`\nu `$’s) in the case of antidots.
Several theoretical descriptions of FIR modes in antidots have been given in the past. Some are based on hydrodynamical or classical electrodynamics models, which either do not take into account the periodicity of the antidot array, or which take it into account either in a circularly symmetric Wigner-Seitz approximation, or incorporate the parameters of the experimental short-period antidot lattices without further approximations.
Despite that all these models have somehow succesfully reproduced the FIR absorption in antidots, the interest in achieving a more microscopic description of the ground state (gs) and FIR response still remains. Quantum mechanical (Hartree) models have been set up which take into account the periodicity of the lattice using confining potentials of different complexity. The difficulties inherent to the handling of many single particle (sp) wave functions and realistic confining potentials which reflect the actual antidot lattice have hampered these microscopic methods to achieve a quantitative description of the FIR absorption in antidots, although some features of the process are qualitatively described.
Recently, we have applied a local spin-density functional (LSDFT) method to calculate the structure of isolate antidots at $`B=0`$, together with a sum rule approach to describe their FIR spectrum. Our aim here is to extend these calculations to the case of finite magnetic fields and to present an account of the gs and dipole response of an antidot within the frame of LSDFT and its time-dependent generalization (TDLSDFT). We recall that these formalisms incorporate the spin degree of freedom and allow to selfconsistently include exchange and correlation effects in the study of the gs and the response of the system, whose importance in the description of 2D electronic structures is nowadays well established.
Rather than isolated antidots, the systems described in this work should be considered as representing long-period antidot lattices. Although no much differences are expected to appear between the ground state structure of an isolated antidot and that of an antidot in a long-period array, apart from the obvious changes at the border of the unit cell, bulk magnetoplasmons are qualitatively different in both cases.
## II The LSDFT and TDLSDFT approaches
We have modeled an antidot of radius $`R`$ in a 2DEG of surface density $`n_s`$ by a positive jellium background of density $`n_J(r)=n_s\mathrm{\Theta }(rR)`$. The gs of the antidot is obtained solving the Kohn-Sham (KS) equations. The problem is simplified by the imposed circular symmetry, which allows one to write the single particle (sp) wave functions as $`\varphi _{nl\sigma }(r,\theta )=u_{nl\sigma }(r)e^{ıl\theta }`$ with $`l=0,\pm 1,\pm 2,\mathrm{}`$, being $`l`$ the sp orbital angular momentum.
We have used effective atomic units defined by $`\mathrm{}=e^2/ϵ=m=`$1, where $`ϵ`$ is the dielectric constant, and $`m`$ the electron effective mass. In units of the bare electron mass $`m_e`$ one has $`m=m^{}m_e`$. In this system of units, the length unit is the effective Bohr radius $`a_0^{}=a_0ϵ/m^{}`$, and the energy unit is the effective Hartree $`H^{}=Hm^{}/ϵ^2`$. In the numerical applications we have always considered GaAs, for which we have taken $`ϵ`$ = 12.4, $`m^{}`$ = 0.067, and $`g^{}=`$0.44. This yields $`a_0^{}`$ 97.94 $`\mathrm{\AA }`$ and $`H^{}`$ 11.86 meV $``$ 95.6 cm<sup>-1</sup>. The Bohr magneton is $`\mu _B=\mathrm{}e/2m_ec`$, and the cyclotron frequency is $`\omega _c=eB/mc`$.
The radial KS equations read
$`[{\displaystyle \frac{1}{2}}({\displaystyle \frac{d^2}{dr^2}}+{\displaystyle \frac{1}{r}}{\displaystyle \frac{d}{dr}}{\displaystyle \frac{l^2}{r^2}}){\displaystyle \frac{\omega _c}{2}}l+{\displaystyle \frac{1}{8}}\omega _c^2r^2V^+(r)`$ (1)
$`+`$ $`V^H+V^{xc}+(W^{xc}+{\displaystyle \frac{1}{2}}g^{}\mu _BB)\eta _\sigma ]u_{nl\sigma }=ϵ_{nl\sigma }u_{nl\sigma },`$ (3)
where $`\eta _\sigma =+1(1)`$ for $`\sigma =()`$, $`V^H=𝑑\stackrel{}{r}^{}n(\stackrel{}{r}^{})/|\stackrel{}{r}\stackrel{}{r}^{}|`$ is the Hartree potential, and $`V^{xc}=_{xc}(n,m)/n|_{gs}`$ and $`W^{xc}=_{xc}(n,m)/m|_{gs}`$ are the variations of the exchange-correlation energy density $`_{xc}(n,m)`$ written in terms of the electron density $`n(r)`$ and of the local spin magnetization $`m(r)n^{}(r)n^{}(r)`$ taken at the ground state. $`_{xc}(n,m)`$ has been built from the 2DEG calculations of Tanatar and Ceperley at zero and full spin polarization, using the Von Barth and Hedin interpolation formula adapted to 2D. The jellium potential $`V^+(r)`$ is analytical:
$$V^+(r)=4n_s\times \{\begin{array}{cc}R_{\mathrm{}}𝐄(r/R_{\mathrm{}})R𝐄(r/R)\hfill & r<R\hfill \\ R_{\mathrm{}}𝐄(r/R_{\mathrm{}})r𝐄(R/r)+r[1(R/r)^2]𝐊(R/r)\hfill & r>R,\hfill \end{array}$$
(4)
where K and E are the complete elliptic integrals of first and second kind, respectively , and $`R_{\mathrm{}}`$ represents a large $`r`$ value. In practice, it is the largest-$`r`$ used in the gs calculation. The KS differential equations have been solved without expanding the sp wave functions in a necessarily truncated basis of Landau orbitals. Our iterative method works for weak and strong $`B`$ fields as well, for which the effective potential is very different. It has the advantage of avoiding to study how the results depend on the size of the basis. As in Ref. , we have worked at a small but finite temperature (0.1 K).
Physically acceptable solutions have to be regular at $`r=0`$, and have to be bound due to the $`\omega _c^2`$ term in the KS equation. Far from $`R`$ they behave as
$$u_{nl\sigma }(r)\left(\frac{r}{}\right)^{|l|}e^{(r/2)^2}L_n^{|l|}\left(\frac{r^2}{2^2}\right),$$
(5)
where $`\sqrt{\mathrm{}/m\omega _c}`$ is the magnetic length and $`L_n^{|l|}`$ is a generalized Laguerre polynomial.
Once the KS ground state has been obtained, we determine the dipole longitudinal response employing linear-response theory. We sketch here how the method is applied.
For independent electrons in the KS mean field, the variation $`\delta n_\sigma ^{(0)}`$ induced in the spin density $`n_\sigma `$ ($`\sigma ,`$) by an external field $`F`$, whose non-temporal dependence we denote as $`F=_\sigma f_\sigma (\stackrel{}{r})|\sigma \sigma |`$, can be written as
$$\delta n_\sigma ^{(0)}(\stackrel{}{r},\omega )=\underset{\sigma ^{}}{}𝑑\stackrel{}{r}^{}\chi _{\sigma \sigma ^{}}^{(0)}(\stackrel{}{r},\stackrel{}{r}^{};\omega )f_\sigma ^{}(\stackrel{}{r}^{}),$$
(6)
where $`\chi _{\sigma \sigma ^{}}^{(0)}`$ is the KS spin density correlation function. In this limit, the frequency $`\omega `$ corresponds to the harmonic time-dependence of the external field $`F`$ and of the induced $`\delta n_\sigma ^{(0)}`$. Eq. (6) is a 2$`\times `$2 matrix equation in the two-component Pauli space. In longitudinal response theory, $`F`$ is diagonal in this space, and its diagonal components are written as a vector $`F\left(\begin{array}{c}f_{}\\ f_{}\end{array}\right)`$. We consider external dipole ($`L`$ = 1) fields of the kind
$$F_{\pm 1}^{(n)}=f(r)e^{\pm ı\theta }\left(\begin{array}{c}1\\ 1\end{array}\right)\mathrm{and}F_{\pm 1}^{(m)}=f(r)e^{\pm ı\theta }\left(\begin{array}{c}1\\ 1\end{array}\right)$$
(7)
which cause, respectively, charge and spin density modes of dipole type. To distinguish the induced densities in each excitation channel, we label them with an additional superscript as $`\delta n_\sigma ^{(0,n)}`$ or $`\delta n_\sigma ^{(0,m)}`$.
The TDLSDFT induced densities are obtained solving the integral equations
$`\delta n_\sigma ^{(A)}(\stackrel{}{r},\omega )=\delta n_\sigma ^{(0,A)}(\stackrel{}{r},\omega )+{\displaystyle \underset{\sigma _1\sigma _2}{}}{\displaystyle 𝑑\stackrel{}{r}_1𝑑\stackrel{}{r}_2\chi _{\sigma \sigma _1}^{(0)}(\stackrel{}{r},\stackrel{}{r}_1;\omega )K_{\sigma _1\sigma _2}(\stackrel{}{r}_1,\stackrel{}{r}_2)\delta n_{\sigma _2}^{(A)}(\stackrel{}{r}_2,\omega )},`$ (8)
where either $`A=n`$ or $`A=m`$, and the kernel $`K_{\sigma \sigma ^{}}(\stackrel{}{r},\stackrel{}{r}^{})`$ is the electron-hole interaction.
Equations (8) have been solved as a generalized matrix equation in coordinate space. Considering angular decompositions of $`\chi _{\sigma \sigma ^{}}`$ and $`K_{\sigma \sigma ^{}}`$ of the kind $`K_{\sigma \sigma ^{}}(\stackrel{}{r},\stackrel{}{r}^{})=_lK_{\sigma \sigma ^{}}^{(l)}(r,r^{})e^{il(\theta \theta ^{})}`$, it is enough to solve them for $`l=1`$ because only this term couples to the external dipole field.
When a magnetic field is perpendicularly applied to the antidot, the $`\pm 1`$ modes are not degenerate and two excitation branches may appear, each having in principle a different circular polarization, i.e., carrying an orbital angular momentum $`\mathrm{\Delta }L_z=\pm 1`$, where $`L_z`$ is that of the gs. In contradistinction with the quantum dot (QD) case, we will see that this is not so for antidots, for which the more intense peaks in both branches have the same polarization.
The induced charge or magnetization densities corresponding to density and spin responses are given by $`\delta n^{(A)}=\delta n_{}^{(A)}+\delta n_{}^{(A)}`$ and $`\delta m^{(A)}=\delta n_{}^{(A)}\delta n_{}^{(A)}`$. ¿From them, the dynamical polarizabilities in the density and spin channels are respectively given by
$`\alpha _{nn}(\omega )`$ $`=`$ $`{\displaystyle 𝑑rrf(r)\delta n^{(n)}(r)}`$ (9)
$`\alpha _{mm}(\omega )`$ $`=`$ $`{\displaystyle 𝑑rrf(r)\delta m^{(m)}(r)}.`$ (10)
In this expression $`\delta n^{(n)}`$ has to be understood as the charge density induced by a spin-independent operator, and $`\delta m^{(m)}`$ as the spin density magnetization induced by a spin-dependent operator. Within longitudinal response theory, cross-channel induced densities such as $`\delta m^{(n)}`$ or $`\delta n^{(m)}`$ may also appear. The above expressions hold for each $`\pm 1`$ circular polarization of the $`F`$ operators defined in Eq. (7). Taking into account both possibilities, we define $`\alpha _{AA}^{(1)}(\omega )\alpha _{AA}(1,\omega )+\alpha _{AA}(1,\omega )`$, whose imaginary part is proportional to the strength function $`S_{AA}^{(1)}(\omega )=\mathrm{Im}[\alpha _{AA}^{(1)}(\omega )]/\pi `$. The peaks appearing in the strength functions are dipole charge density (CDE) or spin density excitations (SDE) caused by the external field. Analogously, the peaks appearing in the strength function which results from using in Eq. (10) the KS induced densities $`\delta n_\sigma ^{(0,A)}`$ instead of the correlated ones $`\delta n_\sigma ^{(A)}`$, correspond to dipole single particle excitations (SPE).
We have used for the $`f(r)`$ function entering Eq. (7) two different choices. One is the standard dipole operator $`f(r)=r`$, and the other one is $`f(r)=1/r`$. As indicated in Ref. , the latter choice is inspired in the small-argument expansion of the irregular Bessel function $`Y_1(qr)`$. The relevance of this operator in the present context is that it mostly causes edge excitations. This will be futher discussed in the next Section. The situation is analogous to that faced in the description of surface modes of 3D cavities in metals, see Ref. for a thorough discussion.
## III Results and discussion
As a case of study we have considered an antidot of $`n_s=0.25`$ $`(a_0^{})^2`$ and $`R=7.5`$ $`a_0^{}`$. It roughly corresponds to one of the systems studied by Zhao et al having $`n_s=2.6\times 10^{11}`$ cm<sup>-2</sup> and the same area. We recall that the antidot arrays studied in this reference have a very short period, whereas we have taken for $`R_{\mathrm{}}`$ defined in Eq. (4) a mean value of 30 $`a_0^{}`$, larger for large $`\nu `$’s, and smaller for small $`\nu `$’s. We have employed $`B`$ values yielding integer filling factors $`\nu =2\pi n_s^2`$ in the range $`1\nu 10`$ which corresponds to $`10.8B1.08`$ T.
Figure 1 shows the particle $`n(r)`$ and spin magnetization densities $`m(r)`$ as a function of $`r`$ for $`\nu =1`$ to 6. Away from $`R`$, the electron density approaches $`n_s`$, and either $`m(r)`$ is zero if $`\nu `$ is even, or the local spin polarization $`\xi (r)m(r)/n(r)`$ equals $`1/\nu `$. The number of sp orbitals needed to numerically achieve this limit increases with $`\nu `$, making it rather involved to obtain high-$`\nu `$ ground states. For instance, some 2500 occupied sp levels have been used to model the $`\nu =10`$ gs.
Figure 2 shows the sp energies $`ϵ_{nl\sigma }`$ as a function of $`l`$ for the same configurations as in Fig. 1. In the bulk, $`ϵ_{nl\sigma }`$ are arranged into Landau bands characterized by the index $`Mn+(|l|l)/2`$ and the value of $`\sigma `$. The filling factor represents the number of occupied Landau bands, each of them labeled as $`(M,\mathrm{or})`$. It is worth noticing that since we have taken $`B`$ in the positive $`z`$ direction and the sp orbital angular momentum is written as $`l`$, most occupied sp levels have $`l0`$, and that for a given $`M`$ the $`(M,)`$ band lies below the $`(M,)`$ one.
These bands are almost degenerate when $`\nu `$ is even because of the smallness of the Zeeman energy and because $`m(r)0`$ in the gs. When $`\nu `$ is odd, the spin dependence of the exchange-correlation energy $`_{xc}`$ is responsible of the large gap between the $`(M,)`$ bands. Defining the $`M`$-th Landau level as the set of $`(M,)`$ and $`(M,)`$ bands with $`M=0`$, 1, 2, …, we see from Fig. 2 that even $`\nu `$ values correspond to completely filled Landau levels configurations, and odd $`\nu `$ values to configurations in which one Landau level is half-filled. If $`M`$ is not too small, the $`M`$-th level is made of sp states having $`n=M`$. As in QD’s, this rule is violated by a few sp states per Landau level at most, see for instance Fig. 1 of Ref. , and the $`\nu =6`$ panel of Fig. 2.
The dipole FIR response has been obtained as outlined in Sect. II. Besides evaluating the occupied Landau levels, two more empty levels have been calculated, which allows an accurate description of modes up to $`2\omega _c`$. Figure 3 shows the charge, spin, and single electron strength functions corresponding to the dipole operator. All peaks displayed in this figure are of cyclotron-like type, i.e., they are excited by the $`e^{+ı\theta }`$ component of the $`F`$ operators in Eq. 7. This makes possible the transfering of strength among them, an effect that has been experimentally observed. We remark in passing that the fact that all these peaks originate from the same excitation operator makes the sum rule approach unsuitable to describe their excitation energies.
Our sign convention implies that at $`B0`$ the gs has a negative total orbital angular momentum $`L_z`$. Thus, these excitations decrease in one unit the absolute value of $`L_z`$; whenever needed, we shall label cyclotron-like modes with a $`()`$ sign. Bulk magnetoplasmons in QD’s also bear this character, whereas edge magnetoplasmons do not (see for instance Refs. ). Indeed, edge magnetoplasmons in QD’s are excited by the $`e^{ı\theta }`$ component of the dipole field and correspond to excitations which increase in one unit the absolute value of $`L_z`$. Whenever needed, we shall label the anticyclotron-like modes with a $`(+)`$ sign. The cyclotron-like character of edge magnetoplasmons in antidots has its microscopic origin in the upwards bending of the Landau levels for small $`l`$ values, whereas the anticyclotron-like character of the edge magnetoplasmons in dots comes from the upwards bending of the Landau levels for large $`l`$ values.
As in quantum dots, SDE’s are weakly redshifted with respect to SPE’s because of the attractive character of the rather small exchange-correlation vertex corrections, whereas CDE’s are strongly blueshifted because of the repulsive character of the intense direct Coulomb interaction. In contradistinction with the QD case, no clear signature of a collective longitudinal spin edge magnetoplasmon appears at small $`B`$ values: the edge spin mode is rather fragmented and weak. We recall that at full magnetizacion longitudinal spin and density responses coincide and only spin modes of transverse type, not studied here, can be excited.
Another interesting feature of longitudinal response theory is that spin and charge density modes are generally coupled if the gs has a large spin magnetization, as it happens for instance at $`\nu =3`$. This means that spin dependent probes may excite charge density modes. Conversely, spin independent probes may excite spin density modes. This coupling has been experimentally detected in QD’s using Raman spectroscopy, and has been theoretically addressed within TDLSDFT. For an antidot the effect is more marked, since $`\xi `$ reaches $`1/\nu `$ instead of a smaller effective value as it happens in QD. This can be seen in Fig. 4.
It is worthwhile to compare antidot with dot results. For this sake, we show in Fig. 5 the dipole charge density response of a QD made of $`N=210`$ electrons confined by the potential created by a jellium disk of radius $`R16.7a_0^{}`$. Outside the edge region, this system has a surface density of $`0.24(a_0^{})^2`$, similar to our case-of study antidot, and it has been thoroughly studied in Ref. . For this dot the $`1\nu 10`$ range corresponds to $`10.3B1.03`$ T.
The energies of the more intense CDE’s of the antidot and of the $`N=210`$ QD have been drawn as a function of $`B`$ in Figs. 6 and 7, respectively (a more complete version of Fig. 7 can be found in Ref. ). Also drawn are the cyclotron frequency $`\omega _c`$, and $`2\omega _c`$ and $`3\omega _c`$ as well. These figures show that the interaction of the bulk magnetoplasmon with the harmonics of the cyclotron resonance $`n\omega _c`$, resembling the Berstein modes, which causes the splitting of the magnetoplasmon near $`2\omega _c`$ and $`3\omega _c`$, is well reproduced by the calculations. Alongside the groups of peaks corresponding to CDE’s we have indicated the change in radial quantum number $`n`$ of the sp wave functions involved in the transition. This change $`\mathrm{\Delta }n`$ is unambiguosly identified in TDLSDFT calculations introducing energy cutoffs in the electron-hole energy denominators $`\mathrm{\Delta }ϵ`$ entering the definition of the free correlation function $`\chi _{\sigma \sigma ^{}}^{(0)}`$: limiting the electron-hole pairs to those having $`\mathrm{\Delta }ϵ<\omega _c`$, only $`\mathrm{\Delta }n=0`$ peaks apear in the response, whereas limiting them to those with $`\mathrm{\Delta }ϵ<2\omega _c`$, no $`\mathrm{\Delta }n=2`$ peaks appear. These cyclotron-like $`\mathrm{\Delta }n=2`$ peaks have been experimentally detected.
The correlations introduced by TDLSDFT have a dramatic effect on the FIR response of the system which might not always be properly recognized. Not only they appreciably shift the CDE energies with respect to the SPE’s but also affect their intensity, causing that peaks having a sizeable strength in the SPE response are nearly washed out, whereas others hardly perceived in the SPE response adquire a large strength. To illustrate it we show in Fig. 8 the single particle and charge density strength functions for $`\nu =7`$ in a logarithmic scale. It can be noticed that in the SPE strength the $`2\omega _c`$ peak, which has an intensity six orders of magnitude smaller than the $`\omega _c`$ peak and cannot be seen at the scale of Fig. 3, emerges as a very collective peak in the CDE strength, of intensity similar to that of the bulk $`\mathrm{\Delta }n=1`$ peak. Similarly, the weak edge SPE’s loose their dimy intensity, transferring it to the very collective $`\mathrm{\Delta }n=0`$ edge CDE. Of course, in actual TDLSDFT calculations all these modes are reciprocally influenced, but the gross effect is as described.
TDLSDFT does reproduce the oscillation of the edge magnetoplasmon energy as a function of filling factor. Figure 9 shows the edge magnetoplasmon energy as a function of $`\nu `$ for the antidot and the $`N=210`$ dot. The oscillations and the out-of-phase effect are tiny effects better seen plotting the second differences of the edge magnetoplasmon energy:
$$\mathrm{\Delta }_2\omega (\nu )=\omega (\nu +1)2\omega (\nu )+\omega (\nu 1).$$
(11)
The energy oscillations are clearly seen in the Fig. 9 insert, as well as the experimental finding that maxima in the edge magnetoplasmon energies correspond to half-filled Landau levels in antidots, and to filled Landau levels in dots.
¿From Figs. 3 and 6 we conclude that for an isolate antidot, TDLSDFT is unable to produce the anticrossing of the edge and the cyclotron modes at small $`B`$ values which is observed in short-period antidot arrays. Yet, the behavior of the antidot and dot $`\mathrm{\Delta }n=0`$ edge branch near the crossing is quite distinct, manifesting a different curvature. The low-$`B`$ behavior of the edge brach experimentally found in short-period antidot arrays is likely influenced by neighboring antidots. Notice that $`1.9a_0^{}`$ at $`\nu =6`$, which is roughly half the effective surface width of the antidot lattice period of Ref. .
Figures 3 and 6 also show that the TDLSDFT antidot response is very fragmented at low $`B`$. This renders tricky to distinguish bulk from edge charge modes. Notice that at $`\nu =10`$ the high energy peak in Fig. 3 is a $`\mathrm{\Delta }n=2`$ mode, whereas the low energy peak is a mixture of $`\mathrm{\Delta }n=0`$ and $`\mathrm{\Delta }n=1`$ peaks. The situation worsens increasing $`\nu `$, which is beyond our capabilities. For instance, to have results at $`B0.5`$ T one would need to carry out calculations for $`M20`$.
A possible way to partially disentangle these modes is to study the charge density response to the $`f(r)=1/r`$ operator. Its $`r`$ dependence makes this very ineffective in exciting cyclotron-like modes while it is well suited to describe edge oscillations. This can be seeing in Fig. 10, where we have plotted the CDE’s caused by $`r`$ and $`1/r`$ for $`\nu =10`$ to 7. In the bottom panel, the intensity of the $`1/r`$ peaks has been multiply by a factor of 50,000 to allow for a sensible comparison with the $`r`$ peaks. It can then be seen in that panel that both operators excite the same $`\mathrm{\Delta }L_z=+1`$ peaks, but with much different intensity, and that indeed, the low energy peak clearly visible in the response is an edge peak.
Finally, we have plotted in the top panel of that figure the CDE caused by the $`e^{ı\theta }`$ component of the $`F`$ operators in Eq. 7. These are anticyclotron-like excitations of very weak intensity: to have them at the same scale as in the bottom panel, one should divide the scale of the $`r`$ modes by a factor of 40, and the scale of the $`1/r`$ modes by a factor of 4, i.e., the $`1/r`$ operator excites the cyclotron-like and anticyclotron-like main peaks with roughly equal intensity. In agreement with the experiment (see especially Fig. 1 of Ref. ), it can be seen on the one hand that their intensity quickly decrease as $`B`$ increases, and on the other hand that their energies are very similar to these of the cyclotron-like modes. This makes these modes experimentaly accessible only using circular polarized radiation.
## IV acknowledgments
It is a pleasure to thank Vidar Gudmundsson for useful discussions. This work has been performed under grants PB98-1247 and PB98-0124 from DGESIC, and SGR98-00011 from Generalitat of Catalunya. A.E. acknowledges support from the Dirección General de Enseñanza Superior (Spain). |
warning/0001/astro-ph0001255.html | ar5iv | text | # METRIC PERTURBATIONS AND INFLATIONARY PHASE TRANSITIONS
## 1 Field Evolution
We work in a spatially flat Friedmann-Robertson-Walker universe with scale factor $`a(t)`$ and take the inflaton to be a real scalar field with Lagrangian
$$L=\frac{1}{2}^\mu \mathrm{\Phi }_\mu \mathrm{\Phi }\left[\frac{3\mu ^4}{2\lambda }\frac{1}{2}\mu ^2\mathrm{\Phi }^2+\frac{\lambda }{4!}\mathrm{\Phi }^4\right].$$
(1)
We break up the field $`\mathrm{\Phi }`$ into its expectation value, defined within the closed time path formalism, and fluctuations about that value:
$$\mathrm{\Phi }(\stackrel{}{x},t)=\varphi (t)+\psi (\stackrel{}{x},t),\varphi (t)\mathrm{\Phi }(\stackrel{}{x},t).$$
By imposing a Hartree resummation, we arrive at the following equations of motion for the inflaton:
$$\ddot{\varphi }+3\frac{\dot{a}}{a}\dot{\varphi }\mu ^2\varphi +\frac{\lambda }{6}\varphi ^3+\frac{\lambda }{2}\psi ^2\varphi =0,$$
(2)
$$\left[\frac{d^2}{dt^2}+3\frac{\dot{a}}{a}\frac{d}{dt}+\frac{k^2}{a^2}\mu ^2+\frac{\lambda }{2}\psi ^2+\frac{\lambda }{2}\varphi ^2\right]f_k=0.$$
(3)
The fluctuation $`\psi ^2`$ is determined from the mode functions $`f_k`$:
$$\psi ^2=\frac{d^3k}{(2\pi )^3}|f_k|^2.$$
(4)
For $`a(t_0)=1`$, the initial conditions on the mode functions are
$$f_k(t_0)=\frac{1}{\sqrt{2\omega _k}},\dot{f}_k(t_0)=\left(\dot{a}(t_0)i\omega _k\right)f_k(t_0),$$
(5)
with
$$\omega _k^2k^2\mu ^2+\frac{\lambda }{2}\psi ^2+\frac{\lambda }{2}\varphi ^2\frac{R(t_0)}{6}.$$
$`R(t_0)`$ is the initial Ricci scalar. For small $`k`$, we modify $`\omega _k`$ either by means of a quench or by explicit deformation so that the frequecies are real.
The gravitational dynamics are determined by the semi-classical Einstein equation. For a minimally coupled inflaton we have
$`{\displaystyle \frac{\dot{a}^2}{a^2}}`$ $`=`$ $`{\displaystyle \frac{8\pi G_N}{3}}[{\displaystyle \frac{1}{2}}\dot{\varphi }^2+{\displaystyle \frac{1}{2}}\dot{\psi }^2+{\displaystyle \frac{1}{2a^2}}(\stackrel{}{}\psi )^2`$ (6)
$`+`$ $`{\displaystyle \frac{3\mu ^4}{2\lambda }}{\displaystyle \frac{1}{2}}\mu ^2(\varphi ^2+\psi ^2)+{\displaystyle \frac{\lambda }{24}}(\varphi ^4+3\psi ^2^2+6\varphi ^2\psi ^2)],`$
where $`G_N`$ is Newton’s gravitational constant, and
$`\dot{\psi }^2(t)`$ $``$ $`{\displaystyle \frac{d^3k}{(2\pi )^3}|\dot{f}_k|^2},`$ (7)
$`\left(\stackrel{}{}\psi (t)\right)^2`$ $``$ $`{\displaystyle \frac{d^3k}{(2\pi )^3}k^2|f_k|^2}.`$ (8)
Each of these integrals is regulated using a cutoff with the divergences absorbed into a renormalization of the parameters of the theory.
A typical field evolution is depicted in Fig. 1.
## 2 Metric Perturbations
Following the procedure of Mukhanov, Feldman and Brandenberger, we arrive at the expression for the density contrast at mode re-entry:
$`|\delta _h(k)|`$ $`=`$ $`\sqrt{{\displaystyle \frac{2}{75\pi }}}{\displaystyle \frac{1}{M_{Pl}\left(\dot{\varphi }^2+\dot{\psi }^2\right)}}[{\displaystyle \frac{3\mu ^4}{2\lambda }}{\displaystyle \frac{1}{2}}\mu ^2(\varphi ^2+\psi ^2)`$ (9)
$`+{\displaystyle \frac{\lambda }{24}}(\varphi ^4+6\varphi ^2\psi ^2+3\psi ^2^2)]^{1/2}`$
$`\times `$ $`[\mu ^4(\varphi ^2+\psi ^2){\displaystyle \frac{\lambda \mu ^2}{3}}(\varphi ^4+6\varphi ^2\psi ^2+3\psi ^2^2)`$
$`+{\displaystyle \frac{\lambda ^2}{36}}(\varphi ^6+15\varphi ^4\psi ^2+45\varphi ^2\psi ^2^2+15\psi ^2^3)]^{1/2}.`$
The computation of the tilt parameter $`n_s1`$ is straightforward, given (9):
$$n_s1\frac{d(\mathrm{ln}|\delta _h(k)|)}{d\mathrm{ln}(k)}|_{k=aH}.$$
(10)
As gravitational wave perturbations do not directly interact with the inflaton field, they may be related directly to the expansion rate. The amplitude of gravitational waves is simply:
$$|\delta _g(k)|=\frac{2}{\sqrt{3\pi }}\frac{H}{M_{Pl}},$$
(11)
All expressions are to be evaluated when the given scale $`k`$ first crosses the horizon, $`k=aH`$.
An example perturbation spectrum is shown in Fig. 2, while Fig. 3 shows the dependence of the spectrum on the initial state for a number of possible evolutions. Both of these figures show distinct regions characterized by a blue spectral tilt.
## Acknowledgments
R.H. was supported in part by the Department of Energy Contract DE-FG02-91-ER40682.
## References |
warning/0001/cs0001001.html | ar5iv | text | # Von Neumann Quantum Logic vs. Classical von Neumann Architecture?
Conference
’2000
> Abstract. The name of John von Neumann is common both in quantum mechanics and computer science. Are they really two absolutely unconnected areas? Many works devoted to quantum computations and communications are serious argument to suggest about existence of such a relation, but it is impossible to touch the new and active theme in a short review. In the paper are described the structures and models of linear algebra and just due to their generality it is possible to use universal description of very different areas as quantum mechanics and theory of Bayesian image analysis, associative memory, neural networks, fuzzy logic.
J. von Neumann is considered as one of “fathers” of modern computers due to his theoretical works and elaboration of computer EDVAC at 1945–50, but already 20 year before it, at 1926–31 he participated in drawing yet another kind of logic — logic of quantum mechanics and only now the different areas of research of one scientist seem going to meet and born new family of ultrascale cybernetic devices — quantum computers.
Currently the quantum information science exists mainly as theoretical area of research and size of quantum registers does not exceed of 3–5 quantum bits (qubits), but in the paper is considered question: does the quantum logic has some useful application as an abstract mathematical model in computer science? Such applications of physical models nowadays are not unusual, for example Boltzmann machines and other methods came to area of artificial neural networks from statistical physics .
It is also useful to draw some analogy with fuzzy sets and logic . Here is used discrete representation of 2D sets on some lattice due to understanding analogy with bit-map pictures.
Let us introduce few data types in some “Pascal-like” notation to come from fuzzy sets to quantum registers. Here $`n`$ and $`N=2^n`$ are integer numbers.
1) $`x`$ : dot = \[1 .. N, 1 .. N\];
Here $`x`$ is represented as point on 2D space for simplicity, because of digitization we can consider space of any dimension as interval of natural numbers $`x`$ : 1 .. $`N^2`$. In this example $`x`$ is visual model for $`n+n=2n`$-bits register.
2) $`I`$ : set\_2D = set of dot;
or
2) $`I`$ : set\_2D = array \[dot\] of Boolean;
The set, “image” can be considered as black and white picture — black set on white sheet. It is necessary $`2^{NN}`$ i.e. $`2(2(2n))`$ bits for the set.
3) $`F`$ : fuzzy\_set = array \[dot\] of real;
The real is interval \[0.0 .. 1.0\] of real numbers represented with some finite precision $`d`$. It can be considered as an analogue of gray-scaled picture. The fuzzy set also can be standardized by condition $`_xF\text{[}x\text{]}=1`$ where $`_x`$ in example under consideration is double sum:
$$\underset{i,j=1}{\overset{N}{}}F_{ij}=1$$
(1)
(in continuous case an integration used instead of the summation and distributions $`F(x)`$ are used instead of arrays).
The condition (1) is used in statistical interpretation of fuzzy set, when it is suggested to choose some point of set with probability proportional of adequacy function $`F\text{[}x\text{]}`$, and sum of all the probabilities is unit. Let us use notation $`p\text{[}x\text{]}`$ for the probabilities.
Now let us introduce notion of quantum set (it is not standard term, mathematical object discussed further corresponds to $`2n`$-qubit register in quantum information science or quantum mechanical system with $`N^2`$ states ):
4) $`Q`$ : qu\_set = array \[dot\] of complex;
Where complex is $`q=u+iw`$, $`|q|1`$. Instead of standardization condition here is used normalization: $`_x|Q\text{[}x\text{]}|^2=1`$ i.e.:
$$\underset{i,j=1}{\overset{N}{}}\left|Q_{ij}\right|^2=1$$
(2)
for example under consideration.
In quantum mechanics the complex numbers $`q\text{[}x\text{]}`$ is called amplitudes, related with classical probabilities as $`p=|q|^2=u^2+w^2`$, i.e. there is standardized fuzzy set related with given normalized quantum set via formula: $`p\text{[}x\text{]}=|q\text{[}x\text{]}|^2`$.
It is possible for simplification to consider case with real amplitudes ($`w=0`$), and let us explain why an “auxiliary” fuzzy set with $`q\text{[}x\text{]}=\sqrt{p\text{[}x\text{]}}`$ has some independent useful application.
Let us consider two sets $`p_1\text{[}x\text{]}`$, $`p_2\text{[}x\text{]}`$ and look for some likelihood function $`H(p_1,p_2)`$ with properties: $`H(p_1,p_2)<1`$ for $`p_1p_2`$, $`H(p,p)=1`$ . For standardized sets such function can be chosen as $`H(p_1,p_2)=_x(p_1\text{[}x\text{]}p_2\text{[}x\text{]})^{1/2}`$. If we use quantum sets $`q_1`$, $`q_2`$ ($`p_1=q_1^2`$, $`p_2=q_2^2`$), the formula is $`H(q_1,q_2)=_xq_1\text{[}x\text{]}q_2\text{[}x\text{]}`$.
In our example $`x`$ was chosen as multi-index of 2D array mostly for simple visualization and any array can be described as one-dimensional. Here $`x=(i,j)`$ can be substituted by index of the 1D array like $`K=(i1)N+j`$, $`K[1..N^2]`$, then formula can be written as:
$$H(q,q^{})=\underset{i,j=1}{\overset{N}{}}q_{ij}q_{ij}^{}\underset{K=1}{\overset{N^2}{}}q_K^{}q_K^{}$$
(3)
and it shows, that $`H(q,q^{})`$ is simply scalar product of two vectors with $`N^2`$ elements and unit lengths (due to normalization condition) and so the function really can be unit only for equivalent vectors.
It should be mentioned also, that the formula (3) is given for real $`q_K^{}`$ for simplicity and in complex case it contains terms with complex conjugation: $`q_K^{}\overline{q}_K^{}`$. It is Hermitian norm . Such case also has applications, for example then we work with complex Fourier transform of some image.
The discussed property of quantum set as “square root of fuzzy set” makes clear, why it can be useful in such abstract area, as image recognition , quite far from initial appearance in quantum mechanics.
Let us discuss now some question related with “hardware”. We had few data structures:
1) $`x`$ : dot, 2) $`I\text{[}x\text{]}`$ : set\_2D, 3) $`p\text{[}x\text{]}`$ : fuzzy\_set
It is sequence with more and more complicated structure with occupation of more and more computer memory. But let us suggest, that register $`x`$ is permanently changing its value by such a law, that after enough period of time $`T`$ it can be found, the $`x`$ had value $`v`$ during time $`t\text{[}v\text{]}`$ and $`\underset{T\mathrm{}}{lim}t\text{[}v\text{]}/T=p\text{[}v\text{]}`$.
The algorithm can be implemented by software, but it also can be considered as some hardware register (like implementation of random number generator in some computers with main difference, that $`p\text{[}v\text{]}`$ depends on $`v`$; or input port of some analog-to-digital converter is scanning some “physical model” of fuzzy set $`p`$).
Then each access to the register produces some value of $`x`$ with probability $`p\text{[}x\text{]}`$, so one $`2n`$-bits stochastic register is enough to implement statistical model of standardized fuzzy set discussed earlier.
Now let us come to qu\_set data type. Why it can be modeled by one quantum register? The example with stochastic register as model of fuzzy set is some analogy (see also ). But procedure of access to such register due to laws of quantum mechanics has some differences with classical statistical register discussed below.
Let us consider some value $`q`$ qu\_set of the register, it is array of numbers $`q\text{[}x\text{]}`$, we may not read all the numbers, but if we access to the register, we read number $`x`$ with probability $`p\text{[}x\text{]}=|q\text{[}x\text{]}|^2`$ and it coincides with functionality of stochastic register described above.
It should be mentioned only, that any access to $`q`$ destroys the quantum register by substitution instead of $`q`$ new array with 1 in element with index $`x`$ and with all other is 0, and so the register should be reset (preparation in terminology of quantum mechanics) in $`q`$ after each access (quantum measurement).
But the quantum register has other useful property, it is possible instead of simple access described above to perform another operation, we prepare some given $`q^{}`$ qu\_set and read the register $`q`$ with using $`q^{}`$ as some “quantum bit-mask”, then with probability $`|H(q,q^{})|^2`$ (see Eq. 3) operation is successful and so by repeating it more times we may found $`|H|`$ with more precision.
Because the $`|H|`$ has useful application as likelihood function, the quantum register can be used as some hardware accelerator for image analysis. Currently such hardware is not accessible and so it was interesting to research advantages and disadvantages of the particular function $`|H|`$ in usual software applications.
It is promising not only because such kind of software would suffer giant speed-up after creation specific quantum hardware, but also because the used mathematical constructions and methods of linear algebra are quite convenient and powerful.
It should be mentioned, that similar mathematical methods already was used in models of associative memory and formal neural network without any relation with quantum mechanical models. Only noticeable difference was using real linear spaces instead of complex and Euclidean norm (Eq. 3) instead of Hermitian (with complex conjugation).
Let us now consider some operations with fuzzy and quantum sets, discuss fuzzy and quantum logic.
For usual sets we have basic operations for $`A`$$`B`$: set\_2D, — intersection: $`AB`$, union: $`AB`$, complementation: $`A^c`$. With using presentation of set as Boolean array the operations in components can be written: $`(a^c)\text{[}x\text{]}=\text{not}a\text{[}x\text{]}`$, $`(ab)\text{[}x\text{]}=a\text{[}x\text{]}b\text{[}x\text{]}`$, $`(ab)\text{[}x\text{]}=a\text{[}x\text{]}b\text{[}x\text{]}`$.
Similarly it is possible to define operations with fuzzy sets by definition of real analogs of Boolean operations not ($`\neg `$), and ($``$), or ($``$).
For example $`\neg a1a`$, $`ab\mathrm{min}(a,b)`$, $`ab\mathrm{max}(a,b)`$ is a good choice, but here is used a second one, more algebraic $`abab`$, $`aba+bab=\neg ((\neg a)(\neg b))`$.
But qu\_set introduced above is not directly used in quantum logic — the linear operators are used here instead of vectors:
$`L`$ : qu\_map = array \[dot,dot\] of complex;
It is matrix for linear map: qu\_set $``$ qu\_set:
$$q_{ij}^{}=\underset{k,l=1}{\overset{N}{}}L_{ij,kl}^{}q_{kl}^{}\text{or}q_I^{}=\underset{K=1}{\overset{N^2}{}}L_{IK}^{}q_K^{}$$
(4)
where indexes $`I`$, $`K`$ are used instead of multi-indexes \[$`i,j`$\] and \[$`k,l`$\] (let us for simplicity use further the indexes like $`I,K:[1..M]`$, $`M=N^2`$).
The operators $`A,B,\mathrm{}`$ qu\_map form an algebra with usual matrix multiplication C=AB:
$$C_{KJ}=\underset{I=1}{\overset{M}{}}A_{KI}B_{IJ}$$
(5)
A special kind of operators, projectors, make possible comparison of the algebra with logic, i.e. Boolean algebra. The projector is operator with property $`P^2=P`$. Let us consider set of orthogonal projectors, i.e. $`P_iP_j=0`$, $`ij`$, then the operators produce Boolean algebra in respect of operations:
$$\neg P1P,PRPR,PRP+RPR$$
(6)
Elements of the algebra have form $`P_{(S)}`$ there $`S`$ : set of 1 .. $`M`$:
$$P_{(S)}=\underset{I(S)}{}P_I$$
(7)
There is relation between $`q`$ qu\_set and some projector $`P_q`$qu\_map: $`P_q\text{[}I,J\text{]}=q\text{[}I\text{]}q\text{[}J\text{]}`$. To describe properties of the projector it is convenient together with $`q`$ considered as row with $`M`$ elements to consider transposed column $`q^+`$ (conjugated for complex case).
Then $`P_q=q^+q`$ and $`qq^+=|H(q,q)|^2=1`$ (the row $`q`$ can be considered as $`1\times M`$ matrix, column $`q^+`$ as $`M\times 1`$ matrix and due to law of multiplication the $`qq^+`$ is $`1\times 1`$ matrix i.e. number and $`q^+q`$ is $`M\times M`$ matrix) and $`P_qP_q=q^+qq^+q=q^+1q=q^+q=P_q`$.
If we consider family of nonintersected sets $`q_1`$, $`q_2,\mathrm{},q_k`$ then $`P_i=q_i^+q_i`$ are orthogonal projectors and so quantum sets in such representations have rather relations with usual logic than with fuzzy one.
For more clear explanation of properties of $`P_i`$ it is possible to use existence of some orthogonal (unitary for complex case) matrix $`U`$ same for all $`P_i`$ such, that all $`P_i^{}=UP_iU^1`$ are diagonal and have very simple form: $`P_1^{}=\text{diag}(1,0,`$$`,0)`$, $`P_2^{}=\text{diag}(0,1,\mathrm{},0)`$, etc..
The classical Boolean structure of the operators $`P_i`$ and their sums $`P_{(S)}`$ (see Eq. 7) is because of all the operators commute . If to choose projectors $`P`$, $`R`$: $`PRRP`$ the Eq. 6 do not produce Boolean algebra, but it is other kind of non-Boolean logic, than fuzzy one.
It should be mentioned, that more direct relation with fuzzy set have so-called mixed quantum states $`R=_iw_iP_i`$, $`_iw_i=1`$ where $`w_i`$ have statistical nature and so here is written analog of standardized fuzzy set.
A representation of some kind of fuzzy operations is example with family of commuting operators , but not projectors. They are described by diagonal matrices: $`D_q\text{[}I,I\text{]}=q\text{[}I\text{]}`$, $`D_q\text{[}I,J\text{]}=0`$, $`IJ`$ and already shown Eq. 6.
Bibliographical notes: In addition to references some general handbook on quantum mechanics like is appropriate for most text of the paper. New area of quantum computation is presented in \[6, 14–17\]. Two works of J. von Neumann devoted to quantum logic and electronic computers are included for completeness. |
warning/0001/math0001156.html | ar5iv | text | # New Solutions of the Einstein-Dirac Equation in Dimension 𝑛=3. 11footnote 1Supported by the SFB 288 and the Graduiertenkolleg ”Geometrie und Nichtlineare Analysis” of the DFG.
Consider a Riemannian spin manifold of dimension $`n3`$ and denote by $`D`$ the Dirac operator acting on spinor fields. A solution of the Einstein-Dirac equation is a spinor field $`\psi `$ solving the equations
$$Ric\frac{1}{2}Sg=\pm \frac{1}{4}T_\psi ,D(\psi )=\lambda \psi .$$
Here $`S`$ denotes the scalar curvature of the space, $`\lambda `$ is a real constant and $`T_\psi `$ is the energy-momentum tensor of the spinor field $`\psi `$ defined by the formula
$$T_\psi (X,Y)=(X_Y\psi +Y_X\psi ,\psi ).$$
Any weak Killing spinor $`\psi ^{}`$ (WK-spinor)
$$_X\psi ^{}=\frac{n}{2(n1)}dS(X)\psi ^{}+\frac{2\lambda }{(n2)S}Ric(X)\psi ^{}\frac{\lambda }{n2}X\psi ^{}+\frac{1}{2(n1)S}XdS\psi ^{}$$
yields a solution $`\psi `$ of the Einstein-Dirac equation after normalization
$$\psi =\sqrt{\frac{(n2)|S|}{|\lambda ||\psi ^{}|^2}}\psi ^{}.$$
In fact, in dimension $`n=3`$ the Einstein-Dirac equation is essentially equivalent to the weak Killing equation (see \[KimF\]). Up to now the following 3-dimensional Riemannian manifolds admitting WK-spinors are known:
1. the flat torus $`T^3`$ with a parallel spinor;
2. the sphere $`S^3`$ with a Killing spinor;
3. two non-Einstein Sasakian metrics on the sphere $`S^3`$ admitting WK-spinors. The scalar curvature of these two left-invariant metrics equals $`S=1\pm \sqrt{5}`$.
The aim of this short note is to announce the existence of a one-parameter family of left-invariant metrics on $`S^3`$ admitting WK-spinors. This family contains the two non-Einstein Sasakian metrics with WK-spinors on $`S^3`$, but does not contain the standard sphere $`S^3`$ with Killing spinors. Moreover, any simply-connected, complete Riemannian manifold $`X^3S^3`$ with WK-spinors such that the eigenvalues of the Ricci tensor are constant is isometric to a space of this one-parameter family.
In order to formulate the result precisely, we fix real parameters $`K,L,MR`$ and denote by $`X^3(K,L,M)`$ the 3-dimensional, simply-connected and oriented Riemannian manifold defined by the following structure equations:
$$\omega _{12}=K\sigma ^3,\omega _{13}=L\sigma ^2,\omega _{23}=M\sigma ^1,$$
or, equivalently:
$$d\sigma ^1=(LK)\sigma ^2\sigma ^3,d\sigma ^2=(M+K)\sigma ^1\sigma ^3,d\sigma ^3=(LM)\sigma ^1\sigma ^2.$$
The 1-forms $`\sigma ^1,\sigma ^2,\sigma ^3`$ are the dual forms of an orthonormal frame of vector fields. Using this frame the Ricci tensor of $`X^3(K,L,M)`$ is given by the matrix
$$Ric=\left(\begin{array}{ccc}2KL& 0& 0\\ 0& 2KM& 0\\ 0& 0& 2LM\end{array}\right).$$
Theorem: Let $`X^3S^3`$ be a complete, simply-connected Riemannian manifold such that:
* the eigenvalues of the Ricci tensor are constant;
* the scalar curvature $`S0`$ does not vanish.
If $`X^3`$ admits a WK-spinor, then $`X^3`$ is isometric to $`X^3(K,L,M)`$ and the parameters are a solution of the equation
$$K^2L(LM)^2M+L^3M^3+KL^2M^2(ML)+K^3(LM)(L+M)^2=0()$$
Conversely, any space $`X^3(K,L,M)`$ such that $`(K,L,M)(0,0,0)`$ is a solution of $`()`$ admits two WK-spinors for one and only one WK-number $`\lambda `$. With respect to the fixed orientation of $`X^3(K,L,M)`$ we have the two cases:
$$\lambda =+\frac{S}{2\sqrt{2}}\sqrt{\frac{S}{S^2|Ric|^2}}\text{if }K<M$$
$$\lambda =\frac{S}{2\sqrt{2}}\sqrt{\frac{S}{S^2|Ric|^2}}\text{if }M<K.$$
The spaces $`X^3(K,L,M)`$ are isometric to $`S^3`$ equipped with a left-invariant metric.
Remark: If the parameters $`K=M`$ coincide, the solution of the equation $`()`$ is given by
$$L=\frac{1}{4}K(1\sqrt{5}),L=\frac{1}{4}K(1+\sqrt{5})$$
and we obtain the Ricci tensors
$$Ric=\left(\begin{array}{ccc}\frac{1}{2}K^2(\sqrt{5}1)& 0& 0\\ 0& 2K^2& 0\\ 0& 0& \frac{1}{2}K^2(\sqrt{5}1)\end{array}\right)$$
or
$$Ric=\left(\begin{array}{ccc}\frac{1}{2}K^2(1+\sqrt{5})& 0& 0\\ 0& 2K^2& 0\\ 0& 0& \frac{1}{2}K^2(1+\sqrt{5})\end{array}\right).$$
The non-Einstein-Sasakian metrics on $`S^3`$ occur for the parameter $`K=1`$ (see \[KimF\]).
Remark: Using the standard basis of the Lie algebra $`\text{so}(3)`$ we can write the left-invariant metric of the space $`X^3(K,L,M)`$ in the following way:
$$\left(\begin{array}{ccc}\frac{1}{|ML||K+M|}& 0& 0\\ & & \\ 0& \frac{1}{|KL||ML|}& 0\\ & & \\ 0& 0& \frac{1}{|KL||K+M|}\end{array}\right).$$
The equation $`()`$ is a homogeneous equation of order six. The transformation $`(K,L,M)(\mu K,\mu L,\mu M)`$ corresponds to a homothety of the metric. Therefore - up to a homothety - the moduli space of solutions is a subset of the real projective space $`P^2(R)`$ given by the equation $`()`$. This subset is a configuration of six curves in $`P^2(R)`$ connecting the three points $`[K:L:M]=[1:0:0],[0:1:0],[0:0:1]`$, corresponding to flat metrics.
In particular, we have constructed two paths of solutions of the Einstein-Dirac equation deforming the non-Einstein Sasakian metrics on $`S^3`$.
The proof of the Theorem as well as the complete computations will be published in a furthercoming paper of the author (see \[F\]). |
warning/0001/hep-ph0001187.html | ar5iv | text | # Polarization of Inclusive Λ_𝑐’s in a Hybrid Model
## 1 INTRODUCTION
Gary R. Goldstein This work is supported in part by funds provided by the U.S. Department of Energy (D.O.E.) #DE-FG02-92ER40702.
Department of Physics
Tufts University
Medford, MA 02155 USA
Inclusively produced strange hyperons can have sizeable polarization over a wide range of energies. Evidence now indicates that charmed hyperons also have sizeable polarization . Many theoretical models have been proposed to explain various aspects of hyperon polarization data with varying success . All try to explain the large negative $`\mathrm{\Lambda }`$ polarization. Because the hyperon data is in the region of high CM energy but relatively small transverse momentum ($`p_T1`$ GeV/c), soft QCD effects should play a major role in any theoretical explanation. Several years ago Dharmaratna and Goldstein developed a hybrid model for $`\mathrm{\Lambda }`$ polarization in inclusive reactions . The model involves hard scattering at the parton level, gluon fusion and light quark pair annihilation, to produce a polarized heavy quark which then undergoes a soft recombination that, in turn, enhances the polarization of the hyperon. This scheme provided an explanation for the characteristic kinematic dependences of the polarization in $`p+p\mathrm{\Lambda }+X`$. The use of perturbative QCD to produce the initial polarization for strange quarks, with their low current or constituent quark mass (compared to $`\mathrm{\Lambda }_{QCD}`$) made the application of perturbation theory marginal, however.
In the heavy quark realm the perturbative contribution is more reliable. Given these circumstances, I have modified the original hybrid model to apply to heavy flavor baryons produced inclusively from either proton or pion beams. The results are encouraging, as the following will show (see ref. for a more complete treatment).
## 2 HYBRID MODEL
All of the models for $`\mathrm{\Lambda }`$ polarization begin with the observation that $`Q`$-flavor hyperons of the type $`\mathrm{\Lambda }_Q[ud]Q`$ have their polarization carried primarily by the $`Q`$; the \[ud\] must be a color anti-triplet isospin 0 spin scalar diquark (to the extent that $`gluons+𝐋+sea`$ contributions can be ignored). How does the $`Q`$ itself get polarized in a production process? Consider $`parton+partonQ_{}+\overline{Q}`$. At tree level in QCD, there can be no single quark polarization for these two-body subprocesses, all diagrams being relatively real. This can be seen when the polarization is written in terms of helicity amplitudes $`f_{a,b,c,d}`$ for particles $`A+BC+D`$ as
$`𝒫_Q`$ $``$ $`{\displaystyle \underset{a,b,d}{}}f_{a,b;c,d}^{}f_{a,b;c^{},d}(\sigma \widehat{𝐧})_{c,c^{}}`$ (1)
$``$ $`Im{\displaystyle f_{a,b;+,d}f_{a,b;,d}^{}},`$
where $`\widehat{𝐧}`$ is the normal to the scattering plane. Hence there has to be a phase difference and a flip–non-flip interference. In QCD with zero quark masses there are only non-flip vertices; helicity flip requires non-zero quark masses. And a relative imaginary part only arises beyond tree level . So the hybrid model incorporates the order $`\alpha _s^2`$ QCD perturbative calculation of interference between tree level and the large number of one loop diagrams to produce massive heavy quark polarization. (Only the imaginary parts of the one loop diagrams were needed, so the Cutkosky rules were used to simplifiy the calculation. For the lengthy results see ref. as well as an independent calculation in ref. .) This gives rise to significant polarization , proportional to $`\alpha _s(Q^2)`$ and to complicated functions of the constituent quark mass. The scale here is $`Q^2m_Q^2\lambda _{QCD}^2`$. The results are illustrated in fig. 1 for the $`g+gQ_{}+\overline{Q}`$ case, with CM energy 26 GeV and outgoing quark flavors $`Q=d,s,c,b`$. The symmetry requires $`𝒫(\pi \theta )=𝒫(\theta )`$, so backward $`Q`$ has $`𝒫<0`$. The magnitude of $`𝒫`$ reaches $`6\%`$ for the b-quark. It is clear that the polarization increases roughly as the quark mass. Similar results are obtained for $`q+\overline{q}Q_{}+\overline{Q}`$.
The cross sections for polarized $`Q`$-quarks (polarized normal to the production plane) must then be convoluted with the relevent structure functions for the hadronic beam and target. The inclusive cross section for $`hadron+hadronQ(\mathrm{o}r)+X`$ is obtained thereby. For protons on protons gluon fusion is the more significant subprocess.
The hadronization process, by which the polarized $`Q`$ recombines with a \[ud\] diquark system to form a $`\mathrm{\Lambda }_Q`$, is crucial for understanding the subsequent hadron polarization. The backward moving, negatively polarized heavy quark must be accelerated to recombine with a fast moving diquark (resulting from remnants of the $`pp`$ or $`\pi p`$ collision) to form the hadron with particular $`x_F`$ while preserving the quark’s $`p_T`$ value. Letting $`x_Q`$ be the Feynman $`x`$ for the heavy quark, the simple form, a linear mapping of the $`Q`$ kinematic region,
$$x_F=a+bx_Q$$
(2)
is used for the recombination. Naively, if the Q has 1/3 of the final hyperon momentum (in its infinite momentum frame) and the diquark carries 2/3 of that momentum, then $`a=2/3`$ and $`b=1`$. The values actually used, $`a=0.86`$ and $`b=0.70`$, were chosen to fit the $`pp\mathrm{\Lambda }+X`$ data (that existed in 1990) at one $`x_F`$ value. These parameters in eqn. 2 are not far from the naive expectation.
This recombination prescription is similar to the semi-classical dynamical mechanism used in the “Thomas precession” model of hyperon polarization , which posits that the s-quark needs to be accelerated by a confining potential or via a “flux tube” at an angle to its initial momentum in order to join with the diquark to form the hyperon. The skewed acceleration gives rise to a spin precession for the s-quark. With the precession rate, $`\omega _𝐓=(\gamma 1)𝐯\times 𝐚/v^2𝐩_𝐐\times 𝚫𝐩_𝐋\widehat{𝐧}`$, an energy shift $`𝐒\omega _𝐓+𝐒\widehat{𝐧}`$ occurs. Hence negative values of $`𝐒\widehat{𝐧}`$ are energetically favored. In the Hybrid Model the $`Q`$ has acquired negative polarization already from the hard subprocess before it is accelerated in the hadronic recombination process. That “seed” polarization gets enhanced by a multiplicative factor $`A2\pi `$ which simulates the Thomas precession. The Hybrid Model combines hard perturbative QCD with this simple model for non-perturbative recombination.
In summary, the hyperon polarization is given as
$$𝒫_{\mathrm{\Lambda }_Q}(x_F,p_T)=A𝒫_Q(x_Q(x_F),p_T)$$
(3)
for each reaction $`g(x_1)+g(x_2)`$ or $`q(x_1)+\overline{q}(x_2)Q\overline{Q}`$, with the mapping function $`x_Q(x_F)`$ obtained by inverting eqn. 2. From eqn. 3 the subprocess polarized cross sections,
$$\frac{d^2\sigma (\mathrm{a}nd)_{i,j}}{dx_Qdp_T}$$
for partons $`(i,j)`$ at $`(x_1,x_2)`$ can be obtained. These cross sections are convoluted with the gluon, quark and antiquark structure functions for the proton and pion , $`g^{p,\pi }(x),q^{p,\pi }(x),\overline{q}^{p,\pi }(x)`$, or generically $`f_i^{p,\pi }(x)`$ leading to
$`{\displaystyle \frac{d^2\sigma (\mathrm{a}nd)}{dx_Qdp_T}}`$ $`=`$ $`{\displaystyle \underset{i,j}{}}{\displaystyle _0^1}𝑑x_1{\displaystyle _0^1}𝑑x_2f_i^{p,\pi }(x_1)`$ (4)
$`f_j^p(x_2){\displaystyle \frac{d^2\sigma (\mathrm{a}nd)_{i,j}}{dx_Qdp_T}}.`$
Next the recombination formula, eqn. 2, is applied to obtain the corresponding $`\mathrm{\Lambda }_Q`$ polarized cross section at $`x_F(=a+bx_Q)`$ and $`p_T`$. The polarization is obtained via
$$𝒫_{\mathrm{\Lambda }_Q}(x_F,p_T)=A\frac{d^2\sigma ()d^2\sigma ()}{d^2\sigma ()+d^2\sigma ()},$$
(5)
in an obvious notation.
Note that the linear form of eqn. 2 maps the $`Q`$-quark Feynman x region $`[1,(1a)/b]`$ into the $`x_F`$ region $`[(ab),+1]`$ for the $`\mathrm{\Lambda }_Q`$. The $`p+pQ`$ differential cross section, $`d^2\sigma /dx_Qdp_T`$ is mapped correspondingly into the $`p+p\mathrm{\Lambda }_Q`$ cross section $`d^2\sigma /dx_Fdp_T`$. The measured cross sections for the latter are known to fall with positive $`x_F`$ and to fall precipitously with $`p_T`$, roughly as
$$(1x_F)^\alpha e^{\beta p_T^2}$$
(6)
overall , where $`\alpha `$ and $`\beta `$ are greater than 1.0 (for $`\pi +p\mathrm{\Lambda }+X`$ $`\alpha ,\beta 3.0`$). However, the directly computed lowest order p+p$``$s-quark cross section grows with $`x_Q`$ in the region (-1,0) and it falls more gradually with $`p_T`$ than the exponential in eqn. 6. Hence the more complete recombination scheme would have to temper the $`x_F`$ dependence and narrow the $`p_T`$ distribution. This will not affect the polarization calculation, though, since the individual up or down polarized cross sections will be alterred in the same way. For a more thorough calculation this should be done, and work is underway on this point. The polarization results are the focus of this work.
## 3 COMPARISON WITH DATA AND PREDICTIONS
Applied to strange $`\mathrm{\Lambda }`$ production, the hybrid model reproduces the detailed $`(x_F,p_T)`$ dependence of the data, with very slow energy dependence , as fig. 2 shows.
Note that an estimated 20 to 30% of the $`\mathrm{\Lambda }`$’s come from $`\mathrm{\Sigma }^0\gamma \mathrm{\Lambda }`$ , so the parameter $`A`$ in eqn. 5 is increased to 7.9. The agreement of the hybrid model with the wide range of data is excellent.
It is worth remarking that recently extensive data have been collected on $`\mathrm{\Lambda }`$ polarization in many exclusive reactions , for which a simple form, $`P=(0.443\pm 0.037)x_Fp_T`$, approximates all the polarization data at $`p_{lab}=27.5`$ GeV/c. That form provides lower bracketing values for the inclusive polarization, as fig. 2 indicates. In the hybrid model all the final states other than the $`\mathrm{\Lambda }`$ arise from the hadronization of the $`\overline{s}`$-quark and the remains of the incoming baryons. Therefore, in the hybrid model it would be anticipated that as the beam energy increases and/or more final states are included in the determination of the $`\mathrm{\Lambda }`$ polarization, more complicated final states will be accompanied by much lower polarization as $`p_T`$ increases beyond 1 GeV/c.
In turning to $`\mathrm{\Lambda }_c`$ production, there is a straightforward scaling up that occurs in the $`𝒫`$ equations for $`g+g`$ and $`q+\overline{q}c+\overline{c}`$. The seed polarization increases by $`3`$. The recombination with a fixed force/mass should have the same Thomas factor, but the overall recombination could scale as $`M_{hadron}`$, so a factor of $`M_{\mathrm{\Lambda }_c}/M_\mathrm{\Lambda }2`$ could apply. The scaled polarization in the reaction $`\pi +p\mathrm{\Lambda }_c+X`$ is obtained from the convolution of eqn. 4 with the $`\pi `$ structure functions for the beam . The predicted kinematic dependences for $`𝒫(x_F,p_T)`$ are shown in fig. 3 (without the hadron mass enhancement).
Integrating over $`x_F`$ from -0.2 to +0.6 allows the comparison with the data of E791 , as fig. 4 shows. The lower curve has taken the additional factor of 2 that could apply to the scaling of the recombination. The higher curve does not have that factor and gives a poorer fit, albeit not far from the large uncertainties in the data points.
## 4 CONCLUSION
In conclusion, these results are encouraging for the hybrid model. The Thomas enhanced gluon fusion model has been modified to include quark-anti-quark annihilation, which should be more prominent for heavy baryon polarization in pion induced reactions, like the above $`\pi ^{}+p\mathrm{\Lambda }_c+X`$. Experimental data can be analyzed into $`x_F`$ as well as $`p_T`$ bins, so the predictions from the hybrid model can be checked in detail. It is important to realize that the results for the $`\mathrm{\Lambda }_c`$ were obtained without changing the parameters of the model that had been applied to the strange hyperons. Aside from the possible enhancement in $`A`$, everything else was simply scaled up by quark mass. This gives further credence to the results herein.
The somewhat ad hoc prescription for the recombination is being studied further in order to accommodate both the polarization and the cross section behavior of eqn. 6, with the kinematic variables $`x_F`$ and $`p_T`$. The overall factor $`A`$ may have some dependence on those variables as well, given that the semi-classical Thomas precession may have such dependence. Furthermore, an investigation of other hyperon production reactions, involving $`\mathrm{\Sigma },\mathrm{\Sigma }_c`$, and $`\mathrm{\Xi }`$, for example, is underway. Will $`\overline{p}+p\mathrm{\Lambda }+X`$ carry significant, near energy independent polarization at collider energies? Can photoproduction of $`\mathrm{\Lambda }`$ produce large polarizations also? These can be answered within the hybrid model.
The related strange meson asymmetries in $`p+pK\mathrm{o}r\pi \mathrm{o}r\mathrm{\Lambda }`$ will be investigated in future work as well.
## 5 ACKNOWLEDGMENTS
The author thanks Austin Napier and appreciates correspondence with members of E791, particularly M.V. Purohit, G.F. Fox and J.A. Appel. He is grateful to the organizers of Hyperon99 for inviting him and for producing a lively conference. |
warning/0001/cond-mat0001437.html | ar5iv | text | # Ferromagnetism below the Stoner limit in La-doped SrB6.
## Abstract
Spin-polarized band calculations for LaSr<sub>7</sub>B<sub>48</sub> show a weak ferro-magnetic state. This is despite a low density-of-states (DOS) and a low Stoner factor. The reason for the magnetic state is found to be associated with a gain in potential energy in addition to the exchange energy, as a spin-splitting is imposed. An impurity like La DOS is essential for this effect. It makes a correction to the Stoner factor, and provides an explanation of the recently observed weak ferro-magnetism in doped hexaborides.
Recent experimental works on doped hexaborides show many surprises . Among the unexpected properties is the observation of a very weak ferromagnetic state that persists up to large temperatures when small amounts of Ca are replaced by La in the hexaboride system, La<sub>x</sub>Ca<sub>(1-x)</sub>B<sub>6</sub> . This is surprising since the density-of-states at the Fermi energy, $`N(E_F)`$, and the Stoner factor are expected to be small in this system. Band calculations in the local density approximation (LDA) on SrB<sub>6</sub> and CaB<sub>6</sub> confirm that the DOS is low, with $`E_F`$ falling in a valley of the DOS, and spin-polarized calculations for EuB<sub>6</sub> show a very weak spill-over of the spin-density from the Eu-f spin to the rest of the valence band. Electronic structure calculations based on the local (spin) density approximation, either by spin-polarized calculations or by calculations of the Stoner factor $`\overline{S}`$, are in general able to detect if a system is ferromagnetic or not . Because the Stoner factor for La<sub>x</sub>Sr<sub>(1-x)</sub>B<sub>6</sub> is far below the limit for magnetism, it is tempting to suggest that the observed magnetism is not a band magnetism, but something else, perhaps excitonic or a polarization of the dilute electron-gas . However, for Ce it is found that the Stoner factor underestimates the tendency towards magnetism, because the stabilization energy is not only coming from exchange, and the Stoner criterion has to be corrected for a potential energy . Even if this correction is small in Ce, there is nothing in principle that prevents a large correction to the Stoner model in other systems. In this letter we show that La doped SrB<sub>6</sub> is such a system. According to the traditional understanding of ferromagnetism, the exchange energy is the source of a magnetic transition, but here we find that the band structure is such that the exchange splitting triggers off charge transfers that contribute to the total gain in energy via the Coulomb interaction.
Electron energies are calculated by the linear muffin-tin orbital (LMTO) method using the LDA. The structure of SrB<sub>6</sub> is simple cubic, with one Sr in the corner, and a B<sub>6</sub> octahedron in the center. The internal parameter $`z`$ which determines the shortest distance from one of the B-atoms to a cube surface, is in these calculations 0.215 of the lattice constant. Additional ’empty’ spheres are inserted at the cube edges, between the Sr-atoms, in order to account for non-sphericity of the potential in the open part of the structure. Calculations where made for the basic unit cell, a double cell and a large supercell of 8 cubes (2 cube lengths in x,y and z-direction), containing totally 56 atoms and 24 empty spheres. The basis set for the largest unitcell includes s,p,d for the atoms, and s,p for empty spheres. The small cells include also f-states for the Sr-sites. The occupied f-part is always very small, even for the La impurity site. The lattice constant is fixed at 7.93 a.u. in each calculation. No lattice relaxation around the La impurity is considered. The number of k-points varies from 286 to 20 irreducible k-points for the unit cells of 10 and 80 sites, respectively.
The bandstructure for undoped SrB<sub>6</sub> agree well with the one in ref. . The paramagnetic (non-spinpolarized) DOS from the supercell calculation for LaSr<sub>7</sub>B<sub>48</sub> is shown in fig. 1. The total DOS at $`E_F,N(E_F)`$, about 48 states/cell/Ry, is very concentrated on the single La site, with $`N_{La}(E_F)`$ 19 states/cell/Ry, cf Table 1. On the remaining 55 sites, each Sr has on the average 1.0 and each B roughly 0.4 states/cell/Ry, respectively. By comparison, a rigid-band model obtained by adding one electron to the band structure without the La impurity (cf. fig. 1) gives very different values: Of the total DOS, 32 states/cell/Ry, each Sr has 1.4 and each B 0.25 states/Ry. Clearly, La behaves a localized impurity site with a very large local $`N_{La}(E_F)`$, much larger than on Sr. The additional valence electron remains located on the La, as can be seen in Table 2. By comparing the charges in the undoped and doped supercell one finds that the remaining Sr$`{}_{7}{}^{}B_{48}^{}`$-spheres receive no additional charge. The difference in charge between the two systems is never larger than 0.01 electron per site, among these sites. The derivative of the DOS becomes very large, about 1400 states/cell/Ry<sup>2</sup>, out of which 700 is coming from the La site, compared to about 30 for each Sr site when the rigid-band model is applied to the undoped bandstructure. This asymmetric and uneven distribution of the DOS near $`E_F`$ is a condition for real space charge transfers as T increases and as an exchange splitting is imposed. The transfer goes in both cases to the La site from the rest of the system.
Spin-polarized calculations for pure SrB<sub>6</sub>, and for a double cell SrLaB<sub>12</sub>, do not give stable moments of a significant size. The supercell calculation of LaSr<sub>7</sub>B<sub>48</sub> using 20 k-points finds a small moment of 0.10 $`\mu _B`$/cell at low temperature. The ferromagnetic state is about 10 mRy/supercell lower in total energy than the paramagnetic solution. If fewer number of k-points are used one finds that the (paramagnetic) DOS and its energy derivative near $`E_F`$ are larger, cf Table 1. The spin moment increases to 0.24 $`\mu _B`$/cell for the lowest number of k-points, with a total energy 11 mRy lower than the paramagnetic state. The moment is always largest on La, with more than 50 percent of the total moment on and near the La impurity. The moment on B-sites far from the impurity is negligible. This fact and the uneven distribution of local DOS show that a rigid-band picture of dilute La-doped hexaborides does not apply. The bands near $`E_F`$ are modified by the spacially non-uniform perturbation of the potential around the impurity site.
The original Stoner model can be applied together with ab-initio paramagnetic band results in order to study the spin susceptibility or the onset of magnetism . An applied exchange splitting, $`\xi `$, leads to a transfer from minority spins to majority spins, and a loss in kinetic energy $`\mathrm{\Delta }K=N\xi ^2`$, where $`N`$ the DOS at $`E_F`$ is assumed to be constant. The same spin transfer leads to a gain in exchange energy $`\mathrm{\Delta }E_x=N^2I_s\xi ^2`$, where the exchange integral, $`I_s`$, can be calculated within LDA . If $`\mathrm{\Delta }E_x\mathrm{\Delta }K`$, there will be a total gain in energy and a ferromagnetic transition can occur. By elimination of factors of $`N`$ and $`\xi ^2`$ one obtains the Stoner criterion, $`\overline{S}=NI_s1`$. If a partial DOS is not constant near $`E_F`$, there will be a possibility of charge transfers as function of $`\xi `$, and additional contributions to the total energy. As in ref. , it is here possible to identify one sub-band, the La-impurity (mainly d) band, which has a large derivative of its DOS, $`N_{La}^{}(E_F)`$. The total DOS at $`E_F`$ is $`N=N_{La}+N_v`$, where $`N_v`$ is the rest of the DOS. By neglect of all other derivatives one can make a model of the charge transfers as function of $`\xi `$ and $`k_BT`$. From ref , we get the charge transfers $`\mathrm{\Delta }q_\xi =N_vN_{La}^{}\xi ^2/N`$ and $`\mathrm{\Delta }Q=N_vN_{La}^{}\pi ^2(k_BT)^2/6N`$, respectively, where it is assumed that $`N_{La}^{}k_BT`$ and $`N_{La}^{}\xi `$ are small.
Next we calculate the Coulomb energy associated with the charge transfer, $`\mathrm{\Delta }E(T)=U_0\mathrm{\Delta }Q`$, from self-consistent paramagnetic calculations at different T. The charge density is $`\rho (T,r)=_{\stackrel{}{k},j}f(ϵ_\stackrel{}{k}^j,E_F,T)\mathrm{\Psi }(ϵ_\stackrel{}{k}^j,r)^2`$ , where $`f`$ is the Fermi-Dirac function, and $`\mathrm{\Psi }(ϵ_k^j,r)`$ is the wavefunction for point $`k`$ and band $`j`$. The electronic total energy, $`E_{tot}`$, contains kinetic, Coulomb and exchange-correlation terms:
$$E_{tot}(T)=fϵN(ϵ)𝑑ϵ(\frac{1}{2}V_e+\epsilon \mu )\rho d^3r+E_M$$
(1)
$`V_e`$ is the electronic part of the Coulomb potential, $`E_M`$ is the Madelung part of the electrostatic energy, and $`\epsilon `$ and $`\mu `$ are the exchange-correlation energy and potential, respectively.
The first term in Eq. 1, the kinetic term, becomes $`E_K=\frac{\pi ^2}{3}(k_BT)^2N(E_F)`$, giving the usual Sommerfeld term in the specific heat $`C_{el}=\frac{E_K}{T}=\gamma T`$. The difference, $`E(T)`$, between the fully calculated $`E_{tot}`$ in eq. 1 and $`E_K`$ defines of the non-kinetic contribution due to charge transfers:
$$E(T)=fϵN(ϵ)𝑑ϵE_{tot}(T)=(\frac{1}{2}V_e+\epsilon \mu )\rho d^3rE_M(T)$$
(2)
and
$$\mathrm{\Delta }E(T)=E(T)E(0)$$
(3)
Calculations of $`E(T)`$ need to be selfconsistent because of relaxation, while $`E_K`$ and the Stoner product are only moderately affected by non-linear response effects from the system. At low T is $`\mathrm{\Delta }E`$ nearly proportional to $`\mathrm{\Delta }Q`$, $`\mathrm{\Delta }EU_0\mathrm{\Delta }Q`$, while it saturates for higher T. The calculated parameter $`U_0`$ varies between 1 and 2.5 Ry/el./La for $`k_BT`$ in the interval 0.5 to 15 mRy. This is from calculations using 10 irreducible k-points. It is probable that the main part of $`\mathrm{\Delta }E`$ is coming from the Madelung term, because of the sizeable charge transfer to La. This value of $`U_0`$ is calculated witout taking into account lattice expansion, but the mechanism of large charge transfers is expected to have an influence on thermal expansion and specific heat.
The charge transfer towards La is a result of the increasing La-DOS near $`E_F`$ (fig. 1). A very similar charge transfer pattern will result from a spin splitting, $`\xi `$, of the para-magnetic bands. The similarity between the two types of charge transfer permits us to write $`\mathrm{\Delta }E_\xi =U_0\mathrm{\Delta }q_\xi =U_0N_{La}^{}N_v\xi ^2/N`$. With this energy included, one obtains a modified Stoner criterion :
$$\overline{S}_U=NI_s+U_0N_{La}^{}N_v/N^21$$
(4)
The calculated Stoner product $`NI_s`$ is 0.21, $`N_{La}^{}`$ 730 per Ry$`{}_{}{}^{2}`$ cell, N $`48`$ and $`N_v29`$ (per Ry $``$ cell), to give $`\mathrm{\Delta }q_\xi 440\xi ^2`$ and $`\mathrm{\Delta }Q=720(k_BT)^2`$, with $`\xi `$ and $`k_BT`$ in Ry. The values make the Coulomb term in eq. 4 very large, about 14. However, even if the Coulomb energy represents a major correction in this case, there are several facts indicating that other energies are involved as well. First, the exchange splitting $`\xi `$ is not uniform over all sites. The spin-polarized calculation shows that $`\xi `$ is finite only at or very near to La, and the energy of a variable $`\xi `$ is not counted. Second, although the amount of charge transfers due to $`\xi `$ and $`k_BT`$ are similar, there are long tails of the Fermi-Dirac distribution which make the radial dependencies of $`\mathrm{\Delta }q_\xi `$ and $`\mathrm{\Delta }Q(T)`$ different. Third, the effective derivative of the La-DOS, $`N_{La}^{}(ϵ,T)`$, (taken as $`(\frac{df}{dϵ})N_f^{}(ϵ)𝑑ϵ`$), is likely to be largest at low T, because the bare DOS has not the same large derivative over a wide energy range. The mechanism of charge transfer and magnetism is therefore expected to disappear at some large T. This is confirmed in the spinpolarized results. By raising $`k_BT`$ to 3 mRy in the calculation with 10 k-points, the magnetic moment goes to zero. It is also interesting to note that the value of $`N_{La}^{}`$ is larger in the calculation with fewer number of k-points, as shown in Table 1. The number of k-points are insufficient for a good convergence in this case, but the variations of $`N_{La}^{}`$ and the magnetic moment in the spin-polarized results as a function of the number of k-points, are similar.
The $`\xi ^2`$ dependence of the charge transfer is a result of assuming at most a linear DOS variation in the simple model. In reality, there are more complex variations of the DOS and its derivatives around $`E_F`$. By imposing a $`\xi `$ on the calculated bands rigidly, one finds that $`\mathrm{\Delta }q_\xi `$ is proportional to $`\xi ^2`$ only below $``$ 1 mRy, it is linear around the interval 2-4 mRy, and it saturates at about 0.007 el./La for $`\xi `$ larger than about 5 mRy. By comparing the energies for large $`\xi `$ one can estimate the maximum $`\xi `$ for a stable moment:
$$N^2I_s\xi ^2+U_00.007=N\xi ^2$$
(5)
With $`U_0`$=1.5, this gives $`\xi `$=16 mRy and a moment of $``$ 0.8 $`\mu _B`$ per cell. These values are much larger than the values from the stable spin-polarized solution, about 3 mRy for $`\xi `$ and 0.1 $`\mu _B`$ for the moment, located on La only. The model can reproduce the essential dependencies of the magnetic instability, and it helps to understand the reasons behind the transition. But for quantitative values one has to rely on the spin-polarized calculations.
The magnetic state in the supercell is fragile, and in a smaller cell not even stable. The fact that an insufficient number of k-points can make the moment to increase, is not meant to represent the real physical situation. It serves to demonstrate that the moment depends on the derivative of the DOS, and thus to confirm the hypothesis that the charge transfer contributes to the stabilization of the magnetic state. The magnetism is in this case sensitive to other features in the electronic structure then what is normal for ferromagnets. A large Stoner factor needs a large DOS, usually a peak, which often is sensitive to disorder and thermal smearing. But a derivative of the DOS can be large in a fairly wide energy interval, aside a peak in the DOS. The contribution to magnetism from charge transfer energies may therefore resist well to thermal effects, and thereby be consistent with the observation of high Curie temperatures.
When it comes to explain the properties of the real hexaboride system from the band results, one has to rely on extrapolations. The very low doping ($`x`$=0.01) for optimal magnetism in real La<sub>x</sub>Ca<sub>(1-x)</sub>B<sub>6</sub> , is much smaller than the doping in the largest supercell ($`x`$=0.125) studied here. Calculations for realistically large supercells are not possible at present. However, there are several indications that the mechanism favorable to magnetism will remain or even be stronger at lower doping. First, by increasing the La content to $`x=\frac{1}{2}`$ in a double cell LaSrB<sub>12</sub>, one finds a more rigid-band like La-DOS than for the dilute impurity in the supercell, and the local La-DOS resembles more the Sr-DOS. Effects of charge transfers will therefore be less pronounced than for a single La-impurity, and no magnetic state was found in this case. Second, the increasing La-La distance for less doping, will maintain a narrow, impurity-like DOS localized on La. The Fermi-energy will not cut this band in a rigid-band manner at a lower energy, where the DOS and its derivative is smaller, even though the number of electrons per average site is decreasing with lower doping. Instead, since the additional charge remains localized at the impurity, one expects that the position of $`E_F`$ remains rather fixed relative to the La-band, with large local values of $`N_{La}`$ and $`N_{La}^{}`$ even for lower doping. On the other hand, in the large volume far from the La-site one expects that charges and DOS conditions will be as for the undoped material, not knowing about the effects of La in part of the system. In that case $`N_v`$ will approach $`N`$, where in $`N`$ only the part of the DOS which is affected by the impurity should enter in eq. 4, and the correction to the Stoner factor will remain large.
The appearance of ferromagnetism is therefore a natural consequence of the energy changes associated with an exchange splitting. These energies can be estimated within a mean-field theory. Favorable conditions for magnetism are non-constant DOS at $`E_F`$ to make transfers possible, and a negative change in total energy as function of the transfers. The non-rigid band like electronic structure around the dilute La impurity shows up in the distribution of the ferromagnetic moment. According to these results, it is mainly La and the nearest sites which takes the major part of the total moment. B-sites remain essentially non-polarized. Alternative theories based on the excitonic mechanism need to take into account the non-rigid band effect of the La-doping. If a rigid-band mechanism did apply, one should expect a uniform distribution of the moment. Theories of spontaneous ferromagnetism of the dilute electron gas have also been proposed . However, the very large electron-gas parameter $`r_s`$ necessary for this ($``$ 30) is much too large. It is the total (and not only the part due to doping) electron density that should be considered, and for this type of compound $`r_s`$ is smaller than 1.5 within the atomic sites and about 3 within the empty sites.
What is shown here is that ferromagnetism occurs in the supercell of LaSr<sub>7</sub>B<sub>48</sub>. This is despite the fact that according to the low Stoner factor, ferromagnetism should not appear. By doing the calculations with different number of k-points it has been possible to show that the actual derivative of the local DOS at $`E_F`$, which is conditional for charge transfers, scales with the stability of the magnetic moment. This demonstrates that the mechanism for a ferromagnetic is not exclusively found in the exchange energy. The necessary condition of a localized, non-constant La-DOS near $`E_F`$, appears to be enforced as the doping decreases, making the mechanism of charge transfers to persist at lower doping. A computational verification of this will need a very large unit cell or a Green-function method for the limit of dilute doping. |
warning/0001/astro-ph0001279.html | ar5iv | text | # Untitled Document
THE FRACTIONAL KINETIC EQUATION
AND
THERMONUCLEAR FUNCTIONS
H.J. Haubold
Outer Space Office, United Nations, Vienna International Centre,
P.O.Box 500, 1400 Vienna, Austria
and
A.M. Mathai
Department of Mathematics and Statistics, McGill University,
805 Sherbrooke Street West, Montreal, Quebec, Canada H3A 2K6
Abstract. The paper discusses the solution of a simple kinetic equation of the type used for the computation of the change of the chemical composition in stars like the Sun. Starting from the standard form of the kinetic equation it is generalized to a fractional kinetic equation and its solutions in terms of H-functions are obtained. The role of thermonuclear functions, which are also represented in terms of G- and H-functions, in such a fractional kinetic equation is emphasized. Results contained in this paper are related to recent investigations of possible astrophysical solutions of the solar neutrino problem.
1. Introduction
A spherically symmetric, non-rotating, non-magnetic, self-gravitating model of a star like the Sun is assumed to be in thermal equilibrium and hydrostatic equilibrium, with a non-uniform chemical composition throughout. The star is characterized by its mass, luminosity, effective surface temperature, radius, central density, and central temperature. For a given mass, four of these variables are independent and are governed by four simultaneous, non-linear, ordinary differential equations of the first order and four boundary conditions. Since there are four equations but more than four unknowns, additional information must be provided through the equation of state, nuclear energy generation rate, and the opacity (constitutive equations). The assumptions of thermal equilibrium and hydrostatic equilibrium imply that there is no time dependence in the equations describing the internal structure of the star (Kourganoff, 1973, Perdang, 1976, Clayton, 1983).
The evolution of a star like the Sun is governed by a second system of differential equations, the kinetic equations, describing the rate of change of chemical composition of the star for each species in terms of the reaction rates for destruction and production of that species (Kourganoff, 1973, Perdang, 1976, Clayton, 1983).
Methods for modeling processes of destruction and production have been developed for bio-chemical reactions and their unstable equilibrium states (Murray, 1989) and for chemical reaction networks with unstable states, oscillations, and hysteresis (Nicolis and Prigogine, 1977). Stability investigations of thermonuclear reactions of stellar interest have not yet been worked out in detail. However, the potentiality of instabilities in thermonuclear chains may not be overlooked, since, as was pointed out once by Eddington, “what is possible in the (Cavendish) Laboratory may not be too difficult in the Sun” (Perdang, 1976, Mestel, 1999).
Consider an arbitrary reaction characterized by a time dependent quantity $`N=N(t)`$. It is possible to equate the rate of change $`dN/dt`$ to a balance between the destruction rate $`d`$ and the production rate $`p`$ of $`N`$, that is $`dN/dt=d+p`$. In general, through feedback or other interaction mechanisms, destruction and production depend on the quantity $`N`$ itself: $`d=d(N)`$ or $`p=p(N)`$. This dependence is complicated since the destruction or production at time $`t`$ depends not only on $`N(t)`$ but also on the past history $`N(\tau ),\tau <t`$, of the variable $`N`$. This may be formally represented by
$$dN/dt=d(N_t)+p(N_t),$$
(1)
where $`N_t`$ denotes the function defined by $`N_t(t^{})=N(tt^{})`$, $`t^{}>0`$. Here $`d`$ and $`p`$ are functionals and eq. (1) represents a functional-differential equation. In the following we study a special case of this equation, namely the equation
$$dN/dt=\alpha N(t)$$
(2)
with a constant $`\alpha >0`$. Eq. (2) implies that spatial fluctuations or inhomogenieties in the quantity $`N(t)`$ are neglected. The standard solution of the differential equation (2) will be briefly discussed in Section 2 and the generalization to a fractional differential equation and its solution will be derived in Section 3. Conclusions will be drawn in Section 4.
2. Standard Kinetic Equation
The production and destruction of species is described by kinetic equations governing the change of the number density $`N_i`$ of species $`i`$ over time, that is,
$$\frac{dN_i}{dt}=\underset{j}{}N_iN_j<\sigma \upsilon >_{ij}+\underset{k,li}{}N_kN_l<\sigma \upsilon >_{kl},$$
(3)
where $`<\sigma \upsilon >_{mn}`$ denotes the reaction probability for an interaction involving species $`m`$ and $`n`$, and the summation is taken over all reactions which either produce or destroy the species $`i`$ (Haubold and Mathai, 1995). For a gas of mass density $`\rho `$, the number density $`N_i`$ of the species $`i`$ is expressed in terms of its abundance $`X_i`$, by the relation $`N_i=\rho N_AX_i/A_i,`$ where $`N_A`$ is Avogadro’s number and $`A_i`$ is the mass of $`i`$ in mass units. The mean lifetime $`\tau _j(i)`$ of species $`i`$ for destruction by species $`j`$ is given by the relation
$$\lambda _j(i)=\frac{1}{\tau _j(i)}=N_j<\sigma \upsilon >_{ij}=\rho N_A\frac{X_j}{A_j}<\sigma \upsilon >_{ij},$$
(4)
where $`\lambda _j(i)`$ is the decay rate of $`i`$ for interactions with $`j`$. Eq. (4) reveals the physical importance of $`<\sigma \upsilon >_{ij}`$ for the kinetic equation (3).
In the case of a nondegenerate, nonrelativistic gas, if a nonresonant charged nuclear reaction proceeds at low energies dominated by Coulomb-barrier penetration, the reaction probability $`<\sigma \upsilon >_{mn}`$ takes the form (Clayton, 1983, Bergstroem et al., 1999)
$$<\sigma \upsilon >_{mn}=(\frac{8}{\pi \mu })^{1/2}\underset{\nu =0}{\overset{2}{}}\frac{1}{(kT)^{\nu +1/2}}\frac{S^{(\nu )}(0)}{\nu !}I_2(\nu ,a,z,\rho ),$$
(5)
where $`I_2`$ represents a thermonuclear function given by
$`I_2(\nu 1,a,z,\rho )`$ $`=`$ $`{\displaystyle _0^{\mathrm{}}}𝑑yy^{\nu 1}e^{ayzy^\rho },\nu >0,a>0,z>0,\rho >0`$ (6)
$`=`$ $`{\displaystyle \frac{a^\nu }{\rho }}H_{0,2}^{2,0}\left[az^{\frac{1}{\rho }}\right|{}_{(\nu ,1),(0,\frac{1}{\rho })}{}^{}],`$
where $`H`$ denotes Fox’s H-function (Mathai, 1993, Haubold and Mathai, 1998), which was introduced into mathematical analysis to unify and extend existing results on symmetrical Fourier kernels.
Fox’s H-function (Mathai and Saxena, 1973, 1978) is defined in terms of a Mellin-Barnes type integral
$$H(z)=H_{p,q}^{m,n}\left[z|_{(b_1,\beta _1),\mathrm{},(b_q,\beta _q)}^{(a_1,\alpha _1),\mathrm{},(a_p,\alpha _p)}\right]=\frac{1}{2\pi i}_Lh(s)z^s𝑑s,$$
(7)
where
$$h(s)=\frac{\{\mathrm{\Pi }_{j=1}^m\mathrm{\Gamma }(b_j+\beta _js)\}\{\mathrm{\Pi }_{j=1}^n\mathrm{\Gamma }(1a_j\alpha _js)\}}{\{\mathrm{\Pi }_{j=m+1}^q\mathrm{\Gamma }(1b_j\beta _js)\}\{\mathrm{\Pi }_{j=n+1}^p\mathrm{\Gamma }(a_j+\alpha _js)\}}$$
(8)
$`i=\sqrt{1}`$, and $`L`$ is a suitable path which will be briefly described in the following. An empty product is interpreted as unity and it is assumed that the poles of $`\mathrm{\Gamma }(b_j+\beta _js),j=1,\mathrm{},m,`$ are separated from the poles of $`\mathrm{\Gamma }(1a_j\alpha _js),j=1,\mathrm{},n;a_1,\mathrm{},a_p,b_1,\mathrm{},b_q`$ are complex numbers; $`\alpha _1,\mathrm{},\alpha _p,\beta _1,\mathrm{},\beta _q`$ are positive real numbers. The poles of $`\mathrm{\Gamma }(b_j+\beta _js),j=1,\mathrm{},m,`$ are at the points
$$s=(b_j+\nu )/\beta _j,j=1,\mathrm{},m,\nu =0,1,\mathrm{}$$
and the poles of $`\mathrm{\Gamma }(1a_j\alpha _js),j=1,\mathrm{},n,`$ are at
$$s=(1a_k+\lambda )/\alpha _k,k=1,2,\mathrm{},n,\lambda =0,1,\mathrm{}$$
The condition of separability of these two sets of poles imposes that there be a strip in the complex s-plane where the H-function has no poles. There are three types of paths $`L`$ possible. These correspond to paths 1,2, and 3 shown in Fig. 1.
Fig. 1.
For all practical problems where H-functions are to be applied one mainly requires paths 2 and 3. Hence the conditions coming from these two paths are given in the following. It is to be pointed out that when more than one path $`L`$ makes sense then it can be shown that they lead to the same function and thus there will be no ambiguity.
Let
$$\mu =\underset{j=1}{\overset{q}{}}\beta _j\underset{j=1}{\overset{p}{}}\alpha _j$$
(9)
and
$$\beta =\mathrm{\Pi }_{j=1}^p\alpha _j^{\alpha j}\mathrm{\Pi }_{j=1}^q\beta _j^{\beta _j}.$$
(10)
The H-function exists for the following cases:
$$\begin{array}{cc}\text{Case (i)}\hfill & q1,\mu >0,H(z)\text{exists for all}z,z0.\hfill \\ \text{Case (ii)}\hfill & q1,\mu =0,H(z)\text{exists for}|z|<\beta ^1.\hfill \\ \text{Case (iii)}\hfill & p1,\mu <0,H(z)\text{exists for all}z,z0.\hfill \\ \text{Case (iv)}\hfill & p1,\mu =0,H(z)\text{exists for}|z|>\beta ^1.\hfill \end{array}$$
In the above cases it is assumed that the basic condition is satisfied that the poles of $`\mathrm{\Gamma }(b_j+\beta _js),j=1,\mathrm{},m`$ and $`\mathrm{\Gamma }(1a_j\alpha _js),j=1,\mathrm{},n`$ are separated. Note that in cases (i) and (ii) the H-function is evaluated as the sum of the residues at the poles of $`\mathrm{\Gamma }(b_j+\beta _js),j=1,\mathrm{},m`$ and in cases (iii) and (iv) the H-function is evaluated as the sum of the residues at the poles of $`\mathrm{\Gamma }(1a_j\alpha _js),j=1,\mathrm{},n`$.
When $`\alpha _1=\mathrm{}=\alpha _p=\beta _1=\mathrm{}=\beta _q=1`$, the H-function reduces to Meijers’s G-function. That is,
$$H_{p,q}^{m,n}\left[z|_{(b_1,1),\mathrm{},(b_q,1)}^{(a_1,1),\mathrm{},(a_p,1)}\right]=G_{p,q}^{m,n}\left[z|_{b_1,\mathrm{},b_q}^{a_1,\mathrm{},a_p}\right].$$
(11)
When $`\rho `$ in eq. (6) is real and rational, then the H-function can be reduced to a G-function by using the multiplication formula for gamma functions.
$$\mathrm{\Gamma }(mz)=(2\pi )^{\frac{1m}{2}}m^{mz\frac{1}{2}}\mathrm{\Pi }_{j=0}^{m1}\mathrm{\Gamma }\left(z+\frac{j}{m}\right),m=1,2,\mathrm{}$$
(12)
In the case of Coulomb-barrier penetration (Gamow factor), $`\rho =\frac{1}{2},I_2`$ reduces to
$$I_2(\nu ,a,z,\frac{1}{2})=\frac{a^{(\nu +1)}}{\pi ^{1/2}}G_{0,3}^{3,0}\left[\frac{az^2}{4}\right|{}_{\nu +1,a,\frac{1}{2}}{}^{}],$$
(13)
where $`G`$ denotes Meijer’s G-function, which was introduced into mathematical analysis in attempts to give meaning to the generalized hypergeometric function $`{}_{p}{}^{}F_{q}^{}`$ in the case $`p>q+1`$.
Proceeding with eq. (3), the first sum in eq. (3) can also be written as
$$\underset{j}{}N_iN_j<\sigma \upsilon >_{ij}=N_i(\underset{j}{}N_j<\sigma \upsilon >_{ij})=N_ia_i,$$
(14)
where $`a_i`$ is the statistically expected number of reactions per unit volume per unit time destroying the species $`i`$. It is also a measure of the speed in which the reaction proceeds. In the following we are assuming that there are $`N_j(j=1,\mathrm{},i,\mathrm{})`$ species $`j`$ per unit volume and that for a fixed $`N_i`$ the number of other reacting species that interact with the i-th species is constant in a unit volume. Following the same argument we have for the second sum in eq. (3) accordingly,
$$+\underset{k,li}{}N_kN_l<\sigma \upsilon >_{kl}=+N_ib_i,$$
(15)
where $`N_ib_i`$ is the statistically expected number of the i-th species produced per unit volume per unit time for a fixed $`N_i`$. Note that the number density of species $`i,N_i=N_i(t)`$, is a function of time while the $`<\sigma \upsilon >_{mn}`$, containing the thermonuclear functions (see eqs. (5) and (6)), are assumed to depend only on the temperature of the gas but not on the time $`t`$ and number densities $`N_i`$. Then eq. (1) implies that
$$\frac{dN_i(t)}{dt}=(a_ib_i)N_i(t).$$
(16)
For eq. (16) we have three distinct cases, $`c_i=a_ib_i>0,c_i<0,`$ and $`c_i=0`$, of which the last case says that $`N_i`$ does not vary over time, which means that the forward and reverse reactions involving species $`i`$ are in equilibrium; such a value for $`N_i`$ is called a fixed point and corresponds to a steady-state behavior. The first two cases exhibit that either the destruction $`(c_i>0)`$ of species $`i`$ or production $`(c_i<0)`$ of species $`i`$ dominates.
For the case $`c_i>0`$ we have
$$\frac{dN_i(t)}{dt}=c_iN_i(t),$$
(17)
with the initial condition that $`N_i(t=0)=N_0`$ is the number density of species $`i`$ at time $`t=0`$, and it follows that
$$N_i(t)dt=N_0e^{c_it}dt.$$
(18)
The exponential function in eq. (18) represents the solution of the linear one-dimensional differential equation (17) in which the rate of destruction of the variable is proportional to the value of the variable. Eq. (17) does not exhibit instabilities, oscillations, or chaotic dynamics, in striking contrast to its cousin, the logistic finite-difference equation (Perdang, 1976, Haubold and Mathai, 1995). A thorough discussion of eq. (17) and its standard solution in eq. (18) is given in Kourganoff (1973).
3. Fractional Kinetic Equation
In the following, for the sake of brevity, the index $`i`$ in eq. (17) will be dropped. The standard kinetic equation (17) can be integrated
$$N(t)N_0=c_0D_t^1N(t),$$
(19)
where $`{}_{0}{}^{}D_{t}^{1}`$ is the standard Riemann integral operator. The generalization of this operator to the fractional integral of order $`p>0`$ is denoted by $`{}_{a}{}^{}D_{t}^{p}`$ and is defined, following Riemann-Liouville, based on the Cauchy formula, by
$${}_{a}{}^{}D_{t}^{p}f(t)=\frac{1}{\mathrm{\Gamma }(p)}_a^t𝑑\tau f(\tau )(t\tau )^{p1},p>0$$
(20)
with
$${}_{a}{}^{}D_{t}^{0}f(t)=f(t)$$
(Oldham and Spanier, 1974, Miller and Ross, 1993). The most general fractional integral operator of the type (20) contains Fox’s H-function (7) as the kernel function. If $`f(t)`$ is continuous for $`ta`$, then integration of arbitrary real order has the property
$${}_{a}{}^{}D_{t}^{p}(_aD_t^qf(t))=_aD_t^{pq}f(t).$$
Replacing the Riemann integral operator by the fractional Riemann-Liouville operator $`{}_{0}{}^{}D_{t}^{\nu }`$ in eq. (19), we obtain a fractional integral equation corresponding to eq. (17)
$$N(t)N_0=c_0^\nu D_t^\nu N(t).$$
(21)
For dimensional reasons, the coefficient $`c`$ in eq. (19), containing the probabilities of the reaction under consideration, had to be replaced by $`c^\nu `$ accordingly.
The Laplace transform of the Riemann-Liouville fractional integral is
$$L\{{}_{0}{}^{}D_{t}^{p}f(t);p\}=p^pF(p),$$
(22)
where
$$F(p)=\mathrm{\Gamma }(p)_{\tau =0}^{\mathrm{}}𝑑\tau e^{p\tau }f(\tau ).$$
In order to solve eq. (21), the integral equation is exposed to a Laplace transformation leading to
$$N(p)=L\{N(t);p\}=N_0\frac{p^1}{1+(\frac{p}{c})^\nu }.$$
(23)
To arrive at a representation of eq. (23) in terms of Fox’s H-function, Mathai and Saxena’s (1978) result can be used,
$`{\displaystyle \frac{z^\beta }{1+az^\alpha }}`$ $`=`$ $`a^{\frac{\beta }{\alpha }}H_{1,1}^{1,1}\left[az^\alpha \right|{}_{(\frac{\beta }{\alpha },1)}{}^{(\frac{\beta }{\alpha },1)}],`$
$`N(p)`$ $`=`$ $`N_0{\displaystyle \frac{1}{c}}H_{1,1}^{1,1}\left[\left({\displaystyle \frac{c}{p}}\right)^\nu \right|{}_{(\frac{1}{\nu },1)}{}^{(\frac{1}{\nu },1)}].`$ (24)
To prepare eq.(24) for an inverse Laplace transform, the following two fundamental properties of an H-function can be used,
$$\frac{1}{k}H_{p,q}^{m,n}\left[z|_{(b_1,\beta _1),\mathrm{},(b_q,\beta _q)}^{(a_1,\alpha _1),\mathrm{},(a_p,\alpha _p)}\right]=H_{p,q}^{m,n}\left[z^k|_{(b_1,k\beta _1),\mathrm{},(b_q,k\beta _q)}^{(a_1,k\alpha _1),\mathrm{},(a_p,k\alpha _p)}\right],k>0,$$
(25)
$$H_{p,q}^{m,n}\left[z|_{(b_1,\beta _1),\mathrm{},(b_q,\beta _q)}^{(a_1,\alpha _1),\mathrm{},(a_p,\alpha _p)}\right]=H_{q,p}^{n,m}\left[\frac{1}{z}|_{(1a_1,\alpha _1),\mathrm{},(1a_p,\alpha _p)}^{(1b_1,\beta _1),\mathrm{},(1b_q,\beta _q)}\right],$$
(26)
leading to
$$N(p)=N_0\frac{1}{c\nu }H_{1,1}^{1,1}\left[\frac{p}{c}\right|{}_{(1\frac{1}{\nu },\frac{1}{\nu })}{}^{(1\frac{1}{\nu },\frac{1}{\nu })}],$$
(27)
where the H-function is defined in eq. (7).
The Laplace transform of Fox’s H-function (7) is given in terms of another H-function by
$$L\{H(z);p\}=\frac{1}{p}H_{q,p+1}^{n+1,m}\left[p\right|{}_{(1,1),(1a_p,\alpha _p)}{}^{(1b_q,\beta _q)}]$$
(28)
for $`0\mu 1`$ in (9), and
$$L\{H(z);p\}=\frac{1}{p}H_{p+1,q}^{m,n+1}\left[\frac{1}{p}\right|{}_{(b_q,\beta _q)}{}^{(0,1),(a_p,\alpha _p)}]$$
(29)
for $`\mu 1`$ in (9), respectively.
Further, having $`H(p)`$, the inverse Laplace transform of this H-function is given by
$$H(z)=L^1\{H(p),z\}=\frac{1}{z}H_{q,p+1}^{n,m}\left[z\right|{}_{(1a_p,\alpha _p),(1,1)}{}^{(1b_q,\beta _q)}]$$
(30)
for $`0\mu 1`$ in (9) and
$$H(z)=L^1\{H(p),z\}=\frac{1}{z}H_{p+1,q}^{m,n}\left[z\right|{}_{(b_q,\beta _q)}{}^{(a_p,\alpha _p),(0,1)}]$$
(31)
for $`\mu 1`$ in (9), respectively.
The above four Laplace transforms hold for
$`\mathrm{max}_{1jn}\mathrm{}\left(\frac{a_p^1}{\alpha _p}\right)<\mathrm{min}_{1jm}\mathrm{}\left(\frac{b_q}{\beta _q}\right).`$
Applying an inverse Laplace transform to the H-function in eq. (27) gives
$$N(t)=N_0\frac{1}{\nu }H_{1,2}^{1,1}\left[ct\right|{}_{(0,\frac{1}{\nu }),(0,1)}{}^{(0,\frac{1}{\nu })}],$$
(32)
which is the solution of the fractional kinetic equation (21). For the H-function in eq. (7) with (32), the following computable representation can be derived (Mathai and Saxena, 1978). When the poles of $`\mathrm{\Pi }_{j=1}^m\mathrm{\Gamma }(b_j\beta _js)`$ are simple, that is,
$$\beta _h(b_j+\lambda )\beta _j(b_h+\nu )$$
for $`jh;j,h=1,\mathrm{},m;\lambda ,\nu =0,1,2,\mathrm{}.`$ Then one obtains the following expansion for the H-function,
$`H_{p,q}^{m,n}(z)`$ $`=`$ $`{\displaystyle \underset{h=1}{\overset{m}{}}}{\displaystyle \underset{\nu =0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\left\{\mathrm{\Pi }_{j=1,jh}^m\mathrm{\Gamma }\left\{b_j\beta _j\frac{(b_h+\nu )}{\beta _h}\right\}\right\}}{\left\{\mathrm{\Pi }_{j=m+1}^q\mathrm{\Gamma }\left\{1b_j+\beta _j\frac{(b_h+\nu )}{\beta _h}\right\}\right\}}}`$ (33)
$`\times `$ $`{\displaystyle \frac{\left\{\mathrm{\Pi }_{j=1}^n\mathrm{\Gamma }\left\{1a_j+\alpha _j\frac{(b_h+\nu )}{\beta _h}\right\}\right\}}{\left\{\mathrm{\Pi }_{j=n+1}^p\mathrm{\Gamma }\left\{a_j\alpha _j\frac{(b_h+\nu )}{\beta _h}\right\}\right\}}}{\displaystyle \frac{(1)^\nu z^{(b_h+\nu )/\beta _h}}{(\nu )!\beta _h}},`$
which exists for all $`z0`$ if $`\mu >0`$ and for $`0<|z|<\beta ^1`$ if $`\mu =0`$, where $`\mu `$ and $`\beta `$ are given in eqs. (9) and (10). Comparing (32) with eq. (33), one obtains the series expansion
$$N(t)=N_0\underset{k=0}{\overset{\mathrm{}}{}}\frac{(1)^k}{\mathrm{\Gamma }(\nu k+1)}(ct)^{\nu k}.$$
(34)
For $`\nu =1`$, the exponential solution of the standard kinetic equation (18) is recovered.
4. Conclusions
The gravitationally stabilized solar fusion reactor, because of its density and temperature, is not a weakly interacting or high temperature plasma. Recently, Kaniadakis et al. (1997) and Coraddu et al. (1998) have explored the possibility that the electrical microfield distribution influences the particle dynamics, collisions are not two-body processes and retain memory beyond single scattering events, and that the velocity correlation function has long-time memory arising from the coupling of collective and individual degrees of freedom. In this connection particle diffusion is a non-Markovian process (anomalous diffusion) and diffusion and frictional coefficients are energy dependent. Based on these results, they conclude that the equilibrium statistical distribution function differs from the Maxwell-Boltzmannian one and is governed by generalized Boltzmann-Gibbs statistics.
Even if the deviations from the Maxwell-Boltzmann distribution, that are compatible with the current knowledge of the solar core behavior, are small, they are sufficient to sensibly modify the sub-barrier particle reaction rates and subsequently solar neutrino fluxes. The above authors have also shown that the respective modifications of the reaction rates do not affect bulk properties of the gravitationally stabilized solar core such as sound of speed or hydrostatic equilibrium which depend on the mean values obtained by averaging over the Maxwell-Boltzmann distribution function (Degli’Innocenti et al. 1998).
Haubold and Mathai (1998) have derived general closed-form representations of particle reaction rates that are suitable to incorporate changes of the Maxwell-Boltzmann distribution function, mindfully taking into account the fact that the whole distribution has physical meaning, contrary to the case of, say, the Gaussian law of errors. In this paper we proceeded one step further in generalizing the standard kinetic equation (17) to a fractional kinetic equation (21) and derived solutions of a fractional kinetic equation that contains the particle reaction rate (or thermonuclear function) (5) as a time constant, and provided the analytic technique to further investigate possible modifications of the reaction rate through a kinetic equation. The Riemann-Liouville operator in the fractional kinetic equation introduces a convolution integral with slowly-decaying power-law kernel, which is typical for memory effects referred to in Kaniadakis et al. (1997) and Coraddu et al. (1998). This technique may open an avenue to accommodate changes in standard solar model core physics (Dar and Shaviv, 1999) as proposed by Shaviv and Shaviv (1999) and Schatzman (1999).
In the solution of the fractional kinetic equation (21), given in eqs. (32) and (34), the standard exponential decay is recovered for $`\nu =1`$. However, eqs. (32) and (34), for $`0<\nu <1`$, show power law behavior for $`t\mathrm{}`$ and are constant (initial value $`N_0`$) for $`t0`$. In the investigations in this paper, we used, as an example, the standard thermonuclear function $`I_2^{(\mathrm{})}`$ = $`I_1`$; the same computations can be done for $`I_2^{(d)}`$, $`I_3`$, or $`I_4`$ with additional parameters showing modifications of the Maxwell-Boltzmann distribution function (Haubold and Mathai, 1998).
All analytic results in this paper have been achieved by the application of the theory of generalized hypergeometric functions, particularly Meijer’s G-function and Fox’s H-function, which seem to be the natural means for tackling the problems referred to above.
Acknowledgments
The investigation of formation of structure in a non-equilibrium solar fusion plasma was discussed in-depth on 12 September 1984 with Hans-Juergen Treder at the Einstein Laboratory for Theoretical Physics in Caputh. Short-time instabilities in solar nuclear reaction kinetics have been explored briefly in a letter of Jean Perdang dated 8 January 1991. The corresponding author of this paper (HJH) is grateful to both colleagues and wishes to put this on record even at such a late date. The same author acknowledges inspiring discussions on non-Markovian processes in a solar nuclear plasma with Piero Quarati at the occasion of the sixth UN/ESA Workshop on Basic Space Science at the Max-Planck-Institute for Radioastronomy in Bonn, 9-13 September 1996.
References
Bergstroem, L., Iguri, S., and Rubinstein, H.: 1999, Constraints on the
variation of the fine structure constant from big bang nucleosynthesis,
Phys. Rev. D60, 045005-1 - 045005-9.
Clayton, D.D.: 1983, Principles of Stellar Evolution and Nucleosynthesis, Second Edition, The University of Chicago Press, Chicago and London.
Coraddu, M., Kaniadakis, G., Lavagno, A., Lissia, M., Mezzorani, G., and Quarati, P.: 1998, Thermal distributions in stellar plasma, nuclear reactions and solar neutrinos,
http://xxx.lanl.gov/abs/nucl-th/981108.
Dar, A. and Shaviv, G.: 1999, The solar neutrino problem - an update, Physics Reports 311, 115-141.
Degli’Innocenti, S., Fiorentini, G., Lissia, M., Quarati, P., and Ricci, B.: 1998, Helioseismology can test the Maxwell-Boltzmann distribution,
http://xxx.lanl.gov/abs/astro-ph/9807078.
Haubold, H.J. and Mathai, A.M.: 1995, A heuristic remark on the periodic variation in the number of solar neutrinos detected on Earth, Astrophys. Space Sci. 228, 113-134.
Haubold, H.J. and Mathai, A.M.: 1998, On thermonuclear reaction rates, Astrophys. Space Sci. 258, 185-199.
Kaniadakis, G., Lavagno, A., Lissia, M., and Quarati, P.: 1998, Anomalous
diffusion modifies solar neutrino fluxes, Physica A261, 359-373,
http://xxx.lanl.gov/abs/astro-ph/9710173.
Kourganoff, V.: 1973, Introduction to the Physics of Stellar Interiors, D. Reidel Publishing Company, Dordrecht.
Mathai, A.M.: 1993, A Handbook of Generalized Special Functions for
Statistical and Physical Sciences, Clarendon Press, Oxford.
Mathai, A.M. and Saxena, R.K.: 1973, Generalized Hypergeometric
Functions with Applications in Statistics and Physical Sciences, Springer-Verlag, Lecture Notes in Mathematics Vol. 348, Berlin Heidelberg New York.
Mathai, A.M. and Saxena, R.K.: 1978, The H-function with Applications in Statistics and Other Disciplines, John Wiley and Sons, New Delhi.
Mestel, L.: 1999, The early days of solar structure theory, Physics Reports 311, 295-305.
Miller, K.S. and Ross, B.: 1993, An Introduction to the Fractional Calculus and Fractional Differential Equations, John Wiley and Sons, New York.
Murray, J.D.: 1989, Mathematical Biology, Biomathematics Texts Vol. 19, Springer-Verlag, Berlin.
Nicolis, G. and Prigogine, I.: 1977, Self-Organization in
Nonequilibrium Systems - From Dissipative Structures to Order Through Fluctuations, John Wiley and Sons, New York.
Oldham, K.B. and Spanier, J.: 1974, The Fractional Calculus, Academic Press, New York.
Perdang, J.: 1976, Lecture Notes in Stellar Stability, Part I and II, Instituto di Astronomia, Padova.
Schatzman, E.: 1999, Role of gravity waves in the solar neutrino problem, Physics Reports 311, 143-150.
Shaviv, G. and Shaviv, N.J.: 1999, Is there a dynamic effect in the screening of nuclear reactions in stellar plasma?, Physics Reports 311, 99-114. |
warning/0001/gr-qc0001074.html | ar5iv | text | # Quantum Isotropization of the Universe
## I Introduction
Two of the main questions one might ask in cosmology are:
1) Is the Universe eternal or it had a beginning, and in the last case, was this beginning given by an initial singularity?
2) Why the Universe we live in is remarkably homogeneous and isotropic, with very small deviations from this highly symmetric state?
The answer given by classical General Relativity (GR) to the first question, indicated by the singularity theorems , asserts that probably the Universe had a singular beginning. As singularities are out of the scope of any physical theory, this answer invalidates any description of the very beginning of the Universe in physical terms. One might think that GR and/or any other matter field theory must be changed under the extreme situations of very high energy density and curvature near the singularity, rendering the physical assumptions of the singularity theorems invalid near this point. One good point of view (which is not the only one) is to think that quantum gravitational effects become important under these extreme conditions. We should then construct a quantum theory of gravitation and apply it to cosmology. For instance, in the euclidean quantum gravity approach to quantum cosmology, a second answer to the first question comes out: the Universe may have had a non-singular birth given by the beginning of time through a change of signature.
In the same way, the naive answer of GR and the standard cosmological scenario to the second question is not at all satisfactory: the reason for the Universe be highly homogeneous and isotropic is a matter of initial conditions. However, solutions of Einstein’s equations with this symmetry are of measure zero; so, why the Universe is not inhomogeneous and/or anisotropic? Inflation is an idea that tries to explain this fact. Nevertheless, in order for inflation to happen, some special initial conditions are still necessary, although much more less stringent than in the case without inflation. Once again, quantum cosmology can help in this matter by providing the physical reasons for having the initial conditions for inflation.
Recently, many papers have been devoted to the application of the Bohm-de Broglie interpretation of quantum mechanics to quantum cosmology . One of the aims of these papers was to try to give an answer to the first question within a new perspective in quantum cosmology. In the framework of the Bohm-de Broglie interpretation of quantum mechanics applied to minisuperspace models, it was shown in References that quantum gravitational effects may indeed become important under extreme situations of very high energy density and curvature, and avoid the initial singularity of the classical models. The solutions that come out are then non-singular (eternal) cosmological models, but with a very hot phase, tending to the usual classical solutions when the energy scales become smaller than the Planck scale. Of course this is a much better answer than the one given by classical GR.
The aim of this paper, as a natural sequence of the references cited above, is to address the second question from the point of view of the Bohm-de Broglie interpretation of quantum cosmology. We want to investigate if quantum mechanical effects can isotropize an anisotropic cosmological model yielding an alternative mechanism for the isotropization of the Universe. Our strategy is to take classical anisotropic models which never isotropizes, quantize them, and examine if quantum effects can at least create isotropic phases. As we will see, quantum effects can not only create isotropic phases but they can also isotropize the models forever, for a large variety of quantum states. Other authors have also studied this problem, mainly taking Bianchi IX models, adopting other interpretations of quantum cosmology, and they arrive at similar results .
The mathematical models we take are Bianchi I models with and without matter. For the case without matter, one of the anisotropic degrees of freedom is suppressed and we end with a two-dimensional minisuperspace model. The classical solutions of this case are all singular and always anisotropic, with the trivial exception of flat space-time. When quantizing this system we find that the solutions of the Wheeler-DeWitt can be written in terms of superpositions of plane wave solutions. Taking gaussian superpositions we show that there are bohmian trajectories representing universes expanding from an initial singularity and others representing bouncing non-singular universes, both with isotropic phases during the course of their evolutions. There are also bohmian non-singular periodic solutions with periodic isotropic phases and non-singular bouncing solutions which are never isotropic. For the case with matter, we take a minimally coupled scalar field with arbitrary coupling constant $`\omega `$. The minisuperspace is four-dimensional. The classical solutions are again singular, and once they begin anisotropic they continue to be anisotropic forever. When quantizing the system, we arrive at a Wheeler-DeWitt equation which corresponds to a massless Klein-Gordon equation in four dimensions. We show that there are plenty of spherical wave solutions whose bohmian trajectories represent expanding universes that isotropizes permanently after some period, many of them which are also non-singular. gaussian superpositions also present this feature.
The paper is organized as follows: in the next section we present the minisuperspace models which we will investigate and their classical solutions. In section III we quantize these models, and obtain their corresponding Wheeler-DeWitt equations and general quantum solutions. In section IV we present the results concerning the quantum isotropization of those solutions through the investigation of the bohmian trajectories we obtain from them. We end up in section V with conclusions and discussions.
## II The classical minisuperspace model
Let us take the lagrangian
$$L=\sqrt{g}e^\varphi (Rw\varphi _{;\rho }\varphi ^{;\rho }).$$
(1)
For $`w=1`$ we have effective string theory without the Kalb-Ramond field. For $`w=3/2`$ we have a conformally coupled scalar field. Performing the conformal transformation $`g_{\mu \nu }=e^\varphi \overline{g}_{\mu \nu }`$ we obtain
$$L=\sqrt{g}[RC_w\varphi _{;\rho }\varphi ^{;\rho }],$$
(2)
where the bars have been omitted, and $`C_w(\omega +\frac{3}{2})`$. The cases of interest correspond to $`C_w>0`$.
The gravitational part of the minisuperspace model we study in this paper is given by the homogeneous and anisotropic Bianchi I line element
$`ds^2`$ $`=`$ $`N^2(t)dt^2+\mathrm{exp}[2\beta _0(t)+2\beta _+(t)+2\sqrt{3}\beta _{}(t)]dx^2+`$ (4)
$`\mathrm{exp}[2\beta _0(t)+2\beta _+(t)2\sqrt{3}\beta _{}(t)]dy^2+\mathrm{exp}[2\beta _0(t)4\beta _+(t)]dz^2.`$
This line element will be isotropic if and only if $`\beta _+(t)`$ and $`\beta _{}(t)`$ are constants . Inserting Equation (4) into the action $`S=Ld^4x`$, supposing that the scalar field $`\varphi `$ depends only on time, discarding surface terms, and performing a Legendre transformation, we obtain the following minisuperspace classical hamiltonian
$$H=\frac{N}{24\mathrm{exp}(3\beta _0)}(p_0^2+p_+^2+p_{}^2+p_\varphi ^2),$$
(5)
where $`(p_0,p_+,p_{},p_\varphi )`$ are canonically conjugate to $`(\beta _0,\beta _+,\beta _{},\varphi )`$, respectively, and we made the trivial redefinition $`\varphi \sqrt{C_w/6}\varphi `$.
We can write this hamiltonian in a compact form by defining $`y^\mu =(\beta _0,\beta _+,\beta _{},\varphi )`$ and their canonical momenta $`p_\mu =(p_0,p_+,p_{},p_\varphi )`$, obtaining
$$H=\frac{N}{24\mathrm{exp}(3y^0)}\eta ^{\mu \nu }p_\mu p_\nu ,$$
(6)
where $`\eta ^{\mu \nu }`$ is the Minkowski metric with signature $`(+++)`$. The equations of motion are the constraint equation obtained by varying the hamiltonian with respect to the lapse function $`N`$
$$\eta ^{\mu \nu }p_\mu p_\nu =0,$$
(7)
and the Hamilton’s equations
$$\dot{y}^\mu =\frac{}{p_\mu }=\frac{N}{12\mathrm{exp}(3y_0)}\eta ^{\mu \nu }p_\nu ,$$
(8)
$$\dot{p}_\mu =\frac{}{y^\mu }=0.$$
(9)
The solution to these equations in the gauge $`N=12\mathrm{exp}(3y_0)`$ is
$$y^\mu =\eta ^{\mu \nu }p_\nu t+C^\mu ,$$
(10)
where the momenta $`p_\nu `$ are constants due to the equations of motion and the $`C^\mu `$ are integration constants. We can see that the only way to obtain isotropy in these solutions is by making $`p_1=p_+=0`$ and $`p_2=p_{}=0`$, which yields solutions that are always isotropic, the usual Friedmann-Robertson-Walker (FRW) solutions with a scalar field. Hence, there is no anisotropic solution in this model which can classically becomes isotropic during the course of its evolution. Once anisotropic, always anisotropic. If we suppress the $`\varphi `$ degree of freedom, the unique isotropic solution is flat space-time because in this case the constraint (7) enforces $`p_0=0`$.
To discuss the appearance of singularities, we need the Weyl square tensor $`W^2W^{\alpha \beta \mu \nu }W_{\alpha \beta \mu \nu }`$. It reads
$$W^2=\frac{1}{432}e^{12\beta _0}(2p_0p_+^36p_0p_{}^2p_++p_{}^4+2p_+^2p_{}^2+p_+^4+p_0^2p_+^2+p_0^2p_{}^2).$$
(11)
Hence, the Weyl square tensor is proportional to $`\mathrm{exp}(12\beta _0)`$ because the $`p`$’s are constants (see Equations (9)), and the singularity is at $`t=\mathrm{}`$. The classical singularity can be avoided only if we set $`p_0=0`$. But then, due to Equation (7), we would also have $`p_i=0`$, which corresponds to the trivial case of flat space-time. Therefore, the unique classical solution which is non-singular is the trivial flat space-time solution.
## III Quantization of the classical model
The Dirac quantization procedure yields the Wheeler-DeWitt equation through the imposition of the condition
$$\widehat{}\mathrm{\Psi }=0,$$
(12)
on the quantum states, with $`\widehat{}`$ defined as in Equation (7) (we are assuming the covariant factor ordering) using the substitutions
$$p_\mu i\frac{}{y^\mu }.$$
(13)
Equation (12) reads
$$\eta ^{\mu \nu }\frac{^2}{y^\mu y^\nu }\mathrm{\Psi }(y^\mu )=0.$$
(14)
### A The empty model
In a first and simple example, we will freeze two degrees of freedom: the matter degree of freedom $`y^3=\varphi =0`$, and one of the anisotropic degrees of freedom, $`y^2=\beta _{}=0`$. We obtain a two-dimensional Klein-Gordon equation whose general solution is
$$\mathrm{\Psi }(\beta _0,\beta _+)=\{F(k)\mathrm{exp}[ik(\beta _0+\beta _+)]+G(k)\mathrm{exp}[ik(\beta _0\beta _+)]\}𝑑k,$$
(15)
with $`F(k)`$ and $`G(k)`$ arbitrary functions of $`k`$. Let us take gaussian superpositions
$$F(k)=G(k)=\mathrm{exp}[\frac{(kd)^2}{\sigma ^2}],$$
(16)
where $`\sigma `$ and $`d`$ are constants. In this case the wave function reads
$$\mathrm{\Psi }_1=\sigma \sqrt{\pi }\left\{\mathrm{exp}\left[\frac{(\beta _0+\beta _+)^2\sigma ^2}{4}\right]\mathrm{exp}[id(\beta _0+\beta _+)]+\mathrm{exp}\left[\frac{(\beta _0\beta _+)^2\sigma ^2}{4}\right]\mathrm{exp}[id(\beta _0\beta _+)]\right\}.$$
(17)
For the case where
$$F(k)=G(k)=\mathrm{exp}[\frac{(k+d)^2}{\sigma ^2}],$$
(18)
the wave function turns out to be
$$\mathrm{\Psi }_2=\sigma \sqrt{\pi }\left\{\mathrm{exp}\left[\frac{(\beta _0+\beta _+)^2\sigma ^2}{4}\right]\mathrm{exp}[id(\beta _0+\beta _+)]+\mathrm{exp}\left[\frac{(\beta _0\beta _+)^2\sigma ^2}{4}\right]\mathrm{exp}[id(\beta _0\beta _+)]\right\}.$$
(19)
In the next section we will calculate the bohmian trajectories corresponding to the solutions (17,19) and see if there are isotropic phases in these quantum models.
### B The scalar field model
In the general case of the four-dimensional minisuperspace model, we will investigate spherical-wave solutions of Equation (14). They read
$$\mathrm{\Psi }_3=\frac{1}{y}\left[f(y^0+y)+g(y^0y)\right],$$
(20)
where $`y\sqrt{_{i=1}^3(y^i)^2}`$.
One particular example is the gaussian superposition of plane wave solutions of Equation (14),
$$\mathrm{\Psi }_4(y^\mu )=\{F(\stackrel{}{k})\mathrm{exp}[i(|\stackrel{}{k}|y^0+\stackrel{}{k}.\stackrel{}{y})]+G(\stackrel{}{k})\mathrm{exp}[i(|\stackrel{}{k}|y^0\stackrel{}{k}.\stackrel{}{y})]\}d^3k,$$
(21)
where $`\stackrel{}{k}(k_1,k_2,k_3)`$, $`\stackrel{}{y}(y^1,y^2,y^3)`$, $`|\stackrel{}{k}|\sqrt{_{i=1}^3(k_i)^2}`$, with $`F(\stackrel{}{k})`$ and $`G(\stackrel{}{k})`$ given by
$$F(\stackrel{}{k})=G(\stackrel{}{k})=\mathrm{exp}[\frac{(|\stackrel{}{k}|d)^2}{\sigma ^2}].$$
(22)
After performing the integration in Equation (21) using spherical coordinates we obtain
$`\mathrm{\Psi }_4(y^0,y)`$ $`=`$ $`{\displaystyle \frac{i\pi ^{3/2}}{y}}\{[2d\sigma +i(y^0y)\sigma ^3]\mathrm{exp}[{\displaystyle \frac{(y^0y)^2\sigma ^2}{4}}]\mathrm{exp}[id(y^0y)][1+\mathrm{\Phi }({\displaystyle \frac{d}{\sigma }}+i(y^0y){\displaystyle \frac{\sigma }{2}})]`$ (24)
$`[2d\sigma +i(y^0+y)\sigma ^3]\mathrm{exp}[{\displaystyle \frac{(y^0+y)^2\sigma ^2}{4}}]\mathrm{exp}[id(y^0+y)][1+\mathrm{\Phi }({\displaystyle \frac{d}{\sigma }}+i(y^0+y){\displaystyle \frac{\sigma }{2}})]\},`$
where $`\mathrm{\Phi }(x)(2/\sqrt{\pi })_0^x\mathrm{exp}(t^2)𝑑t`$ is the probability integral. The wave function $`\mathrm{\Psi }_4`$ is a spherical solution with the form of Equation (20) with $`g=f`$. In order to simplify $`\mathrm{\Psi }_4`$, we will take the limit $`\sigma ^2>>d`$ and $`(y^0\pm y)\sigma >>1`$ in Equation (24) yielding
$$f(z)\frac{16\pi d}{\sigma ^2z^3}+i2\pi \left(\frac{2}{z^2}+\sigma ^2\right).$$
(25)
## IV The quantum bohmian trajectories and results
In this section, we will apply the rules of the Bohm-de Broglie interpretation to the wave functions we have obtained in the previous section. We first summarize these rules for the case of homogeneous minisuperspace models. In this case, the general minisuperspace Wheeler-DeWitt equation is
$$[\widehat{p}_\alpha (t),\widehat{q}^\alpha (t)]\mathrm{\Psi }(q)=0,$$
(26)
where $`p_\alpha (t)`$ and $`q^\alpha (t)`$ represent the homogeneous degrees of freedom coming from the gravitational and matter fields. Writing $`\mathrm{\Psi }=R\mathrm{exp}(iS/\mathrm{})`$, and substituting it into (26), we obtain the equation
$$\frac{1}{2}f^{\alpha \beta }(q^\mu )\frac{S}{q^\alpha }\frac{S}{q^\beta }+U(q^\mu )+Q(q^\mu )=0,$$
(27)
where
$$Q(q^\mu )=\frac{1}{2R}f^{\alpha \beta }\frac{^2R}{q^\alpha q^\beta }.$$
(28)
The quantities $`f^{\alpha \beta }(q^\mu )`$ and $`U(q^\mu )`$ are the minisuperspace particularizations of the DeWitt metric , and of the scalar curvature density of the space-like hypersurfaces together with matter potential energies, respectively. The causal interpretation applied to quantum cosmology states that the trajectories $`q^\alpha (t)`$ are real, independently of any observations. Equation (27) is the Hamilton-Jacobi equation for them, which is the classical one amended with the quantum potential term (28), responsible for the quantum effects. This suggests to define
$$p_\alpha =\frac{S}{q^\alpha },$$
(29)
where the momenta are related to the velocities in the usual way
$$p_\alpha =f_{\alpha \beta }\frac{1}{N}\frac{q^\beta }{t},$$
(30)
and $`f_{\alpha \beta }`$ is the inverse of $`f^{\alpha \beta }`$. To obtain the quantum trajectories, we have to solve the following system of first order differential equations, called the guidance relations:
$$\frac{S(q^\gamma )}{q^\alpha }=f_{\alpha \beta }\frac{1}{N}\frac{q^\beta }{t}.$$
(31)
Equations (31) are invariant under time reparametrization. Hence, even at the quantum level, different choices of $`N(t)`$ yield the same space-time geometry for a given non-classical solution $`q^\alpha (t)`$. There is no problem of time in the causal interpretation of minisuperspace quantum cosmology<sup>*</sup><sup>*</sup>*This is not the case, however, for the full superspace (see Reference )..
For the minisuperspace we are investigating, the guidance relations in the gauge $`N=12\mathrm{exp}(3y_0)`$ are (see Equations (8))
$$p_\mu =\frac{S}{y^\mu }=\eta _{\mu \nu }\dot{y}^\nu ,$$
(32)
where $`S`$ is the phase of the wave function.
### A The empty model
Taking the phase of the solution (17), and inserting it into Equations (32), remembering that $`y^2=\beta _{}`$ and $`y^3=\varphi `$ have been suppressed, we get the following system of planar equations:
$$\dot{\beta }_0=\frac{\beta _+\sigma ^2\mathrm{sin}(2d\beta _0)+2d\mathrm{sinh}(\beta _0\beta _+\sigma ^2)}{2[\mathrm{cos}(2d\beta _0)+\mathrm{cosh}(\beta _0\beta _+\sigma ^2)]},$$
(33)
$$\dot{\beta }_+=\frac{2d\mathrm{cos}(2d\beta _0)+2d\mathrm{cosh}(\beta _0\beta _+\sigma ^2)\beta _0\sigma ^2\mathrm{sin}(2d\beta _0)}{2[\mathrm{cos}(2d\beta _0)+\mathrm{cosh}(\beta _0\beta _+\sigma ^2)]}.$$
(34)
The line $`\beta _0=0`$ divides configuration space in two symmetric regions. The line $`\beta _+=0`$ contains all singular points of this system, which are nodes and centers. The nodes appear when the denominator of the above equations, which is proportional to the absolute value of the wave function, is zero. No trajectory can pass through them. They happen when $`\beta _+=0`$ and $`\mathrm{cos}(2d\beta _0)=1`$, or $`\beta _0=(2n+1)\pi /2d`$, $`n`$ an integer, with separation $`\pi /d`$. The center points appear when the numerators are zero. They are given by $`\beta _+=0`$ and $`\beta _0=2d[\mathrm{cotan}(d\beta _0)]/\sigma ^2`$. They are intercalated with the node points. As $`\beta _0\mathrm{}`$ these points tend to $`n\pi /d`$, and their separations cannot exceed $`\pi /d`$. As one can see from Equations (33) and (34), the classical solutions (10) are recovered when $`\beta _0\mathrm{}`$ and $`\beta _+\mathrm{}`$, the $`\beta `$’s become proportional to $`t`$. The quantum potential given in Equation (28), which now reads,
$`Q(\beta _0,\beta _+)`$ $`=`$ $`{\displaystyle \frac{1}{2R}}\left({\displaystyle \frac{^2R}{\beta _0^2}}+{\displaystyle \frac{^2R}{\beta _+^2}}\right)`$ (35)
$`=`$ $`{\displaystyle \frac{[\beta _+\sigma ^2\mathrm{sin}(2d\beta _0)+2d\mathrm{sinh}(\beta _0\beta _+\sigma ^2)]^2[\beta _0\sigma ^2\mathrm{sin}(2d\beta _0)2d\mathrm{cos}(2d\beta _0)2d\mathrm{cosh}(\beta _0\beta _+\sigma ^2)]^2}{8[\mathrm{cos}(2d\beta _0)+\mathrm{cosh}(\beta _0\beta _+\sigma ^2)]^2}},`$ (36)
becomes negligible in these limits.
A field plot of this planar system is shown in Figure 1 for $`\sigma =d=1`$. We can see plenty of different possibilities, depending on the initial conditions. Near the center points we can have oscillating universes without singularities and with periodic isotropic phases, with amplitude of oscillation of order $`1`$. For negative values of $`\beta _0`$, the universe arises classically from a singularity but quantum effects become important forcing it to recollapse to another singularity, recovering classical behaviour near it. Isotropic phases may happen near their maximum size. For positive values of $`\beta _0`$, the universe contracts classically but when $`\beta _0`$ is small enough quantum effects become important creating an inflationary phase which avoids the singularity. The universe contracts to a minimum size and after reaching this point it expands forever, recovering the classical limit when $`\beta _0`$ becomes sufficiently large. In this case isotropic phases may happen when the universe is near its minimum size. In both cases we see that isotropic phases happen only when quantum effects are important. We can see that, for $`\beta _0`$ negative, we have classical limit for small scale factor while for $`\beta _0`$ positive we have classical limit for big scale factor. In these models with $`d=1`$, the isotropic phases last very shortly. Only for $`d<<1`$ can the isotropic phases be arbitrarily large because, as said above, the separation of the singular points goes like $`\pi /d`$.
For the wave function $`\mathrm{\Psi }_2`$, the analysis goes in the same way but we have to interchange $`\beta _0`$ with $`\beta _+`$ in Figure 1. In this case, we have also periodic solutions but the others are anisotropic universes arising classically from a singularity, experiencing quantum effects in the middle of their expansion when they become approximately isotropic, and recovering their classical behaviour for large values of $`\beta _0`$. Depending on the initial conditions, the isotropic phase can be arbitrarily large. There are no further possibilities.
### B The scalar field model
Let us study the spherical wave solutions (20) of Equation (14). The guidance relations (32) are
$$p_0=_0S=\mathrm{}\left(\frac{_0\mathrm{\Psi }_3}{\mathrm{\Psi }_3}\right)=\dot{y}^0,$$
(37)
$$p_i=_iS=\mathrm{}\left(\frac{_i\mathrm{\Psi }_3}{\mathrm{\Psi }_3}\right)=\dot{y}^i,$$
(38)
where $`S`$ is the phase of the wave function. In terms of $`f`$ and $`g`$ the above equations read
$$\dot{y}^0=\mathrm{}\left(\frac{f^{}(y^0+y)+g^{}(y^0y)}{f(y^0+y)+g(y^0y)}\right),$$
(39)
$$\dot{y}^i=\frac{y^i}{y}\mathrm{}\left(\frac{f^{}(y^0+y)g^{}(y^0y)}{f(y^0+y)+g(y^0y)}\right),$$
(40)
where the prime means derivative with respect to the argument of the functions $`f`$ and $`g`$, and $`\mathrm{}(z)`$ is the imaginary part of the complex number $`z`$.
From Equations (40) we obtain that
$$\frac{dy^i}{dy^j}=\frac{y^i}{y^j},$$
(41)
which implies that $`y^i(t)=c_j^iy^j(t)`$, with no sum in $`j`$, where the $`c_j^i`$ are real constants, $`c_j^i=1/c_i^j`$, and $`c_1^1=c_2^2=c_3^3=1`$. Hence, apart some positive multiplicative constant, knowing about one of the $`y^i`$ means knowing about all $`y^i`$. Consequently, we can reduce the four equations (39) and (40) to a planar system by writing $`y=C|y^3|`$, with $`C>1`$, and working only with $`y^0`$ and $`y^3`$, say. The planar system now reads
$$\dot{y}^0=\mathrm{}\left(\frac{f^{}(y^0+C|y^3|)+g^{}(y^0C|y^3|)}{f(y^0+C|y^3|)+g(y^0C|y^3|)}\right),$$
(42)
$$\dot{y}^3=\frac{\mathrm{sign}(y^3)}{C}\mathrm{}\left(\frac{f^{}(y^0+C|y^3|)g^{}(y^0C|y^3|)}{f(y^0+C|y^3|)+g(y^0C|y^3|)}\right).$$
(43)
Note that if $`f=g`$, $`y^3`$ stabilizes at $`y^3=0`$ because $`\dot{y}^3`$ as well as all other time derivatives of $`y^3`$ are zero at this line. As $`y^i(t)=c_j^iy^j(t)`$, all $`y^i(t)`$ become zero, and the cosmological model isotropizes forever once $`y^3`$ reaches this line. Of course one can find solutions where $`y^3`$ never reaches this line, but in this case there must be some region where $`\dot{y}^3=0`$, which implies $`\dot{y}^i=0`$, and this is an isotropic region. Consequently, quantum anisotropic cosmological models with $`f=g`$ always have an isotropic phase, which can become permanent in many cases.
As a concrete example, let us take the gaussian $`\mathrm{\Psi }_4`$ given in Equation (24). It is a spherical wave solution of the Wheeler-DeWitt equation (14) with $`f=g`$, and hence it does not necessarily have isotropic phases as described above for the case $`f=g`$. Figure 2 shows a field plot of this system for $`d/\sigma ^2=10^4`$ and $`C=2`$. In the region $`|\varphi |>>|\beta _0|`$ we have periods of isotropic expansion because $`\varphi =\mathrm{const}.`$ implies $`\beta _\pm =\mathrm{const}.`$. Depending on the value of $`|\varphi |`$, this isotropic phase can be arbitrarily long. These trajectories are periodic and without singularities. Due to quantum effects, no trajectory crosses the point $`\beta _0=\varphi =0`$. Figure 3 shows the bohmian trajectories for the wave function given in Equation (24), now with $`d/\sigma ^2=10^4`$ and $`C=2`$. They have some qualitative behaviours different from the precedent case. The isotropic phases, now in contraction, are no longer periodic. They are parts of universes that contract anisotropically from infinity, experience a period of isotropy, and contract anisotropically to a singularity. Figure 3 shows that no trajectories crosses the point $`\beta _0=\varphi =0`$.
In order to get some analytical insight over Figures 2 and 3, we present the planar system obtained from the guidance relations corresponding to the wave function (24) in the approximation (25):
$$\dot{\beta }_0=\frac{\frac{2d}{\sigma ^2}(3\beta _0^4+6C^2\beta _0^2\varphi ^2C^4\varphi ^4)}{\beta _0^2(\beta _0^2C^2\varphi ^2)^2+\frac{4d^2}{\sigma ^4}(3\beta _0^2+C^2\varphi ^2)^2},$$
(44)
$$\dot{\varphi }=\frac{\frac{16d}{\sigma ^2}\beta _0^3\varphi }{\beta _0^2(\beta _0^2C^2\varphi ^2)^2+\frac{4d^2}{\sigma ^4}(3\beta _0^2+C^2\varphi ^2)^2},$$
(45)
where we have reset $`y^0=\beta _0`$ and $`y^3=\varphi `$. This approximation is not reliable in the lines $`\beta _0=\pm C\varphi `$. As one can see immediately from these equations, $`\beta _\pm =\mathrm{const}.`$ whenever $`|\varphi |>>|\beta _0|`$, and the sign of $`d`$ defines the trajectories direction.
## V Conclusion
Adopting the Bohm-de Broglie interpretation of quantum cosmology, we have shown that quantum effects can generate an efficient alternative mechanism for the isotropization of cosmological models. Anisotropic classical models which never isotropize may present arbitrarily large isotropic phases during the course of their evolutions if quantum effects are taken into account, without needing to introduce any classical inflationary phase. The models studied were Bianchi I models in empty space or filled with a free massless scalar field.
There are questions and developments which should be undertaken within this approach: the dependence of the above results on boundary conditions of the Wheeler-DeWitt equation, their generalizations to other Bianchi models, and the conditions for which quantum effects can also induce homogeneity on cosmological models which are classically inhomogeneous. The last issue is by far the most interesting but also the most complicated one because the implementation of the Bohm-de Broglie interpretation for inhomogeneous cosmological models is much more subtle and involved . These questions will be the subject of our future investigations.
## ACKNOWLEDGMENTS
We would like to thank the Cosmology Group of CBPF for useful discussions, and CNPq and CAPES of Brazil for financial support. One of us (NPN) would like to thank the Laboratoire de Gravitation et Cosmologie Relativistes of Université Pierre et Marie Curie, where part of this work has been done, for hospitality. |
warning/0001/cond-mat0001402.html | ar5iv | text | # Submonolayer epitaxy with impurities
## I Introduction
Recent progress in the fabrication of atomically smooth interfaces by Molecular Beam Epitaxy (MBE) has lead to an increasing appreciation of the dramatic, detrimental or beneficial, effects that small amounts of impurities may have on the morphology of growing films. Adsorbates acting as surfactants can stabilize layer-by-layer growth of metal and semiconductor surfaces. On the other hand, for the simple case of Pt(111) homoepitaxy it was recently shown that minute coverages of CO strongly increase the step edge barriers for interlayer transport, thus enhancing three-dimensional mound growth. The effect of additional surface species on growth and nucleation is of obvious importance also in more complex, technologically relevant deposition techniques such as chemical vapor deposition. In either case the detailed atomistic kinetics and energetics of the interaction between adsorbate and deposited material influence the growth mode to a degree which makes it very difficult to formulate general rules for large classes of growth systems.
As a first step towards an improved understanding of the generic effects of impurities on epitaxial growth, in the present paper we introduce a minimal model which, we hope, is simple enough to extract some insights of fairly general validity, and yet possesses sufficient flexibility to include most physically relevant microscopic features. The model is based on the standard solid-on-solid description for the growth of a simple cubic crystal. The impurities are represented by a second particle species, which can be codeposited with the growing material or predeposited prior to growth. Impurities diffuse and interact attractively with the deposit atoms (adatoms), but they do not attract each other and hence do not nucleate islands. The details of the model are described in Section II. A brief account of some preliminary results was given in an earlier communication.
Despite its simplicity, the model contains a large number of parameters: impurity and growth fluxes, substrate temperature, and energy barriers for half a dozen kinetic processes. To focus our efforts, we concentrate on modeling a situation in which the impurities decorate the island edges, forming a monatomic chain along the island perimeter. Preferential adsorption of impurities at step edges is suggested by bond counting arguments, and has often been invoked to explain the strong effect of submonolayer adsorbate coverages on growth behavior, e.g., through a change of the barrier for interlayer transport.
It will be shown in Section III that the growth of decorated islands requires, in addition to a suitable choice of binding energies, the possibility of impurity-adatom exchange. Such a process, which is a two-dimensional analog of the exchange mechanism responsible for the floating of surfactants in multilayer growth, was recently demonstrated to play a crucial role in the submonolayer homoepitaxy of Si(001) in the presence of hydrogen.
In Section IV the influence of the adsorbates on the island density is investigated, both for codeposited and predeposited impurities. In the absence of impurities the scaling relation
$$N(F/D)^\chi $$
(1)
between island density $`N`$, deposition flux $`F`$ and adatom diffusion coefficient $`D`$ is well established theoretically, numerically and experimentally. In two dimensions, rate equation analysis yields the expression
$$\chi =\frac{i^{}}{i^{}+2}$$
(2)
for the exponent $`\chi `$ in terms of the size $`i^{}`$ of the largest unstable cluster.
There are several conceivable mechanisms by which attractive impurities could alter the relationship (1). First, impurities may act as nucleation centers, thus effectively decreasing $`i^{}`$ and therefore $`\chi `$; in the extreme case of immobile adatom traps the limit of spontaneous nucleation with $`i^{}=\chi =0`$ would be realized. Second, impurities decorating the island edges may induce energy barriers to attachment. Kandel has shown that, provided these barriers are sufficiently strong, the exponent $`\chi `$ in (1) is increased such that (2) is replaced by $`\chi =2i^{}/(i^{}+3)`$. Both mechanisms imply an increase of the island density compared to the case of pure homoepitaxy.
Our simulations indicate that none of these two mechanisms are operative under the conditions used in our model: The addition of impurities is found to increase the island density in all cases, but the scaling of $`N`$ with the flux $`F`$ remains unaffected within the accuracy of the simulation. An analysis of the relevant microscopic processes reveals that, within our model, even completely decorated island edges do not provide efficient barriers to attachment, and therefore the scenario of Kandel does not apply.
In view of (1), an increase of the island density at fixed $`\chi `$ suggests that the main effect of the impurities is to reduce the mobility $`D`$ of adatoms. The reduction of the adatom mobility has been identified as the most important mechanism contributing to the surfactant action of Sb on Ag. Decreasing $`D`$ reduces the island size and favors the growth of ramified, rather than compact islands. Both effects enhance interlayer transport, since the adatoms landing in the second layer have more opportunities to descend, and thus promote layer-by-layer growth. In our model the island size decreases but the island shapes remain compact, because the edge decoration facilitates edge diffusion (see Section III).
For an analytic description of the relation between impurity coverage, adatom mobility and island density, in Section V we develop a simple rate equation approach which provides a semi-quantitative explanation for many (though not all) features observed in the simulations. Some conclusions and open questions are formulated in Section VI.
## II Model
The growth model employed in this work has been briefly described in an earlier paper. It is a solid-on-solid model with two surface species $`A`$ and $`B`$, where $`A`$-particles correspond to the growing material, and $`B`$-particles represent the impurities. The simulation starts on a flat substrate composed only of $`A`$-atoms. The basic microscopic processes are deposition and migration; desorption is not allowed. Two deposition modes are considered: (i) simultaneous deposition (codeposition) of both species and (ii) predeposition of a certain impurity coverage prior to growth. In the case of codeposition the fluxes $`F_A`$ and $`F_B`$ of the two species may differ.
The migration of a surface atom is modeled as a nearest–neighbor hopping process with the rate $`R_D=k_0\mathrm{exp}(E_D/k_BT)`$, where $`k_0=10^{13}`$ Hz is an adatom vibration frequency, $`E_D`$ is the hopping barrier, $`T`$ is the substrate temperature and $`k_B`$ is Boltzmann’s constant. The hopping barrier is the sum of a term from the substrate $`E_{\mathrm{sub}}`$ and a contribution from each lateral nearest neighbor $`E_\mathrm{n}`$. Both contributions depend on local composition: For each term we have the four possibilities $`AA`$, $`AB`$, $`BA`$ and $`BB`$. The hopping barrier of an atom $`X`$ (of type $`A`$ or $`B`$) is then
$$E_D^X=\underset{Y=A,B}{}\left(n_0^YE_{\mathrm{sub}}^{XY}+n_1^{XY}E_\mathrm{n}^{XY}\right),$$
(3)
where $`E_{\mathrm{sub}}^{XY}`$ is the hopping barrier for a free $`X`$ adatom on a substrate atom $`Y`$, $`n_0^Y`$ is equal to one if a substrate atom is of type $`Y`$ and zero otherwise, $`n_1^{XY}`$ is the number of nearest-neighbor $`X`$-$`Y`$ pairs, and $`E_\mathrm{n}^{XY}`$ is the corresponding contribution to the barrier (symmetric in $`X`$ and $`Y`$). Lateral interactions between impurity atoms are neglected ($`E_\mathrm{n}^{BB}=0`$).
In the simulations reported in this paper we used $`E_{\mathrm{sub}}^{AA}=0.8`$ eV, $`E_{\mathrm{sub}}^{AB}=0.1`$ eV, $`E_{\mathrm{sub}}^{BA}=1.0`$ eV, $`E_{\mathrm{sub}}^{BB}=0.1`$ eV, and the substrate temperature $`T=500`$ K. The low values of $`E_{\mathrm{sub}}^{AB}`$ and $`E_{\mathrm{sub}}^{BB}`$ ensure that atoms deposited on top of an impurity instantaneously descend to the substrate. Growth and impurity fluxes were varied in the interval ranging from 0.00025 ML/s to 0.25 ML/s. The system sizes ranged from 300$`\times `$300 to 500$`\times `$500. The nearest-neighbor coupling $`E_\mathrm{n}^{AA}`$ between $`A`$-atoms controls the size of the critical nucleus $`i^{}`$. It may be stronger ($`E_\mathrm{n}^{AA}>E_\mathrm{n}^{AB}`$) or weaker ($`E_\mathrm{n}^{AA}<E_\mathrm{n}^{AB}`$) than the coupling to the impurities. In equilibrium at low temperatures, the former case leads to the formation of islands composed inside mainly of $`A`$ atoms with $`B`$ atoms bounded near the edges, while in the latter case it is energetically more favorable when $`B`$ atoms are inside the islands.
However, our simulations show that growth leads to intermixing of $`A`$ and $`B`$ atoms in both cases, $`E_\mathrm{n}^{AA}>E_\mathrm{n}^{AB}`$ and $`E_\mathrm{n}^{AA}<E_\mathrm{n}^{AB}`$. Thus the energetic bias favoring segregation is not sufficient to obtain configurations with impurities mostly at island edges (decorated islands). To achieve this, we have to introduce an additional thermally activated process, which allows an $`A`$ atom approaching an island to exchange with an impurity covering the island edge. A similar process was introduced previously in the context of homoepitaxy on Si(001) with predeposited hydrogen. In that work, an $`A`$-atom was allowed to exchange with an impurity provided the $`A`$-atom was not bonded to another $`A`$-atom at a nearest-neighbor site. In our case this modification turned out not to be sufficient, since impurities were still found to be progressively trapped inside islands during growth. We therefore allow the exchange of an $`A`$ atom with an impurity also when it has a single bond to another $`A`$ atom in a nearest-neighbor position. Using this rule, which is analogous to the exchange process invoked in the case of three-dimensional growth with surfactants, we obtain well decorated islands with impurities floating on the island edges during growth (Fig. 1b, and c), see Section III.
In principle, the rate of the exchange process could depend on the number of nearest-neighbor bonds (zero or one) of the $`A`$-atom. We observed that the difference of both rates is not crucial provided both processes are active. The rates of these processes are taken as $`k_{\mathrm{ex}}=k_0\mathrm{exp}(E_{\mathrm{ex}}/k_BT)`$, where $`E_{\mathrm{ex}}`$ are the corresponding activation barriers. For both processes there is a maximum activation barrier above which the decorated geometry is not observed. In the following, the barriers for both types of exchange are for simplicity set to be equal.
## III Island morphology
In Fig. 1 we show examples of typical configurations with the same partial coverage of both species $`\theta _A=\theta _B=0.1`$ ML (i.e., the total coverage $`\theta =\theta _A+\theta _B=0.2`$ ML) obtained by codeposition of adatoms and impurities with fluxes $`F_A=F_B=0.004`$ ML/s. Figures 1a, 1b, 1c illustrate the effect of varying the relation between $`E_\mathrm{n}^{AA}`$ and $`E_\mathrm{n}^{AB}`$ for $`E_\mathrm{n}^{AA}=0.3`$ eV. Several features can be identified. First, the island density increases with $`E_\mathrm{n}^{AB}`$. This can be explained by the observation (cf. Fig. 1c for $`E_\mathrm{n}^{AB}=0.4`$ eV) that for larger $`E_\mathrm{n}^{AB}`$, free $`A`$-atoms start to be captured by impurities and many small islands containing impurities and a few $`A`$ atoms appear on the surface in addition to already existing decorated islands. These small islands act as nucleation centers that lead to the increase of the island density. For large $`E_\mathrm{n}^{AB}`$, almost all impurities will capture an $`A`$ adatom for a certain time. As we argue in Section V, this effect causes reduction of adatom mobility by the impurities.
Whereas the island density increases, the density of free impurities decreases with increasing $`E_\mathrm{n}^{AB}`$ due to (i) the stronger $`A`$-$`B`$ bond favoring the binding of impurities at island edges and (ii) the increase of the island density that leads to smaller islands with more perimeter sites. (Notice that we do not obtain decorated islands for simultaneous deposition at very large fluxes or at very early stages of growth since there are not enough impurities available to cover all perimeter sites.)
The degree of edge decoration also strongly depends on the value of $`E_\mathrm{n}^{AB}`$. Edge decoration is not observed for small $`E_\mathrm{n}^{AB}=0.1`$ eV, it is only partial for $`E_\mathrm{n}^{AB}=0.2`$ eV, and it becomes perfect for $`E_\mathrm{n}^{AB}=0.4`$ eV (cf. Figure 1). Hence, in order to get decorated islands, the barrier $`E_\mathrm{n}^{AB}`$ has to be larger than a minimal value. A simple detailed balance argument shows that the fraction $`f_0`$ of uncovered edge sites is given by
$$f_0=(1+\theta _Be^{E_\mathrm{n}^{AB}/k_BT})^1,$$
(4)
and thus at $`T=500`$ K a barrier of $`E_\mathrm{n}^{AB}0.2`$ eV is required. As we shall see in Section V, the condition $`f_01`$ also implies that the diffusion of $`A`$-atoms is slowed down considerably by the impurities.
Another important parameter determining decoration of island edges is the exchange barrier $`E_{\mathrm{ex}}`$ that was in our simulations varied from 0.8 eV to 2 eV. For a small value of $`E_{\mathrm{ex}}`$, impurities are driven toward island edges, whereas for large $`E_{\mathrm{ex}}`$, the exchange process is not active and impurities are often incorporated inside the islands. The impurities were observed to be floating on the island edges for small $`E_{\mathrm{ex}}`$, both for $`E_\mathrm{n}^{AA}>E_\mathrm{n}^{AB}`$ (Fig. 1b) and for $`E_\mathrm{n}^{AA}<E_\mathrm{n}^{AB}`$ (Fig. 1c). Fig. 1d shows a configuration for $`E_{\mathrm{ex}}=2`$ eV in the case $`E_\mathrm{n}^{AA}>E_\mathrm{n}^{AB}`$. The surface morphology is similar to the one observed for $`E_\mathrm{n}^{AA}<E_\mathrm{n}^{AB}`$ (not shown), where quite a regular checkerboard
structure is produced with almost no free impurities on the surface. Thus in order to obtain decorated islands, $`E_{\mathrm{ex}}`$ has to be lower than a threshold value which in the present case is about 1.2 eV.
A further remarkable feature of the configurations displayed in Figure 1(b) and (c) is that the compact square island shape is maintained as the island density increases. In fact, careful inspection shows that the kink density on the well decorated edges is smaller than when the decoration is incomplete. This reflects the enhancement of edge diffusion by the impurities: The energy barrier for an adatom moving along the decorated step edge within the impurity layer is $`E_{\mathrm{ex}}`$ which, under the conditions of Figure 1(b),(c), is smaller than both the barrier $`E_{\mathrm{sub}}^{AA}+E_\mathrm{n}^{AA}`$ for diffusion along an uncovered step and the barrier $`E_{\mathrm{sub}}^{AA}+E_\mathrm{n}^{AB}`$ for edge diffusion on the outside of the impurity layer. Clearly this is true only if the exchange of singly bonded $`A`$-atoms is allowed, which underlines the importance of this type of the exchange process.
## IV Island density scaling
### A Simultaneous growth
In the previous section, we described qualitatively the growth morphology for one fixed value of the deposition flux. Here we present results for the behavior of the island density $`N`$ as a function of flux $`F`$ and coverage $`\theta `$, and discuss its dependence on the kinetic parameters. Results for each set of parameters were obtained by averaging over several independent simulation runs.
In the presence of impurities, islands are composed of both $`A`$ and $`B`$ atoms. We define the size of an island as the number of $`A`$-atoms in a connected cluster of $`A`$-atoms forming the island. This definition is appropriate for growth with impurities segregating on the edges of the islands. However, visual inspection of configurations showed that for $`E_{\mathrm{ex}}>1.2`$, there exist islands containing several mutually disconnected clusters of $`A`$-atoms. Hence, our definition cannot be applied straightforwardly for large $`E_{\mathrm{ex}}`$. In the following, we restrict ourselves to situations where the intermixing inside the islands is negligible. Simulations for larger values of $`E_{\mathrm{ex}}`$ indicate that the flux dependence of the island density flattens (the exponent $`\chi `$ in (1) decreases), but due to the ambiguity in the definition of the island density in presence of intermixing, we did not attempt to assess the physical significance of this observation.
#### 1 Flux dependence
Fig. 2 shows the island density $`N`$ as a function of the adatom flux $`F_A`$ for several coverages $`\theta `$ and different energy barriers $`E_\mathrm{n}^{AB}`$ and $`E_{\mathrm{ex}}`$ (inset). The energy barrier $`E_\mathrm{n}^{AA}=0.3`$ eV is fixed, and the impurities and adatoms are codeposited with the same flux $`F_B=F_A`$. For comparison we also show data for homoepitaxial growth without impurities at two coverages $`\theta =0.05`$ ML and $`\theta =0.1`$ ML. We can see that for $`E_\mathrm{n}^{AB}=0.2`$ eV and $`E_{\mathrm{ex}}=1`$ eV, the island density is quite close to the corresponding value in homoepitaxy. With increasing interaction energy between adatoms and impurities, the island density dramatically increases, but the exponent $`\chi `$ in the power law relation (1) between flux and the island density remains nearly unchanged. For example, we find $`\chi 0.54`$ for $`E_\mathrm{n}^{AB}=0.2`$, $`\chi 0.45`$ for $`E_\mathrm{n}^{AB}=0.4`$, and $`\chi 0.54`$ for homoepitaxial growth, which means that the effective critical nucleus size is $`i^{}2`$ in this range of parameters. According to Kandel’s rate equation theory, the scaling exponent should then become $`\chi 0.8`$ in the presence of strong barriers to attachment. In our model this is not observed, because the bonding of the adatoms to the impurity-covered edges keeps them near the edge long enough for an exchange to occur.
The inset of Fig. 2 shows that the island density is further increased if the exchange barrier $`E_{\mathrm{ex}}`$ is set to a larger value $`E_{\mathrm{ex}}=1.2`$ eV. The data for $`E_{\mathrm{ex}}=1.2`$ eV and $`E_\mathrm{n}^{AB}=0.4`$ indicate a slight decrease of the exponent $`\chi `$.
Figure 3 shows results obtained by varying the ratio $`F_B/F_A`$ of impurity to adatom flux. In one set of simulations, using $`E_\mathrm{n}^{AB}=0.2`$ eV, the impurity flux was kept constant at $`F_B=0.016`$ ML/s, while the adatom flux $`F_A`$ was varied. For large fluxes $`F_A>F_B`$, the island density
is seen to approach the data obtained for homoepitaxy, indicating that the impurities have no effect, while for small fluxes $`F_A<F_B`$, the flux dependence is described by an effective power law $`NF_A^\chi ^{}`$ with $`\chi ^{}0.36<\chi `$. An interpretation of this behavior will be given at the end of Section V B 2
In the second set of simulations shown in Figure 3, which were carried out using $`E_\mathrm{n}^{AB}=0.4`$ eV, both fluxes were varied keeping the ratio $`F_B/F_A=2`$ constant. This is seen to further increase the island density without changing the flux dependence. In this sense, an increase in the coverage of impurities (by increasing $`F_B`$) is equivalent to increasing their effectiveness through an increase of the bond energy $`E_n^{AB}`$. A quantitative formulation of this statement will be given in Section V.
#### 2 Coverage dependence
A new feature in comparison with homoepitaxy is a stronger coverage dependence of the island density. This is seen in Fig. 2 for both weak ($`E_\mathrm{n}^{AB}=0.2`$ eV) and strong ($`E_\mathrm{n}^{AB}=0.4`$ eV) interaction with impurities, but it is more pronounced for strong interaction, in particular at larger fluxes. We followed the coverage dependence in more detail for fixed flux, and in addition to the island density we also measured the density of free adatoms $`n`$. The results obtained at a medium flux $`F_A=0.004`$ ML/s are compared with homoepitaxy in Fig. 4. Both in impure growth and in homoepitaxy the island density shows an initial regime of rapid increase followed by a “saturation” regime in which it increases much more slowly with coverage. However, the residual coverage dependence in
the saturation regime is stronger in the presence of impurities, and furthermore the onset of the saturation regime is delayed as the interaction between adatoms and impurities increases. A quantitative description of this effect will be provided in Section V B 2.
The density of adatoms exhibits a completely different behavior as compared to homoepitaxy. We observe that for growth with impurities, the adatom density is comparable with the island density up to the coverage $`\theta _A=0.1`$ ML, whereas in homoepitaxy the adatom density rapidly decreases after reaching a maximum at the beginning of the saturation regime (cf. Section V A). Other surprising features are the power-law increase $`n\theta _A^{0.75}`$ observed over almost two decades in the case of strongly interacting impurities, and the weak oscillations of the adatom density for coverages $`\theta _A>0.01`$ (cf. Fig. 4). We will return to the behavior of the adatom density in Section V B 2.
#### 3 Next-nearest-neighbor interaction
A modification of the model in which the barrier for diffusion $`E_B^X`$ contains an additional contribution from each lateral next-nearest neighbor of the opposite type was also studied. This implies that a term $`n_2^{AB}E_{\mathrm{nn}}^{AB}`$ is added to the right hand side of Eq. (3). Here $`n_2^{AB}`$ is the number of next-nearest neighbors of the type opposite to the atom under consideration, and $`E_{\mathrm{nn}}^{AB}`$ is the corresponding contribution to the activation barrier. We do not consider next-nearest-neighbor contributions from pairs of particles of the same type. For simplicity, the new parameter $`E_{\mathrm{nn}}^{AB}`$ is set equal to $`E_\mathrm{n}^{AB}`$.
Our motivation for introducing the additional term is a desire to study an improvement in the decoration of island edges by impurities, and the resulting decrease of the fraction $`f_0`$ of uncovered edge sites. The additional interaction also enhances the probability of nucleation around impurities because the number of sites at which an adatom can be captured is considerably higher. The configuration shown in Fig. 5a demonstrates that now we obtain almost perfect decoration also for $`E_\mathrm{n}^{AB}=E_{\mathrm{nn}}^{AB}=0.1`$ eV. Fig. 5b illustrates the nucleation of small islands.
For $`E_\mathrm{n}^{AB}=E_{\mathrm{nn}}^{AB}=0.1`$ eV, the island density is nearly the same as for homoepitaxy for all fluxes studied, and decoration is perfect provided there is sufficient amount of impurities available. This shows that also in the presence of next-nearest-neighbor interactions the decorated
edges are unable to block efficiently the attachment of adatoms. For larger values of $`E_\mathrm{n}^{AB}=E_{\mathrm{nn}}^{AB}`$, the additional interaction causes an increase of the island density and a decrease of the scaling exponent $`\chi `$. For example, the effective value for $`E_\mathrm{n}^{AB}=E_{\mathrm{nn}}^{AB}=0.2`$ eV is $`\chi 0.42`$ and for $`E_\mathrm{n}^{AB}=E_{\mathrm{nn}}^{AB}=0.4`$ eV it drops to $`\chi 0.3`$. This suggests that the nucleation of small islands described above effectively lowers the size $`i^{}`$ of the critical nucleus.
### B Predeposition of impurities
We performed simulations with predeposition of impurities for the same set of parameters as for codeposition. In order to obtain a morphology with island edges decorated by impurities, we need an appropriate value of $`E_{\mathrm{ex}}`$. Complete decoration also requires a sufficient amount of impurities available on the surface. The data presented are for a predeposited coverage $`\theta _B=0.1`$ ML. Examples of morphologies for $`E_\mathrm{n}^{AB}=0.2`$ eV and $`E_\mathrm{n}^{AB}=0.4`$ eV are shown in Fig. 6 and look qualitatively similar to Figures 1 b and c.
The $`F_A`$-dependence of the island density is compared with the results for codeposition in Fig. 7. The island densities in the predeposition regime are slightly higher than for codeposition. This is qualitatively plausible, since the predeposited impurities are present on the surface throughout the deposition process and hence their effect on growth accumulates over time. The corresponding curves are shifted by a factor independent of the flux. The slope remains the same as for codeposition. The difference from codeposition is that there is no appreciable
coverage dependence for $`\theta _A`$ larger than $`0.05`$ ML, except at larger fluxes in the case of strongly interacting impurities.
The detailed coverage dependence for a fixed flux is shown in Fig. 8. We see that the behavior of both the island density and the adatom density is qualitatively similar to homoepitaxy. The island density saturates and at the same time the adatom density starts to decrease. The impurities only cause a shift of the crossover to the saturation regime to a higher coverage, the shift being larger for stronger interaction between adatoms and impurities. We shall return to this effect in Section V B 1.
## V Rate equation theory
In this section, we develop a simple rate equation approach to explain, at least qualitatively, the main impurity effects on the island density which were presented in Section IV. Our basic assumption is that the impurities affect the growth process only by slowing down the diffusion of adatoms. To obtain a simple analytic expression for the effective adatom diffusion coefficient $`\overline{D}(\theta _B)`$ in the presence of an impurity coverage $`\theta _B`$, we further replace the (mobile) impurities by static traps with binding energy $`E_\mathrm{n}^{AB}`$. Then standard results for diffusion in random media yield
$$\overline{D}(\theta _B)=\frac{D}{1\theta _B+\theta _Be^{E_\mathrm{n}^{AB}/k_BT}}\frac{D}{1+\theta _B\varphi }$$
(5)
where $`D=k_0e^{E_{\mathrm{sub}}^{AA}/k_BT}`$ is the diffusion coefficient of a single adatom on the clean substrate, and the abbreviation
$$\varphi =e^{E_\mathrm{n}^{AB}/k_BT}1$$
(6)
has been introduced.
The first conclusion that can be drawn from Eq. (5) is that predeposited impurities, $`\theta _B`$ = const., significantly affect the adatom diffusion only if $`\theta _B1/\varphi `$. In the case of codeposition $`\theta _B=F_Bt`$ and $`\overline{D}`$ becomes time- or coverage-dependent. It is then useful to rewrite (5) in terms of the coverage $`\theta _A`$ of $`A`$-atoms as
$$\overline{D}=\frac{D}{1+\theta _A/\theta ^{}},$$
(7)
with the characteristic coverage
$$\theta ^{}=\frac{F_A}{F_B}\varphi ^1.$$
(8)
For coverages $`\theta _A\theta ^{}`$ the impurities begin to significantly affect the adatom mobility. The expression (8) quantifies the statement made above in Section IV A 1 that an increase of the flux ratio $`F_B/F_A`$ in codeposition is equivalent to an increase of the $`A`$-$`B`$ binding energy $`E_\mathrm{n}^{AB}`$. For $`F_A=F_B`$ and $`T=500`$ K, we have $`\theta ^{}0.01`$ for $`E_\mathrm{n}^{AB}=0.2`$ eV and $`\theta ^{}10^4`$ for $`E_\mathrm{n}^{AB}=0.4`$ eV. In the following, these two sets of parameters will be referred to as the case of weak and strong impurities, respectively.
### A Pure growth
We proceed by combining (5) with the simplest analytic model of nucleation, consisting of two coupled rate equations for the island density $`N`$ and the adatom density $`n`$. In the absence of impurities, the equations for a critical island size $`i^{}=1`$ read
$$\frac{dn}{dt}=F_A4Dn(2n+N)$$
(9)
$$\frac{dN}{dt}=4Dn^2.$$
(10)
The main features of the solution of (9,10) with initial condition $`n=N=0`$ can be described as follows (see the article by Tang for a lucid presentation): In the early time regime the adatom density increases linearly by deposition, $`nF_At=\theta _A`$, and accordingly the island density grows as
$$N(4/3)(D/F_A)\theta _A^3.$$
(11)
In the late time regime the adatoms are mainly captured by preexisting islands. This implies that $`nF_A/DNN`$ and the island density grows more slowly, as
$$N(F_A/12D)^{1/3}\theta _A^{1/3},$$
(12)
while the adatom density decreases as $`n\theta _A^{1/3}`$. The transition between the two regimes occurs at a coverage
$$\theta _1(F_A/D)^{1/2}.$$
(13)
Keeping the coverage fixed while increasing the flux therefore takes the system from the late time regime, where $`NF^{1/3}`$, into the early time regime with $`NF^1`$, with a maximum in the island density attained at a critical flux $`F^cD\theta ^2`$.
To generalize these estimates to the case $`i^{}>1`$, we replace the nucleation equation (10) by
$$\frac{dN}{dt}Dn^{i^{}+1}.$$
(14)
Then the early time behavior becomes $`N(D/F)\theta _A^{i^{}+2}`$, while in the late time regime
$$N(F_A/D)^{i^{}/(i^{}+2)}\theta _A^{1/(i^{}+2)},$$
(15)
in agreement with the expression (2) for the scaling exponent $`\chi `$. The transition coverage $`\theta _1`$ is estimated by matching the two behaviors, which yields
$$\theta _1(F_A/D)^{2/(i^{}+3)}.$$
(16)
For $`i^{}=1`$ the coverage dependence of the densities of islands and adatoms observed in microscopic simulations is in accordance with the rate equation theory. In the reversible case $`i^{}>1`$, the simple rate equations are quantitatively inappropriate, though the key qualitative features – the existence of an early time regime of a rapid increase of the island density, followed by a “precoalescence saturation regime” with little change in $`N`$ – remain.
### B Impure growth
#### 1 Predeposition
The effect of predeposited impurities is obtained simply by replacing $`D`$ by the constant expression (5) for $`\overline{D}`$ in (15) and (16). Consequently the island density in the late time regime increases by a factor $`(1+\theta _B\varphi )^{i^{}/(i^{}+2)}`$ which is independent of flux or $`A`$-coverage, and the onset of saturation is delayed by a factor $`(1+\theta _B\varphi )^{2/(i^{}+3)}`$. This is in qualitative agreement with the coverage dependence of densities of islands and adatoms displayed in Figure 8, which shows the same overall behavior as in the homoepitaxial case, only shifted to larger coverages and higher densities. Quantitatively, the numerically observed increase in the island density is consistent with the factor $`(1+\theta _B\varphi )^{i^{}/(i^{}+2)}`$ if the size of the critical nucleus is set to $`i^{}=1`$. On the other hand, if the critical nucleus size is assumed to be $`i^{}=2`$ as suggested by the numerical value of $`\chi `$, then the theory is seen to overestimate the increase in the island density and underestimate the delay of the onset of saturation.
As was mentioned in Section V A, the island density at fixed coverage $`\theta _A`$ shows a maximum at a critical flux $`F^c`$, which is determined by setting the saturation coverage equal to $`\theta _A`$. For predeposited impurities with $`\theta _B\varphi 1`$ this is given by
$$F_A^c(D/\varphi \theta _B)\theta _A^{(i^{}+3)/2}.$$
(17)
Further discussion of predeposition in relation to codeposition will be provided below.
#### 2 Codeposition
In the case of codeposition the situation is richer due to the coverage dependence of $`\overline{D}`$. First, the cases $`\theta _1<\theta ^{}`$ and $`\theta _1>\theta ^{}`$ have to be distinguished. In the first case the impurities only affect the late time regime. For $`\theta _A\theta ^{}`$ we can approximate (7) by $`\overline{D}D\theta ^{}/\theta _A`$. Inserting this into the nucleation equation (14) and setting $`nF/\overline{D}N`$ we obtain the expression
$$N(F_A/D\theta ^{})^{i^{}/(i^{}+2)}(\theta _A)^{(i^{}+1)/(i^{}+2)}$$
(18)
which replaces (15) in the pure case. It can be seen that the scaling of $`N`$ with flux $`F_A`$ remains the same, i.e. the exponent $`\chi `$ is not affected. The impurities increase the island density by a factor $`(\theta _A/\theta ^{})^{i^{}/(i^{}+2)}`$ which, in contrast to the case of predeposition, is coverage dependent. This is qualitatively consistent with the residual coverage dependence of the island density seen in Figure 2, which is nearly absent in the corresponding predeposition data in Figure 7. In quantitative terms, however, the coverage dependence in (18) is much stronger than what is observed in the simulations. The dependence of (18) on $`\theta ^{}`$ shows that increasing the ratio $`F_B/F_A`$ at constant $`F_A`$ and $`\theta _A`$ will also increase the island density, in accordance with the simulation data shown in Figure 3.
Noting that for codeposition $`\theta _B=(F_B/F_A)\theta _A=\theta _A/(\varphi \theta ^{})`$, Eq. (18) can be rewritten as $`N(F_A\theta _B\varphi /D)^{i^{}/(i^{}+2)}\theta _A^{1/(i^{}+2)}`$, which is identical to the expression for predeposition with $`\theta _B\varphi 1`$. A more careful calculation shows that the island density for predeposition exceeds that for codeposition by a constant factor $`(i^{}+1)^{1/(i+2)}`$, if systems of the same impurity coverage $`\theta _B`$ are compared. This is qualitatively consistent with Figure 7, though the simulation data indicate that the factor is larger for weak impurities than for strong ones.
A surprising consequence of the rate equations with coverage-dependent diffusion is that the adatom density increases with coverage in the late time regime. This follows from balancing the deposition term on the right hand side of (9) against the island capture term $`4\overline{D}Nn4D(\theta ^{}/\theta _A)Nn`$, which yields
$$n(F_A/D\theta ^{})^{2/(i^{}+2)}\theta _A^{1/(i^{}+2)}.$$
(19)
In fact, when $`\theta _1<\theta ^{}`$ the adatom density is a nonmonotonic function of coverage: It increases as $`n\theta _A`$ for $`\theta _A<\theta _1`$, decreases as $`n\theta _A^{1/(i^{}+2)}`$ for $`\theta _1<\theta _A<\theta ^{}`$,
and increases again according to (19) for $`\theta _A>\theta ^{}`$. This is illustrated in Figure 9, which is reminiscent of the simulation data for the adatom density in Figure 4. However, in contrast to the simulations, the late time adatom density given by (19) is small compared to the island density (18), and neither the intermediate scaling regime $`n(\theta _A)^{0.75}`$ nor the oscillations of $`n`$ seen in Figure 4 are reproduced by the rate equations. Here and in the following figures we show results obtained by numerical integration of the rate equations (9,10) for $`i^{}=1`$, with $`D`$ replaced by $`\overline{D}`$. This is sufficient for a qualitative comparison, and relieves us of the necessity to explicitly treat the growth dynamics of the intermediate unstable clusters.
Consider next the case $`\theta _1>\theta ^{}`$. For coverages in the early time regime which satisfy $`\theta _A\theta ^{}`$, we set $`\overline{D}D\theta ^{}/\theta _A`$ and obtain, using $`n\theta _A`$, the early time behavior $`N(D\theta ^{}/F_A)\theta _A^{i^{}+1}`$. The onset of the saturation regime then occurs at a coverage
$$\stackrel{~}{\theta }_1(F_A/D\theta ^{})^{2/(i^{}+1)}$$
(20)
which exceeds the corresponding expression (16) for the pure system by a factor $`(\theta _1/\theta _{})^{2/(i^{}+1)}`$. Thus both predeposited and codeposited impurities delay the onset of the saturation regime.
The critical flux at which the island density attains a maximum is now given by
$$F_B^c(D/\varphi )\theta _A^{(i^{}+1)/2}$$
(21)
which is seen to become identical to the predeposition expression (17) by setting $`\theta _B=(F_B/F_A)\theta _A`$. A quantitative evaluation using the parameters of the simulations
shows that the critical flux is beyond the range of simulated fluxes for the case of weak impurities, while it should be observable for strong impurities. This is illustrated in Figure 10. It therefore appears natural to identify the saturation of the flux dependence of the island density found in the simulations (see Figures 2 and 7) with the maximum predicted by the rate equations, although it should be emphasized that the simulations show no clear evidence of a decrease of $`N`$ beyond the plateau.
Finally, we address the effect of changing the adatom flux $`F_A`$ while keeping the impurity flux $`F_B`$ constant. At fixed $`\theta _A`$ this implies a crossover from $`\theta _A\theta ^{}`$ for $`F_AF_B\varphi \theta _A`$ to $`\theta _A\theta ^{}`$ for $`F_AF_B\varphi \theta _A`$. In the high flux regime the island density $`N`$ is unaffected by the impurities, while in the low flux regime $`N(F_A/D)^\chi (\theta _A/\theta ^{})^\chi (F_B\varphi /D)^\chi `$ becomes independent of $`F_A`$, since $`\theta ^{}F_A`$. This is illustrated in Figure 11, which should be compared to Figure 3. The behavior is qualitatively similar, however instead of a plateau at low fluxes the MC simulations show a second scaling regime $`NF_A^\chi ^{}`$ with a nonzero exponent $`\chi ^{}<\chi `$.
#### 3 The effect of edge decoration
In the formulation of the rate equations we have assumed that all deposited impurities contribute to the impurity coverage in the expression (5) for the effective diffusion coefficient, thus neglecting the fact that a certain fraction of impurities is bound at the island edges. This assumption is self-consistent only if the density $`n_e`$ of edge sites, as predicted by the rate equations, is small compared to the deposited coverage $`\theta _B`$ of impurities at all times.
For compact islands the density of edge sites is of the order of $`N\sqrt{A}`$, where $`A(\theta _An)/N`$ is the area of an island, and hence
$$n_e\sqrt{N(\theta _An)}.$$
(22)
In the saturation regime $`n,N\theta _A`$, thus $`n_e\sqrt{\theta _AN}`$. In the early time regime $`n\theta _A`$ to leading order, and the density of edge sites is determined by the next-to-leading correction. Analysis of the rate equations shows that $`n_eN`$ both for $`\theta _A<\theta ^{}`$ and for $`\theta _A>\theta ^{}`$, which implies that the island size does not increase with coverage in this regime.
Using the estimates for $`n_e`$, the importance of the impurities bound at the edges can be worked out for specific cases. Since evidently $`n_e\theta _A`$ always, a sufficient condition for the irrelevance of edge decoration is that $`\theta _B>\theta _A`$ at all times. This is true for predeposited impurities up to a coverage $`\theta _A=\theta _B`$, and for codeposition with $`F_AF_B`$. Among the situations treated earlier in this section, the only case where corrections due to edge decoration may be expected is deposition at fixed impurity flux $`F_B`$ in the transition region $`F_BF_AF_B\varphi \theta _A`$ (see Figure 11). For fluxes $`F_A>F_B\varphi \theta _A`$ the impurities were seen to be irrelevant even if the full impurity coverage contributes to slowing down the adatoms. Since the edge decoration decreases the impurity concentration on the terraces, it is likely that its only effect will be to shift the point where the island density becomes equal to its homoepitaxial value towards smaller deposition fluxes.
## VI Conclusions
A brief glance at the results presented in this paper may lead to disappointment: No dramatic change of the exponent $`\chi `$ has been found contrary to expectations in any of the modifications of the simulation model. However, upon closer inspection our work reveals several nontrivial features that should not be overlooked in the large amount of numerical data.
(i) We have established that a perfect decoration of island edges in our model requires a process of adatom exchange that is completely analogous to the exchange discussed for surfactant-mediated growth of semiconductors. A feature worth remembering is the enhancement of diffusion of adatoms along island edges via the exchange process with impurities attached to these edges. This mechanism of smoothing on a one-dimensional substrate provides another perspective and possible interpretation of the smooth growth on a two-dimensional substrate in the presence of a surfactant. Smoothing of island shapes should be experimentally verifiable and may have practical implications.
Given the fact that size and shape of islands (including the number of kinks at island edges) in the submonolayer regime of growth determine the developing surface morphology (cf. most of the experimental papers cited below, and in particular Ref. ), a possibility to control both of them by adding impurities seems very attractive.
(ii) From a theoretical perspective, we have been able to obtain insight into (and even semi-quantitative agreement with) the simulation results, using rather simple rate equation theory. The surprising resistance to change of the exponent $`\chi `$ can be understood within this theory, as well as other features of the simulations, such as the strikingly different behavior of the adatom density during impure growth with codeposition as compared to the case of homoepitaxy.
Our research also leads to new questions. One would like to understand better the details of behavior observed, in particular the oscillations of the adatom density seen in simulations (Fig. 8). Another open question concerns “monovalent” impurities that can bond to island edges but do not bond to adatoms approaching these edges, and which therefore should bring about an efficient passivation of the island boundaries. Preliminary simulations suggest that this effect alone does not bring about any significant change in the value of the exponent $`\chi `$.
Acknowledgements. JK acknowledges the kind hospitality of the Institute of Physics, Czech Academy of Sciences, Prague, while part of this paper was prepared. This work was supported by Volkswagenstiftung, by the COST project P3.130 and by DFG within SFB 237. |
warning/0001/hep-th0001212.html | ar5iv | text | # Geometry of Orientifolds with NS-NS 𝐵-flux
## I Introduction and Summary
In recent years six and four dimensional orientifolds have been extensively studied, and much progress has been made in understanding such string compactifications. Various $`𝒩=1`$ supersymmetric six dimensional orientifold vacua were constructed, for instance, in . Generalizations of these constructions to $`𝒩=1`$ supersymmetric four dimensional orientifold vacua have also been discussed, for instance, in .
Most of the aforementioned discussions have been confined to orientifolds with vanishing NS-NS antisymmetric tensor backgrounds. However, generalizations to cases with non-trivial NS-NS $`B`$-flux have also been considered. Thus, in For a more recent discussion of toroidal Type I compactifications, see . toroidal Type I compactifications with non-zero $`B`$-field were studied. In it was pointed out that even though the NS-NS 2-form is projected out of the closed sting spectrum by the orientifold projection $`\mathrm{\Omega }`$, quantized expectation values of $`B_{ij}`$ are allowed with $`i,j`$ corresponding to the toroidally compactified coordinates. In particular, since the components of $`B_{ij}`$ are defined up to unit shifts $`B_{ij}B_{ij}+1`$, and $`B_{ij}`$ is odd under the world-sheet parity reversal $`\mathrm{\Omega }`$, the allowed background values for the components of $`B_{ij}`$ are 0 and 1/2. Furthermore, in it was found that the rank of the gauge group coming from D9-branes wrapped on tori with non-vanishing half-integer $`B`$-flux is reduced from 16 (that is, the rank of the original $`SO(32)`$ gauge group) down to $`16/2^{b/2}`$, where $`b`$ is the rank (which is always even) of the matrix $`B_{ij}`$. This rank reduction is evident once one considers the cylinder partition function for D9-branes wrapped on such tori. However, the partition function alone does not provide a clear geometric interpretation of the rank reduction phenomenon. In particular, one might conclude that in the cases with non-zero $`B`$-field we have only $`32/2^{b/2}`$ D9-branes instead of the usual 32 D9-branes. This, however, does not appear to be the case. Thus, as was originally pointed out in , the $`\mathrm{\Omega }`$ orientifold of Type IIB with non-zero $`B`$-flux can be viewed as a toroidal Type I compactification with all 32 D9-branes in the presence of non-commuting Wilson lines. This point was further elaborated in more detail in . For the reasons that will become clear in a moment, in this paper we will review the approach of in detail.
Further progress in understanding toroidal orientifolds with non-zero $`B`$-flux was made in , where D-branes transverse to tori with non-trivial $`B_{ij}`$ were considered. In particular, it was argued in that if we, say, consider O7-planes transverse to a 2-torus $`T^2`$ with $`B_{12}=1/2`$ in the directions of $`T^2`$, then we have two types of orientifold planes O7<sup>-</sup> and O7<sup>+</sup>, where the O$`7^{}`$-plane refers to the usual orientifold plane with the $`SO`$ type of orientifold projection on the Chan-Paton charges, whereas the O$`7^+`$-plane refers to the orientifold plane with the $`Sp`$ type of projection. More concretely, out of the four O7-planes (located at the 4 points on $`T^2`$ fixed under the reflection $`R:X_{1,2}X_{1,2}`$, where $`R`$ is now a part of the orientifold projection $`\mathrm{\Omega }R(1)^{F_L}`$) three are of the O7<sup>-</sup> type, while one is of the O7<sup>+</sup> type (to be contrasted with the case with trivial $`B`$-flux where all four O7-planes are of the O7<sup>-</sup> type). The R-R charges of the O7<sup>-</sup>\- and O7<sup>+</sup>-planes are $`8`$ and $`+8`$, respectively, so that the total R-R charge to be canceled by D7-branes is $`16`$. This implies that we must introduce 16 (instead of 32) D7-branes in the presence of non-zero $`B_{12}`$, hence “rank reduction”. However, as we will point out in the following, the mechanisms for the rank reduction in the cases of D-branes wrapped on tori with non-zero $`B_{ij}`$ vs. D-branes transverse to such tori are different. In the former case we still have 32 D-branes, and the rank reduction occurs due to non-commuting Wilson lines . In the latter case the number of D-branes is not 32 but $`32/2^{b/2}`$ due to the fact that some of the orientifold planes are no longer of the $`SO`$ but $`Sp`$ type. Thus, in this case there is really no “rank reduction”. Moreover, the two cases are not T-dual to each in the usual sense as T-duality is more subtle in the presence of the $`B`$-flux (some aspects of T-duality in this context were discussed in ). We will make this more precise in the following using the approach of , which, in particular, makes it clear that one of the O7-planes must indeed be of the $`Sp`$ type.
So far we have mentioned toroidal orientifold compactifications which preserve 16 supersymmetries. To obtain backgrounds with reduced supersymmetry we need to consider compactifications with non-trivial holonomy. A step in this direction was made in , where K3 orientifolds with non-zero $`B`$-field were discussed. More concretely, in the following backgrounds were considered. Start with Type IIB on $`\mathrm{K3}=T^4/𝐙_M`$, where $`M=2,3,4,6`$ so that the orbifold action on $`T^4`$ is crystallographic. Then consider the $`\mathrm{\Omega }`$ orientifold, where $`\mathrm{\Omega }`$ acts not as in the smooth caseOrientifolds of Type IIB on smooth K3 surfaces with non-zero $`B`$-flux were discussed in . but as in , that is, $`\mathrm{\Omega }`$ maps the $`g^k`$ twisted sector to the $`g^{Mk}`$ twisted sector , where $`g`$ is the generator of the orbifold group $`𝐙_M`$, and $`k=0,\mathrm{},M1`$. Let us assume that there is a non-zero half-integer $`B`$-flux of rank $`b`$ in the directions of K3. This is the setup of . A priori we expect that in the $`𝐙_3`$ case we have no D5-branes (for the above choice of the orientifold projection), but some number of D9-branes. In the other three cases, namely, $`𝐙_2,𝐙_4,𝐙_6`$, we expect both D9- and D5-branes to be present. The aforementioned question of how many D9-branes we have in these backgrounds, namely, whether we have 32 or $`32/2^{b/2}`$ D9-branes, becomes extremely relevant, at least in the cases where we have both D9- and D5-branes. The reason for this is the following. As was originally pointed out in , the 59 sector states arise with multiplicity of $`2^{b/2}`$, which is 1 in the absence of the $`B`$-field, but becomes non-trivial whenever the $`B`$-field is non-zero. This fact becomes evident, as was explained in , if one examines the boundary states for the D9- and D5-branes in the presence of the $`B`$-field<sup>§</sup><sup>§</sup>§Equivalently, one can examine the annulus partition function, and see that the multiplicity of states in the 59 sector is indeed $`2^{b/2}`$. This was done for the $`𝐙_2`$ models in .. In particular, this multiplicity of states is in accord with the tadpole/anomaly cancellation requirements in, say, the $`𝐙_2`$ model . However, just by looking at the boundary states (or, equivalently, the partition function), the geometric interpretation of this multiplicity of states in the 59 sector is obscure. In particular, in string theory we expect that no two states should have identical vertex operators. Thus, one should be able to distinguish the otherwise degenerate 59 states (in the presence of the $`B`$-field) from each other by some quantum numbers. In this paper we will give an explicit answer to this question. In particular, we will use the approach of , and point out that the number of D5-branes (if present) is $`32/2^{b/2}`$, whereas the number of D9-branes is always 32, albeit the rank of the 99 gauge group is $`16/2^{b/2}`$ (for the reasons mentioned above). The 59 sector degeneracy then is related to the fact that in these sectors there are $`2^{b/2}`$ different vertex operators with otherwise identical quantum numbers, and the former are distinguished by $`2^{b/2}`$ Chan-Paton degrees of freedom that 32 D9-branes have on top of those corresponding to the unbroken gauge group. In particular, as we will see in the following, these degrees of freedom correspond to $`2^{b/2}`$ different charges carried by the 59 sector states under a discrete gauge symmetry, namely, $`(𝐙_2)^{(b/2)}`$, arising in the 99 sector. In fact, one can turn this point around and argue that the multiplicity of states in the 59 sector gives us a hint for the number of D9-branes being 32 and not $`32/2^{b/2}`$.
Understanding the geometry underlying toroidal orientifolds with NS-NS $`B`$-flux, which is one of the aims of this paper, is a useful tool for shedding light on other non-trivial orientifold compactifications with $`B`$-flux, say, with reduced number of supersymmetries. In particular, in this paper we will revisit the K3 orientifolds with $`B`$-flux originally discussed in . It turns out that there are various subtleties arising in these models. For instance, as we will argue in the following, the conformal field theory $`T^4/𝐙_2`$ orbifold does not appear to be a consistent background for the corresponding orientifolds with $`B`$-flux. This is related to the fact that in the presence of non-trivial $`B`$-flux we must have both O5<sup>-</sup>\- as well as O$`5^+`$-planes at various $`𝐙_2`$ orbifold fixed points, and this is incompatible with the presence of the twisted $`B`$-flux in the conformal field theory orbifold. The $`𝐙_6`$ as well as $`𝐙_4`$ models with $`B`$-flux also suffer from this problem. In fact, we will discuss other subtleties arising in the $`𝐙_6`$ and $`𝐙_4`$ models, which appear to be related, just as it turns out to be the case for their $`𝐙_2`$ counterparts, to the fact that here we have 59 sectors whose presence appears to be incompatible with the lack of vector structure dictated by non-zero $`B`$-flux. On the other hand, $`𝐙_3`$ models with, say, D5-branes only (but no mixed 59 sector) appear to be consistentHere we should point out that the other aforementioned models might be consistent in some other sense, but we fail to find consistent constructions for them within the (perhaps limited) orientifold framework..
The key points of the above discussion carry over to analogous compactifications on Calabi-Yau orbifolds. In particular, every time we have, say, a $`𝐙_3`$ twist acting in the compact directions with non-zero $`B`$-flux, extra care is needed in making sure that tadpole cancellation is compatible with geometric constraints. We illustrate these issues by revisiting the four dimensional models of , in some of which we also observe various subtleties.
The remainder of this paper is organized as follows. In section II we discuss toroidal orientifolds with $`B`$-flux. Our discussion here is essentially a generalization of non-commutative toroidal compactifications in the presence of orientifold planes. This, generalization, however, is not completely straightforward as including orientifold planes brings in additional subtle issues. In section III we revisit K3 orientifolds with $`B`$-flux, and explain in detail the aforementioned subtle inconsistencies arising in some of these backgrounds. In section IV we discuss four dimensional cases with $`SU(3)`$ as well $`SU(2)`$ holonomy. In particular, here we construct consistent orientifold models with $`B`$-flux which do not suffer from difficulties encountered in the K3 cases. We comment on various issues in section V.
## II Toroidal Orientifolds with $`B`$-flux
In this section we will review the effects of the $`B`$-field in toroidal orientifolds. We will use the approach of , where the geometric interpretation of such backgrounds is evident. In subsection A we will discuss the cases where D-branes (and the corresponding orientifold planes) wrap a torus with non-zero $`B`$-field. In subsection B will will consider the cases where D-branes are transverse to such tori. In subsection C we will generalize our discussion to the cases where both types of D-branes are present. There we will also discuss in detail T-duality in orientifolds with $`B`$-flux, and, in particular, argue that the results of this section are consistent with T-duality considerations.
### A D-branes and O-planes Wrapped on Tori with $`B`$-flux
Consider Type IIB (it is straightforward to generalize our discussion to Type IIA) on $`𝐑^{1,9d}T^d`$ in the presence of some number $`n`$ of D$`p`$-branes completely wrapping the $`d`$-torus ($`pd`$). We would like to study the effect of turning on a quantized (half-integer) $`B`$-field in the directions of $`T^d`$. For the sake of clarity we will consider D9-branes wrapping $`T^2`$. Generalizations to other cases are completely straightforward.
Thus, let us start from Type IIB on $`𝐑^{1,7}T^2`$ in the presence of some number $`n`$ of D9-branes. Let us first assume the the $`B`$-field in the compactified directions is zero: $`B_{12}=0`$. Let $`g_{ij}`$ be the metric on $`T^2`$. Then in the suitable normalization the left- and right-moving closed string momenta are given by
$$P_{L,R}=\frac{1}{2}\stackrel{~}{e}^im_i\pm e_in^i,$$
(1)
where $`e_i`$ are two-component vielbeins satisfying $`e_ie_j=g_{ij}`$, while $`\stackrel{~}{e}^i`$ are their duals: $`\stackrel{~}{e}^i\stackrel{~}{e}^j=\stackrel{~}{g}^{ij}`$, where $`\stackrel{~}{g}^{ij}`$ is the inverse of $`g_{ij}`$. Also, the integers $`m_i`$ and $`n^i`$ are the momentum and winding numbers, respectively. As usual, $`T^2`$ can be viewed as a quotient $`𝐑^2/\mathrm{\Lambda }`$, where the lattice $`\mathrm{\Lambda }\{e_in^i\}`$, and the coordinates $`X_i`$ on $`T^2`$ are identified via $`X_iX_i+e_i`$.
Next, consider a freely acting $`𝐙_2𝐙_2`$ orbifold of this theory defined as follows. Let $`S_i`$ be a half-lattice shift in the $`X_i`$ direction, that is, $`S_iX_i=X_i+e_i/2`$. Note that the set $`\{I,S_1,S_2,S_3\}`$ forms a freely acting orbifold group isomorphic to $`𝐙_2𝐙_2`$. Here $`I`$ is the identity element, and $`S_3S_1S_2`$. Thus, in this orbifold we have the untwisted sector labeled by $`I`$, and three twisted sectors labeled by $`S_1,S_2,S_3`$. The left- and right-moving momenta can now be written as
$$P_{L,R}(\alpha ^1,\alpha ^2)=\frac{1}{2}\stackrel{~}{e}^im_i\pm e_i(n^i+\frac{1}{2}\alpha ^i),$$
(2)
where $`(\alpha ^1,\alpha ^2)=(0,0)`$, $`(1,0)`$, $`(0,1)`$ and $`(1,1)`$ in the untwisted, $`S_1`$, $`S_2`$ and $`S_3`$ twisted sectors, respectively. The most general action of $`S_i`$ on the left- and right-moving momenta (compatible with modular invariance) is given by
$$S_i|P_L,P_R_{(\alpha ^1,\alpha ^2)}=ϵ_i(\alpha ^1,\alpha ^2)\mathrm{exp}(\pi im_i)|P_L,P_R_{(\alpha ^1,\alpha ^2)},$$
(3)
where $`ϵ_i(0,0)1`$, $`ϵ_1(1,0)=ϵ_2(0,1)=1`$, $`ϵ_1(0,1)=ϵ_2(1,0)=ϵ_{1,2}(1,1)=ϵ`$. Here $`ϵ`$ can take two values: $`\pm 1`$. If $`ϵ=+1`$, then we have the usual freely acting $`𝐙_2𝐙_2`$ orbifold. It is straightforward to show that the resulting theory corresponds to a compactification on $`(T^2)^{}`$ with the metric $`g_{ij}^{}=g_{ij}/4`$, and zero $`B`$-field $`B_{12}=0`$. However, if $`ϵ=1`$, in which case we have discrete torsion between the generators $`S_1`$ and $`S_2`$ of the two $`𝐙_2`$ subgroups of the orbifold group (that is, $`S_1`$ acts with an extra minus sign in the $`S_2`$ twisted sector, and, consequently, $`S_2`$ acts with the extra minus sign in the $`S_1`$ twisted sector; both $`S_1`$ and $`S_2`$ act with an extra minus sign in the $`S_3`$ twisted sector), then it is straightforward to show that the resulting theory corresponds to a compactification on $`(T^2)^{}`$ with the metric $`g_{ij}^{}=g_{ij}/4`$, but now the $`B`$-field is non-zero: $`B_{12}=1/2`$. In deriving these results it is important to note that (3) implies that only even momentum numbers $`m_i`$ survive the orbifold projection in all four sectors if there is no discrete torsion (that is, $`ϵ=1`$), while in the case with discrete torsion (that is, $`ϵ=1`$) the momenta kept after the orbifold projection are given as follows. Both $`m_1`$ and $`m_2`$ are even in the untwisted sector. In the $`S_1`$ twisted sector $`m_1`$ is even and $`m_2`$ is odd. In the $`S_2`$ twisted sector $`m_1`$ is odd and $`m_2`$ is even. Finally, in the $`S_3`$ twisted sector both $`m_1`$ and $`m_2`$ are odd. Taking all of this into account, we can see from (2) and (3) that the resulting left- and right-moving momenta are given by
$$P_{L,R}=\frac{1}{2}\stackrel{~}{e}^i(m_i^{}B_{ij}n^j)\pm e_i^{}n^i,$$
(4)
where the new momentum and winding numbers $`m_i^{}`$ and $`n^i`$ are now arbitrary integers, the new vielbeins $`e_i^{}`$ and their duals $`\stackrel{~}{e}^i`$ are related to the original ones via $`e_i^{}=e_i/2`$, $`\stackrel{~}{e}^i=2\stackrel{~}{e}^i`$, and the $`B`$-field is zero for $`ϵ=1`$, while $`B_{12}=1/2`$ for $`ϵ=1`$.
Thus, what we have learned from the above discussion is that we can describe a compactification on $`(T^2)^{}`$ with half-integer $`B`$-field as a freely acting $`𝐙_2𝐙_2`$ orbifold that involves half-shifts along the two cycles of $`T^2`$ (the metric on $`(T^2)^{}`$ is four times as small as that on $`T^2`$, that is, both of the cycles on $`(T^2)^{}`$ are half the size of the corresponding cycles on $`T^2`$) with discrete torsion between the generators $`S_1`$ and $`S_2`$ of the two $`𝐙_2`$ subgroups of the orbifold group $`𝐙_2𝐙_2`$. This approach proves to be convenient in the context of D-branes (and O-planes) wrapping tori with non-zero $`B`$-field as we can recast the latter problem into the corresponding freely acting orbifold of a setup where D-branes are wrapping a torus with zero $`B`$-field. In particular, as we will see in a moment, this provides a geometrization of the cases where D-branes wrap tori with non-zero $`B`$-field.
Let us now see what happens to open strings stretched between $`n`$ D9-branes wrapped on $`(T^2)^{}`$. As we have already mentioned, to study this system we can start from $`n`$ D9-branes wrapped on $`T^2`$ without the $`B`$-field, and then consider the aforementioned freely acting $`𝐙_2𝐙_2`$ orbifold. Thus, before orbifolding we have open string states which can be obtained by the corresponding Kaluza-Klein (KK) compactification of the ten dimensional open string spectrum on $`T^2`$. In particular, among other quantum numbers (corresponding to the open string oscillator modes) we have the Kaluza-Klein momenta $`m_i𝐙`$ with the contributions to the masses of the corresponding open string states given by $`M_𝐦^2=\frac{1}{2}P_𝐦^2`$, where $`P_𝐦\stackrel{~}{e}^im_i`$ (here $`𝐦(m_1,m_2)`$). The generators $`S_1`$ and $`S_2`$ of the $`𝐙_2𝐙_2`$ orbifold group act on the momentum states as follows:
$$S_i|P_𝐦=\mathrm{exp}(\pi im_i)|P_𝐦.$$
(5)
However, we must also specify the action of the orbifold group elements on the Chan-Paton charges of D9-branes. It is described by $`n\times n`$ matrices that form a representation of $`𝐙_2𝐙_2`$. Thus, we are free to choose the Chan-Paton matrix $`\gamma _I`$ corresponding to the identity element $`I`$ of the orbifold group as $`\gamma _I=I_n`$, where here and in the following $`I_m`$ will denote the $`m\times m`$ identity matrix. As to the twisted Chan-Paton matrices $`\gamma _{S_1}`$ and $`\gamma _{S_2}`$ (as well as $`\gamma _{S_3}`$), they must satisfy certain constraints depending on the choice of $`ϵ`$, that is, depending upon whether we have discrete torsion or not. In particular, if $`ϵ=1`$, then $`\gamma _{S_1}`$ and $`\gamma _{S_2}`$ must commute. This follows from the fact that in this case we simply rescale the two cycles on $`T^2`$ to obtain $`(T^2)^{}`$ without the $`B`$-field, and if the matrices $`\gamma _{S_i}`$ are non-trivial, then they act as (discrete) Wilson lines corresponding to the two cycles on $`T^2`$. However, if $`ϵ=1`$, that is, if we have non-zero $`B`$-field ($`B_{12}=1/2`$), then the string consistency (in particular, the closed to open string coupling consistency) requires that $`\gamma _{S_1}`$ and $`\gamma _{S_2}`$ must anticommute . In this case we can still view the action of the orbifold group on the Chan-Paton Charges as having (discrete) Wilson lines, but now they are no longer commuting . Thus, to summarize, the Chan-Paton matrices must satisfy
$$\gamma _{S_1}\gamma _{S_2}=ϵ\gamma _{S_2}\gamma _{S_1}.$$
(6)
In the case without discrete torsion ($`ϵ=1`$) we thus have gauge bundles with vector structure. In the case with discrete torsion ($`ϵ=1`$) we have gauge bundles without vector structure (and the corresponding generalized second Stieffel-Whitney class is non-vanishing) .
Note that in the cases without discrete torsion the matrices $`\gamma _{S_i}`$ can essentially be arbitrary as long as they commute and form a (projective) representation of $`𝐙_2𝐙_2`$. In particular, their traces are not fixed - there are no tadpoles associated with the twists $`S_i`$ as they are freely acting. Thus, the action of the $`𝐙_2𝐙_2`$ orbifold on the Chan-Paton charges can be trivial, that is, $`\gamma _{S_i}=I_n`$, which corresponds to having trivial Wilson lines. If we do not include an orientifold plane, then the gauge symmetry is $`U(n)`$ with this choice of Wilson lines, while if Wilson lines are non-trivial, then the gauge group $`G`$ is a subgroup of $`U(n)`$ with rank $`r(G)=n`$. In the case of D9-branes the oriented open string theory suffers from massless tadpoles, which can be canceled if we introduce the O9<sup>-</sup>-plane (with the $`SO`$ type of orientifold projection on the Chan-Paton charges), and choose the number $`n`$ of D9-branes to be $`n=32`$. Then in the case of trivial Wilson lines we have the usual $`SO(32)`$ gauge group, while non-trivial Wilson lines break $`SO(32)`$ to its subgroup $`G`$ with rank $`r(G)=16`$.
However, in the case with discrete torsion the situation is qualitatively different. In particular, all three twisted Chan-Paton matrices $`\gamma _{S_a}`$, $`a=1,2,3`$, must be traceless. This follows from the fact that for these matrices to form a (projective) representation of $`𝐙_2𝐙_2`$, they must satisfy
$`\gamma _{S_a}^2=\eta _{aa}I_n,`$ (7)
$`\gamma _{S_a}\gamma _{S_b}=\eta _{ab}\gamma _{S_c},abc,`$ (8)
where not all the nine structure constants $`\eta _{ab}`$ are independent but satisfy the following relations:
$`\eta _{33}=\eta _{11}\eta _{22},`$ (9)
$`\eta _{21}=\eta _{12},`$ (10)
$`\eta _{13}=\eta _{31}=\eta _{11}\eta _{12},`$ (11)
$`\eta _{23}=\eta _{32}=\eta _{22}\eta _{12}.`$ (12)
Here $`\eta _{11},\eta _{22},\eta _{12}`$ a priori independently take values $`\pm 1`$. It then follows that $`\mathrm{Tr}(\gamma _a)0`$. Moreover, even if we consider the oriented open string theory, the number of D9-branes must be even: $`n=2N`$. In fact, we can now see the rank reduction phenomenon mentioned in the previous section. Thus, the rank of the unbroken gauge group is no longer $`n`$ but twice as small, that is, $`N`$. This follows from the fact that the Wilson lines corresponding to the two cycles of $`(T^2)^{}`$ do not commute, which can be seen from the fact that $`\gamma _{S_1}`$ and $`\gamma _{S_2}`$ anticommute. If we consider an unoriented open string theory, that is, if we introduce an orientifold 9-plane, the rank of the gauge group is then no longer $`n/2`$ (as is the case without discrete torsion), but rather $`N/2`$ (in the presence of an orientifold plane $`N`$ must also be even). Intuitively it should be clear that to cancel all the tadpoles we must introduce the O9<sup>-</sup>-plane (and not the O9<sup>+</sup>-plane), and choose the number of D9-branes to be $`n=32`$. However, we would like to derive this result rigorously as understanding this point will be important for the subsequent discussions.
To do this, let us start with the Klein bottle amplitude $`𝒦`$. It is obtained from the torus amplitude by inserting the orientifold projection $`\mathrm{\Omega }`$ into the trace over the Hilbert space of closed string states. This implies that the only states contributing into the Klein bottle amplitude are left-right symmetric closed string states. The oscillator modes are not going to be important in the following as their contributions are the same with or without the $`B`$-flux, so let us focus on the left- and right-momentum contributions. Note, in particular, that the momentum numbers $`m_i`$ are invariant under the action of $`\mathrm{\Omega }`$, while the winding numbers $`n^i`$ change sign under the action of $`\mathrm{\Omega }`$. This implies that only the states with zero winding (but arbitrary momentum) numbers contribute to the Klein bottle amplitude. Such states are the same regardless of the $`B`$-flux, which can be readily seen from (4). Thus, the Klein bottle amplitude is independent of the $`B`$-field. We will write the Klein bottle amplitude in the language of the $`𝐙_2𝐙_2`$ freely acting orbifold of $`T^2`$ with discrete torsion (here we do not display the oscillator contributions):
$$𝒦=\left(\frac{1}{2}\right)\left(\frac{1}{4}\right)\underset{𝐦}{}q^{\frac{1}{2}P_𝐦^2}\left[1+(1)^{m_1}+(1)^{m_2}+(1)^{m_1+m_2}\right],$$
(13)
where $`P_𝐦=\stackrel{~}{e}^im_i`$, $`𝐦=(m_1,m_2)`$, and $`m_i𝐙`$. Also, the first numerical prefactor of $`(1/2)`$ is related to the orientifold projection, while the second numerical prefactor of $`(1/4)`$ is related to the $`𝐙_2𝐙_2`$ orbifold projection. As we have already mentioned, the Klein bottle amplitude (13) does not depend on whether we have discrete torsion or not, that is, $`𝒦`$ in (13) is the same as the Klein bottle amplitude for the $`𝐙_2𝐙_2`$ freely acting orbifold without discrete torsion. Note that $`𝒦`$ is written in terms of the metric on $`T^2`$. We can rewrite it in terms of the metric on $`(T^2)^{}`$ as follows:
$$𝒦=\left(\frac{1}{2}\right)\underset{𝐦^{}}{}q^{\frac{1}{2}P_𝐦^{}^2},$$
(14)
where where $`P_𝐦^{}=\stackrel{~}{e}^im_i^{}`$, $`𝐦^{}=(m_1^{},m_2^{})`$, and $`m_i^{}𝐙`$. In arriving at (14) we have explicitly performed the $`𝐙_2𝐙_2`$ orbifold projections in (13) which keep only the states with $`m_i2𝐙`$. These states are then rewritten in terms of new momenta $`m_i^{}𝐙`$ with the dual vielbeins $`\stackrel{~}{e}^i=2\stackrel{~}{e}^i`$ on $`(T^2)^{}`$.
Let us now discuss the annulus amplitude. We will write the latter in the language of D9-branes wrapped on $`T^2`$ with the subsequent $`𝐙_2𝐙_2`$ freely acting orbifold action. Thus, the annulus amplitude is given by:
$`𝒜=\left({\displaystyle \frac{1}{2}}\right)\left({\displaystyle \frac{1}{4}}\right){\displaystyle \underset{𝐦}{}}q^{\frac{1}{2}P_𝐦^2}[(\mathrm{Tr}(\gamma _I))^2+`$ $`(1)^{m_1}(\mathrm{Tr}(\gamma _{S_1}))^2+`$ (16)
$`(1)^{m_2}(\mathrm{Tr}(\gamma _{S_2}))^2+(1)^{m_1+m_2}(\mathrm{Tr}(\gamma _{S_3}))^2].`$
Here we have chosen a specific orientation on the annulus so that the traces over the Chan-Paton charges give $`(\mathrm{Tr}(\gamma _{S_a}))^2`$, while if we chose the opposite orientation, we would instead have $`\mathrm{Tr}(\gamma _{S_a})\mathrm{Tr}(\gamma _{S_a}^1)`$.
Note that $`\mathrm{Tr}(\gamma _I)=n`$, and $`\mathrm{Tr}(\gamma _{S_a})=0`$, $`a=1,2,3`$, and we can formally rewrite the annulus amplitude via
$`𝒜_1=\left({\displaystyle \frac{1}{2}}\right){\displaystyle \underset{𝐦}{}}q^{\frac{1}{2}P_𝐦^2}(\mathrm{Tr}(I_N))^2,`$ (17)
so that naively one could reinterpret this annulus amplitude as corresponding to that of $`N=n/2`$ D9-branes (instead of $`n`$ D9-branes). This interpretation, however, would be erroneous. Indeed, one should get a hint of this from the sum over the momenta $`𝐦`$. These are the momenta written in terms of the metric on the original torus $`T^2`$. However, we would have to rewrite it in terms of the metric on the torus $`(T^2)^{}`$ \- after all we are considering D9-branes wrapped $`(T^2)^{}`$ (plus the $`B`$-field), and not on $`T^2`$, the latter merely being the starting point for the $`𝐙_2𝐙_2`$ orbifold with discrete torsion which is a way of obtaining $`(T^2)^{}`$ with the $`B`$-field. However, if we chose to reinterpret the annulus partition function via (17), where we essentially would be trying to ameliorate all the traces of the $`𝐙_2𝐙_2`$ orbifold, we would have to write the momenta in terms of the metric on $`(T^2)^{}`$. Thus, we would ultimately arrive at the conclusion that the momenta would have to take half-integer (instead of usual integer) values. Indeed, $`P_𝐦`$ in (17) is given by $`P_𝐦=\stackrel{~}{e}^im_i`$, which can be rewritten in terms of the dual vielbeins $`\stackrel{~}{e}^i`$ on $`(T^2)^{}`$ as $`P_{\widehat{𝐦}}=\stackrel{~}{e}^i\widehat{m}_i`$, where $`\widehat{m}_im_i/2`$ (recall that $`\stackrel{~}{e}^i=2\stackrel{~}{e}^i`$), so that the new momenta $`\widehat{m}_i`$ would be half-integer. This signals that such an interpretation would indeed be erroneousThis erroneous interpretation has been adopted in most of the literature on orientifolds with non-zero $`B`$-flux. Often it does not lead to inadequate description of the massless spectra of such orientifolds. For instance, this interpretation, which was essentially adopted in , gives the correct results in the case of toroidal orientifolds with $`B`$-flux due to their relative simplicity (these theories have 16 supercharges). In fact, this can be understood from the fact that the annulus amplitude in (17) would give the same massless spectrum as that in (16). Moreover, it would predict the same degeneracy of states at the massive KK levels as (16), but the vertex operators at the massive levels in the two interpretations would be different. As we will see in the following, in more complicated cases such as K3 or Calabi-Yau orientifolds with $`B`$-flux knowing the correct structure of vertex operators becomes relevant already at the massless level, for instance, in the 59 open string sector.. (In particular, note that in the Klein bottle amplitude (14) the sum is over integer momenta $`m_i^{}`$ written in terms of the metric on $`(T^2)^{}`$, while in the annulus amplitude (17) the sum is over half-integer momenta $`\widehat{m}_i`$.) Instead, the correct physical interpretation is that we have $`n=2N`$ (and not $`N`$) D9-branes. The effect of the $`B`$-field then can be understood as in the interpretation provided by the annulus amplitude (16) via the $`𝐙_2𝐙_2`$ orbifold construction. Thus, we see that merely examining the partition functions is not sufficient to understand the geometric structure of orientifolds with $`B`$-field as the former only provide us with the multiplicities of states at various string levels, but may not carry the complete information about the structure of the vertex operators.
At any rate, let us start with the annulus amplitude (16), and proceed further. Thus, we would like to discuss the Möbius strip amplitude next. It is given by
$`=\lambda \left({\displaystyle \frac{1}{2}}\right)\left({\displaystyle \frac{1}{4}}\right){\displaystyle \underset{𝐦}{}}q^{\frac{1}{2}P_𝐦^2}[`$ $`\mathrm{Tr}(\gamma _\mathrm{\Omega }^1\gamma _\mathrm{\Omega }^T)+(1)^{m_1}\mathrm{Tr}(\gamma _{\mathrm{\Omega }S_1}^1\gamma _{\mathrm{\Omega }S_1}^T)+`$ (19)
$`(1)^{m_2}\mathrm{Tr}(\gamma _{\mathrm{\Omega }S_2}^1\gamma _{\mathrm{\Omega }S_2}^T)+(1)^{m_1+m_2}\mathrm{Tr}(\gamma _{\mathrm{\Omega }S_3}^1\gamma _{\mathrm{\Omega }S_3}^T)],`$
where $`\lambda =1`$ for the $`SO/Sp`$ orientifold projection (that is, $`\lambda =\pm 1`$ for the O9<sup>±</sup>-plane). We can simplify this expression by noting that
$`\mathrm{Tr}(\gamma _\mathrm{\Omega }^1\gamma _\mathrm{\Omega }^T)=\mathrm{Tr}(\gamma _I),`$ (20)
$`\mathrm{Tr}(\gamma _{\mathrm{\Omega }S_a}^1\gamma _{\mathrm{\Omega }S_a}^T)=\mathrm{Tr}(\gamma _{(S_a)^2})=\eta _{aa}\mathrm{Tr}(\gamma _I).`$ (21)
In the second line we have used the fact that $`\gamma _{(S_a)^2}=\gamma _{S_a}^2=\eta _{aa}I_n`$. Thus, the Möbius strip amplitude is given by
$`=\lambda \left({\displaystyle \frac{1}{2}}\right)\left({\displaystyle \frac{1}{4}}\right){\displaystyle \underset{𝐦}{}}q^{\frac{1}{2}P_𝐦^2}\mathrm{Tr}(\gamma _I)\left[1+\eta _{11}(1)^{m_1}+\eta _{22}(1)^{m_2}+\eta _{33}(1)^{m_1+m_2}\right],`$ (22)
where $`\eta _{33}=\eta _{11}\eta _{22}`$.
Note that the quantity in the square brackets in (22) is always $`+2`$ or $`2`$. Thus, we can write
$$1+\eta _{11}(1)^{m_1}+\eta _{22}(1)^{m_2}+\eta _{33}(1)^{m_1+m_2}2\rho _𝐦(\eta _{aa}),$$
(23)
where $`\rho _𝐦(\eta _{aa})=\pm 1`$. In fact, for a given choice of $`\eta _{aa}`$, $`\rho _𝐦`$ by definition only depends on whether $`m_1`$ and $`m_2`$ are even or odd. So formally we can rewrite the Möbius strip amplitude as follows:
$`_1=\lambda \left({\displaystyle \frac{1}{2}}\right){\displaystyle \underset{𝐦}{}}q^{\frac{1}{2}P_𝐦^2}\mathrm{Tr}(I_N)\rho _𝐦(\eta _{aa}).`$ (24)
Thus, naively one could reinterpret this Möbius strip amplitude along the lines of (17) as corresponding to that of $`N=n/2`$ D9-branes (instead of $`n`$ D9-branes), with the $`SO/Sp`$ orientifold projection at integer KK levels $`\widehat{m}_i`$ (recall that $`\widehat{m}_im_i/2`$) if $`\lambda \rho _𝐦(\eta _{aa})=1`$ for $`m_1,m_22𝐙`$, while at the half-integer KK levels $`\widehat{m}_i`$ the type of the orientifold projection is determined by the corresponding signs $`\lambda \rho _𝐦(\eta _{aa})`$ with either $`m_1`$ or $`m_2`$ or both odd. Such an interpretation, however, would be erroneous for the same reasons as in the annulus case discussed above.
Next, we turn to the tadpole cancellation conditions. To extract the tadpoles, we must rewrite the open string loop-channel Klein bottle $`𝒦`$, annulus $`𝒜`$ and Möbius strip $``$ amplitudes in terms of the corresponding closed string tree-channel exchange expressions. In doing so, as usual, one must be careful with the relative normalizations between the proper times on these three surfaces. The modular transformations that map the loop-channel expressions to the tree-channel expressions amount to Poisson resummations of the momentum sums in $`𝒦`$, $`𝒜`$ and $``$ (they also act non-trivially on the characters corresponding to the oscillator contributions). It is not difficult to see that after Poisson resummations the terms in (13), (16) and (19) containing $`(1)^{m_1}`$, $`(1)^{m_2}`$ and $`(1)^{m_1+m_2}`$ do not contain terms corresponding to the massless closed string exchange, and, therefore, do not contribute to the tadpoles. This, actually, has been anticipated from the fact that the $`𝐙_2𝐙_2`$ orbifold is freely acting. The remaining terms, which do contribute to the tadpoles, are the same (up to a universal overall factor of $`1/4)`$ as those in the $`\mathrm{\Omega }`$ orientifold of Type IIB on $`T^2`$ (with zero $`B`$-field) in the presence of $`n`$ D9-branes. This makes extracting the tadpoles in our case straightforward. Thus, the tadpoles are given by (here $`c`$ is a universal numerical constant):
$`\mathrm{Tad}(𝒦)=c\mathrm{Vol}(T^2)32^2,`$ (25)
$`\mathrm{Tad}(𝒜)=c\mathrm{Vol}(T^2)(\mathrm{Tr}(\gamma _I))^2,`$ (26)
$`\mathrm{Tad}()=c\mathrm{Vol}(T^2)64\lambda \mathrm{Tr}(\gamma _I),`$ (27)
so that the total tadpole factorizes into a perfect square
$$\mathrm{Tad}=c\mathrm{Vol}(T^2)\left[32+\lambda \mathrm{Tr}(\gamma _I)\right]^2.$$
(28)
Since $`\mathrm{Tr}(\gamma _I)=n`$, we conclude that the orientifold projection must be of the $`SO`$ type ($`\lambda =1`$), that is, we must include the O9<sup>-</sup>-plane (and not the O9<sup>+</sup>-plane), as well as $`n=32`$ D9-branes.
Thus, we see that the number of D9-branes is indeed 32, and the orientifold plane is of the O9<sup>-</sup> type, which induces the $`SO`$ type of projection on the D9-branes. This, however, does not imply that we cannot obtain, say, $`Sp`$ gauge symmetry, which is expected to arise in orientifolds with $`B`$-flux . In order to understand this point we must study the possible choices of the twisted Chan-Paton matrices $`\gamma _{S_i}`$, which is related to the question of possible gauge bundles on $`T^2`$ in the presence of the $`B`$-field.
To begin with let us note that it suffices to consider a $`2\times 2`$ representation for $`\gamma _{S_i}`$ as the full $`2N\times 2N`$ matrices can be obtained as the $`N`$-fold copy of the corresponding $`2\times 2`$ matrices. Thus, we can write
$$\gamma _{S_a}=\gamma _aI_N,$$
(29)
where the matrices $`\gamma _a`$, $`a=1,2,3`$, form the aforementioned $`2\times 2`$ representation. These matrices must satisfy $`\gamma _a^2=\eta _{aa}I_2`$ as well as $`\mathrm{Tr}(\gamma _a)=0`$ conditions, which, in particular, imply that $`det(\gamma _a)=\eta _{aa}`$. Recall that $`\eta _{33}=\eta _{11}\eta _{22}`$, so either all three $`\gamma _a`$ matrices have determinant $`+1`$ (if $`\eta _{11}=\eta _{22}=1`$), or two have determinant $`1`$, while one has determinant $`+1`$ (for the other three choices of $`\eta _{11}`$ and $`\eta _{22}`$). In the former case $`\gamma _a`$ are $`SU(2)`$ matrices, and form a 2-dimensional representation of the non-Abelian dihedral $`D_4`$ subgroup of $`SU(2)`$. In the latter case $`\gamma _a`$ form a 2-dimensional representation of the “double cover” of the $`D_4`$ subgroup of $`SU(2)`$, which we will denote by $`D_4^{}`$. Note that $`D_4^{}`$ is not a subgroup of $`SU(2)`$ but is a subgroup of $`SO(3)`$. Up to equivalent representations, we can write $`\gamma _a`$ for the above two cases as follows:
$`D_4:\gamma _1=i\sigma _3,\gamma _2=i\sigma _2,\gamma _3=i\eta _{12}\sigma _1,`$ (30)
$`D_4^{}:\gamma _1=\sigma _3,\gamma _2=\sigma _1,\gamma _3=i\eta _{12}\sigma _2.`$ (31)
Here $`\sigma _1,\sigma _2,\sigma _3`$ are the usual $`2\times 2`$ Pauli matrices. It is not difficult to show that the $`𝐙_2𝐙_2`$ orbifold projection breaks the $`SO(32)`$ gauge group on 32 D9-branes at the O9<sup>-</sup>-plane down to $`Sp(16)`$ in the $`D_4`$ case<sup>\**</sup><sup>\**</sup>\**In our notations $`Sp(2r)`$ has rank $`r`$., and $`SO(16)`$ in the $`D_4^{}`$ case. More precisely, the unbroken gauge symmetries are $`Sp(16)𝐙_2`$ and $`SO(16)𝐙_2`$, respectively. The extra discrete $`𝐙_2`$ gauge symmetry will be important in the subsequent discussions, so let us elaborate on its appearance in more detail.
Let us first consider the $`D_4^{}`$ case. Note that the first $`𝐙_2`$ twist $`\gamma _{S_1}`$ acting on the Chan-Paton charges breaks $`SO(32)`$ down to $`SO(16)SO(16)`$. The second $`𝐙_2`$ twist $`\gamma _{S_2}`$ breaks the latter down to its diagonal subgroup $`SO(16)_{\mathrm{diag}}`$ times the discrete $`𝐙_2`$ gauge symmetry associated with the permutation of the two $`SO(16)`$ subgroups. Similarly, in the $`D_4`$ case $`\gamma _{S_1}`$ breaks $`SO(32)`$ down to $`U(16)`$, and $`\gamma _{S_2}`$ breaks the latter down to $`Sp(16)`$ times the discrete $`𝐙_2`$ subgroup of the $`U(1)`$ subgroup of $`U(16)`$ (under which, for instance, the fundamental and antifundamental representations of $`SU(16)`$ are charged).
Before we end this subsection, a few comments are in order. First, above we have described how to obtain the $`SO(16)`$ and $`Sp(16)`$ gauge symmetries via the $`𝐙_2𝐙_2`$ freely acting orbifold construction. Since these theories have 16 supercharges, it is clear that it should be possible to construct other points in the moduli space, whose generic points have $`U(1)^8`$ gauge symmetry, but at other points one should be able to obtain enhanced unitary gauge subgroups with rank 8, in particular, $`U(8)`$. Moreover, one expects that the moduli space of gauge bundles on $`T^2`$ without vector structure should be connected, so that it should be possible to continuously interpolate between the $`SO(16)`$ and $`Sp(16)`$ points. A detailed discussion of gauge bundles on $`T^2`$ without vector structure can be found in . Here we will briefly review some of the basic relevant facts. Thus, the matrices $`\gamma _{S_i}`$ given above correspond to points in the moduli space which can be described via the $`𝐙_2𝐙_2`$ freely acting orbifold. However, the matrices $`\gamma _{S_i}`$ can be more generally viewed as describing the gauge bundle on $`T^2`$ without vector structure if we think about them as anticommuting Wilson lines. Thus, we can relax all the constraints we have imposed on $`\gamma _{S_i}`$ except for the non-commutative property: $`\gamma _{S_1}\gamma _{S_2}=\gamma _{S_2}\gamma _{S_1}`$. The most general solution to this constraint can be (up to equivalent representations) written as
$$\gamma _{S_i}=\gamma _i\mathrm{\Gamma }_i,$$
(32)
where the unitary $`N\times N`$ matrices $`\mathrm{\Gamma }_i`$ commute. In particular, note that there is no longer a restriction on $`\mathrm{\Gamma }_i^2`$. Thus, starting from the solution corresponding to $`D_4`$ we can smoothly interpolate to the solution corresponding to $`D_4^{}`$ with the intermediate points corresponding to $`U(N/2)`$ or its subgroups of rank $`N/2`$.
Another point we would like to mention is generalization to higher tori with the rank of $`B_{ij}`$ $`b>2`$. It is clear that, say, in the case of $`T^4`$ we can represent non-zero $`B`$-field essentially along the same lines as we have done so far for $`T^2`$. Instead of being most general here, for illustrative purposes let us consider the case of $`T^4=T^2T^2`$ (generalizations should be clear). We can introduce two Wilson lines corresponding to the two cycles on the first $`T^2`$, call the corresponding Chan-Paton matrices $`\gamma _{S_1},\gamma _{S_2}`$, and two other Wilson lines corresponding to the two cycles on the second $`T^2`$, call the corresponding Chan-Paton matrices $`\gamma _{T_1},\gamma _{T_2}`$. If we take $`\gamma _{S_1}`$ and $`\gamma _{S_2}`$ anticommuting, as well as $`\gamma _{T_1}`$ and $`\gamma _{T_2}`$ anticommuting, but $`\gamma _{S_i}`$ and $`\gamma _{T_j}`$ commuting, this corresponds to rank $`b=4`$ half-integer $`B`$-field on $`T^4`$. Such Chan-Paton matrices can be easily constructed. Thus, let $`\gamma _a`$ be a set of three anticommuting $`2\times 2`$ matrices, and let $`\beta _a`$ be another set of three anticommuting $`2\times 2`$ matrices, each set satisfying the constraints discussed above in the $`T^2`$ case. Then we can choose the Chan-Paton matrices corresponding to the four Wilson lines on $`T^4`$ as follows ($`N^{}n/4`$):
$`\gamma _{S_a}=\gamma _aI_2I_N^{},`$ (33)
$`\gamma _{T_a}=I_2\beta _aI_N^{}.`$ (34)
Note that the type of the unbroken gauge group can be determined as follows. If $`\gamma _a`$ and $`\beta _a`$ both are of the $`D_4^{}`$ or $`D_4`$ type, then the unbroken gauge group is $`SO(N^{})`$. If one of them is of the $`D_4^{}`$ type while the other one is of the $`D_4`$ type, then the unbroken gauge group is $`Sp(N^{})`$. (More precisely, the unbroken gauge symmetry includes the corresponding $`𝐙_2𝐙_2`$ subgroup.) Other points in the moduli space interpolating between these special configurations can be obtained in complete parallel with the $`T^2`$ case. Also, generalizations to higher tori, in particular, $`T^6`$ should be evident from the above discussions.
Finally, note that, say, in the case of $`T^2`$ having two non-commuting Wilson lines implies that some of the components of the non-Abelian gauge field strength $`F_{12}`$ (in the language of the original $`SO(32)`$ gauge symmetry) in the compact directions are non-zero. (In the four dimensional language this implies that some of the D-term components are non-zero.) This is precisely the statement that the corresponding generalized second Stieffel-Whitney class of the gauge bundle in non-vanishing. Note that this is perfectly consistent with the low energy supersymmetry as the moduli corresponding to these directions are absent in the low energy spectrum of the $`SO(16)`$ or $`Sp(16)`$ gauge theory with $`16`$ supercharges. In particular, the moduli space of gauge bundles of the $`SO(32)`$ theory on $`T^2`$ with vector structure is $`2\times 16`$ dimensional, whereas that of gauge bundles of the $`SO(16)/Sp(16)`$ theory on $`T^2`$, which lacks vector structure, is $`2\times 8`$ dimensional. In fact, these two components of the moduli space of gauge bundles on $`T^2`$ are disconnected. Under Type I-heterotic duality the component with vector structure maps to the corresponding part of the Narain moduli space, while, as was originally pointed out in , the component without vector structure maps to the corresponding part of the moduli space of CHL strings (with rank 8 gauge symmetry) in 8 dimensions.
### B D-branes and O-planes Transverse to Tori with $`B`$-flux
In this subsection we will discuss toroidal orientifolds with D-branes and O-planes transverse to tori with non-zero $`B`$-field. From we expect that in these cases we will have both O<sup>-</sup>-planes and O<sup>+</sup>-planes. We will see that this is indeed the case using the $`𝐙_2𝐙_2`$ freely acting orbifold of described in the previous subsection. However, before we discuss orientifolds with non-zero $`B`$-flux, we will take a detour into a discussion of possible types of O<sup>±</sup>-planes as there are two different types of O<sup>+</sup>-planes as well as O<sup>-</sup>-planes<sup>††</sup><sup>††</sup>††I would like to thank Amihay Hanany for a valuable discussion on this point..
To begin with, let us start with Type IIB in ten dimensions. A priori we can orientifold Type IIB by two different orientifold actions, which we will refer to as $`\mathrm{\Omega }_\pm `$. One of these actions, namely, $`\mathrm{\Omega }_{}`$ (which we have been denoting by $`\mathrm{\Omega }`$ in the previous discussions) is the usual orientifold projection which leads to the Type I theory in ten dimensions with 16 supercharges and $`SO(32)`$ gauge group. Note that the action of $`\mathrm{\Omega }_{}`$ on the Chan-Paton charges of D9-branes is antisymmetric in both Neveu-Schwarz and Ramond open string sectors. The corresponding orientifold plane has R-R charge $`32`$, which requires introduction of 32 D9-branes to cancel the R-R charge as each D9-brane has R-R charge $`+1`$. The R-R charge cancellation implies that the corresponding R-R tadpoles are also canceled. The fact that the NS-NS tadpoles also cancel then follows from supersymmetry. However, in cases without supersymmetry NS-NS and R-R tadpole cancellations are a priori independent. To keep track of both NS-NS and R-R tadpoles it is convenient to introduce the notion of the “NS-NS charge” for D-branes and O-planes as a book-keeping device (albeit there is no real conserved charge associated with NS-NS tadpoles). Thus, let us assign NS-NS charge $`+1`$ to a single D9-brane. Then the NS-NS charge of the orientifold plane corresponding to the action of $`\mathrm{\Omega }_{}`$ is $`32`$. In this language the orientifold plane corresponding to the action of $`\mathrm{\Omega }_+`$ has R-R charge $`+32`$ and NS-NS charge $`+32`$. The action of $`\mathrm{\Omega }_+`$ on the Chan-Paton charges of D9-branes is now symmetric in both NS and R open string sectors. Note that the gauge group on $`M`$ D9-branes with the O-plane corresponding to the action of $`\mathrm{\Omega }_+`$ is $`Sp(M)`$. In fact, the corresponding gauge theory is supersymmetric. However, in ten dimensions such a theory would be anomalous as we cannot cancel the R-R charge: both the O-plane and D9-branes have positive R-R charges. Here we can ask if we could cancel the R-R charges by introducing anti-D9-branes, which we will denote via D$`\overline{9}`$. Each D$`\overline{9}`$-brane has R-R charge $`1`$ and NS-NS charge $`+1`$. So if we introduce 32 D$`\overline{9}`$-branes together with the aforementioned O-plane, we can cancel the R-R tadpoles, so that the resulting theory would be anomaly free. However, this theory would be non-supersymmetric, and NS-NS tadpoles would not be canceled.
Let us systematize the above discussion. Let us denote the NS-NS and R-R charges of a given object by $`Q_{\mathrm{NS}}`$ and $`Q_\mathrm{R}`$, respectively. Then D9-branes and D$`\overline{9}`$-branes have the following $`(Q_{\mathrm{NS}},Q_\mathrm{R})`$ charges:
$`\mathrm{D9}\mathrm{brane}:(+1,+1),`$ (35)
$`\mathrm{D}\overline{9}\mathrm{brane}:(+1,1).`$ (36)
We can introduce four different types of O9-planes according to their NS-NS and R-R charges:
$`\mathrm{O9}^{}:(32,32)\mathrm{\Omega }_{},`$ (37)
$`\mathrm{O9}^{++}:(+32,+32)\mathrm{\Omega }_+,`$ (38)
$`\mathrm{O9}^+:(+32,32)\mathrm{\Omega }_+(1)^{F_L+F_R},`$ (39)
$`\mathrm{O9}^+:(32,+32)\mathrm{\Omega }_{}(1)^{F_L+F_R},`$ (40)
where at the end of each line we have indicated the type of the Type IIB orientifold which produces each of these objects ($`F_L`$ and $`F_R`$ are the usual left- and right-moving space-time fermion numbers). We will discuss the above four orientifolds more explicitly in a moment. However, before we do this, let us discuss the action of the orientifold projections on D9- and D$`\overline{9}`$-branes induced by the above orientifold planes. In the following “S” refers to symmetrization on Chan-Paton charges, while “A” refers to antisymmetrization. The first entry corresponds to the NS (that is, space-time bosonic) open string sector on the corresponding D9/D$`\overline{9}`$-branes, while the second entry corresponds to the R (that is, space-time fermionic) sector. The orientifold projections of the above O9-planes on the D9-branes are given by:
$`\mathrm{O9}^{}:(A,A),(32+M,32+M),`$ (41)
$`\mathrm{O9}^{++}:(S,S),(32+M,32+M),`$ (42)
$`\mathrm{O9}^+:(S,A),(32+M,32+M),`$ (43)
$`\mathrm{O9}^+:(A,S),(32+M,32+M),`$ (44)
where at the end of each line we have indicated the total NS-NS and R-R charges of the background with the corresponding O9-plane and $`M`$ D9-branes. Let us discuss these four cases in more detail.
$``$ O9-plane plus $`M`$ D9-branes. The theory is supersymmetric, the gauge group is $`SO(M)`$, and the fermions on D9-branes are in the antisymmetric representation. The R-R tadpoles cancel for $`M=32`$, that is, we have an anomaly free theory in this case. Note that NS-NS tadpoles also cancel. This is the familiar Type I theory with $`SO(32)`$ gauge symmetry.
$``$ O9<sup>++</sup>-plane plus $`M`$ D9-branes. The theory is supersymmetric, the gauge group is $`Sp(M)`$, and the fermions on D9-branes are in the symmetric representation. The R-R (and NS-NS) tadpoles, however, cannot be canceled, so that the theory is anomalous.
$``$ O9<sup>+-</sup>-plane plus $`M`$ D9-branes. The closed string sector is supersymmetric at the tree level. The open string sector is non-supersymmetric. The gauge group is $`Sp(M)`$, and the fermions on the D9-branes are in the antisymmetric representation. The R-R tadpoles cancel for $`M=32`$, that is, we have an anomaly free theory in this case. This is the non-supersymmetric $`Sp(32)`$ string theory recently proposed in . Note, however, that NS-NS tadpoles do not cancel in this theory, so that the flat Minkowski metric with constant dilaton (more precisely, the corresponding $`𝒩=1`$ supergravity) does not appear to be the correct background for this theory. At present it is unclear whether there is a consistent background for this theory which could be reached via the Fischler-Susskind mechanism.
$``$ O9<sup>-+</sup>-plane plus $`M`$ D9-branes. The closed string spectrum is supersymmetric at the tree level. The open string sector is non-supersymmetric. The gauge group is $`SO(M)`$, and the fermions on D9-branes are in the symmetric representation. The NS-NS tadpoles cancel for $`M=32`$, but the R-R tadpoles cannot be canceled, so that the theory is anomalous.
Next, let us discuss the orientifold projections induced by the above O9-planes on D$`\overline{9}`$-branes:
$`\mathrm{O9}^{}:(A,S),(32+M,32M),`$ (45)
$`\mathrm{O9}^{++}:(S,A),(32+M,32M),`$ (46)
$`\mathrm{O9}^+:(S,S),(32+M,32M),`$ (47)
$`\mathrm{O9}^+:(A,A),(32+M,32M),`$ (48)
where, as before, at the end of each line we have indicated the corresponding total NS-NS and R-R tadpoles for the system of the corresponding O9-plane and $`M`$ D$`\overline{9}`$-branes. It is not difficult to see that the system of the O9<sup>α,β</sup>-plane plus $`M`$ D$`\overline{9}`$-branes gives the same theory as the system of the O9<sup>α,-β</sup>-plane plus $`M`$ D9-branes, where $`\alpha ,\beta =\pm `$.
Here we would like to make one remark. In the above language we can describe the Type O open plus closed string theory, which is the $`\mathrm{\Omega }_{}`$ orientifold of Type 0B string theory, as follows. Type 0B can be viewed as the $`(1)^{F_L+F_R}`$ orbifold of Type IIB. This implies that Type O can be viewed as a Type IIB orientifold, where the orientifold group is $`𝒪=\{1,(1)^{F_L+F_R},\mathrm{\Omega }_{},\mathrm{\Omega }_{}(1)^{F_L+F_R}\}`$. Thus, we have two orientifold planes in this theory, namely, the O9-plane and the O9<sup>-+</sup>-plane, which together have the following NS-NS and R-R charges: $`(Q_{\mathrm{NS}},Q_R)=(64,0)`$. These can be canceled by introducing 32 D9-branes together with 32 D$`\overline{9}`$-branes<sup>‡‡</sup><sup>‡‡</sup>‡‡In Type 0B, unlike Type IIB, there are two different types of D-branes and anti-D-branes, so other solutions to the tadpole cancellation conditions also exist .. Note that in this language, for instance, it is evident why there are no fermionic states in the open string spectrum: the O9 orientifold projection on D9-brane fermions is antisymmetric, while the O9<sup>-+</sup> orientifold projection on the same fermions is symmetric, so that fermions are completely projected out. The same conclusion holds for the D$`\overline{9}`$-brane fermions.
The reason why we discussed four different kinds of O-planes is that in our discussion of orientifolds with $`B`$-flux we will encounter O-planes with the $`Sp`$ type of orientifold projection. However, as we have already pointed out, one must distinguish two possible O-planes of this type as well as two O-planes with the $`SO`$ type of orientifold projection. In the following we will only encounter O-planes of O and O<sup>++</sup> type<sup>\**</sup><sup>\**</sup>\**In the following we will discuss O$`p`$-planes with $`p<9`$ for which the above discussion of the possible types of O-planes can be straightforwardly generalized. In particular, we can start from the aforementioned four types of O9-planes and use the standard T-duality arguments to arrive at the corresponding lower dimensional O$`p`$-planes.. We will refer to them as O<sup>-</sup> and O<sup>+</sup>, respectively. Also, just as in the previous sections, we will refer to $`\mathrm{\Omega }_{}`$ as $`\mathrm{\Omega }`$. Note that the NS-NS and R-R charges for each of these O-planes are equal. The same holds for D-branes (we will not need to introduce anti-D-branes in the following), so in the following it will suffice to just consider R-R charges. This is simply a manifestation of the fact that all the theories we will consider in the following are supersymmetric.
Next, we would like to return to orientifolds with $`B`$-flux, and consider cases where O-planes and the corresponding D-branes are transverse to tori with non-zero $`B`$-flux. To begin with, let us start with the simplest case. Thus, consider the $`\mathrm{\Omega }R(1)^{F_L}`$ orientifold of Type IIB on $`𝐑^{1,7}T^2`$, where $`R`$ acts as $`RX_{1,2}=X_{1,2}`$ on the compact coordinates, and the $`B`$-flux on $`T^2`$ is zero. To consider the case with non-zero $`B`$-flux, let us use the $`𝐙_2𝐙_2`$ freely acting orbifold construction discussed in the previous subsection. In fact, we will discuss both cases with and without discrete torsion, whose comparison will be helpful in understanding the types of orientifold planes arising in the former case.
Thus, let us consider the $`𝐙_2𝐙_2`$ orbifold of $`T^2`$ whose generators $`S_1`$ and $`S_2`$ act as shifts $`S_iX_i=X_i+e_i/2`$. Note that there are four fixed points of $`R`$ on $`T^2`$ located at $`(X_1,X_2)=(0,0),(e_1/2,0),(0,e_2/2),(e_1/2,e_2/2)`$. However, these four fixed points are identified by the combined action of the $`S_1`$ and $`S_2`$ shifts, so that we expect one orientifold 7-plane at $`(X_1,X_2)=(0,0)`$. On the other hand, we have additional fixed points due to the elements $`RS_a`$, $`a=1,2,3`$ (recall that $`S_3S_1S_2`$). In particular, for a given element $`RS_a`$ we have four fixed points, which are identified by the shifts $`S_1`$ and $`S_2`$. So we have one independent fixed point corresponding to each element $`RS_a`$: $`(X_1,X_2)=(e_1/4,0)`$ for $`RS_1`$, $`(X_1,X_2)=(0,e_2/4)`$ for $`RS_2`$, and $`(X_1,X_2)=(e_1/4,e_2/4)`$ for $`RS_3`$. At each of these three fixed points we expect one orientifold 7-plane.
Next, we would like to understand what types of O7-planes we have in this background. In the case without discrete torsion we have an O7<sup>-</sup>-plane at each of the aforementioned four fixed points. In the case with discrete torsion, however, one of the O7-planes is of the O7<sup>+</sup> type, while the other three are of the O7<sup>-</sup> type. To show this, let us consider the Klein bottle in each case. Thus, we have four sectors of the $`𝐙_2𝐙_2`$ orientifold labeled by $`(\alpha ^1,\alpha ^2)=(0,0),(1,0),(0,1),(1,1)`$, and the left- and right-moving closed string momenta are given in these sectors by (2), where the windings $`n^i`$ are arbitrary integers, whereas the momenta $`m_i`$ are integers subject to the orbifold projections which can be read off (3). Let us denote the Klein bottle contribution coming from a given sector $`(\alpha ^1,\alpha ^2)`$ via $`𝒦(\alpha ^1,\alpha ^2)`$. Let us denote the contribution coming from the tree-level closed string exchange between the O7-plane located at the fixed point labeled by $`(\beta ^1,\beta ^2)`$ and the O7-plane located at the fixed point labeled by $`(\gamma ^1,\gamma ^2)`$ via $`\stackrel{~}{𝒦}(\beta ^1,\beta ^2;\gamma ^1,\gamma ^2)`$. Here the fixed point labeled by $`(\beta ^1,\beta ^2)`$ is given by $`(X_1,X_2)=(\beta ^1e_1/4,\beta ^2e_2/4)`$, $`\beta ^i=0,1`$. Then we have the following relation:
$$𝒦(\alpha ^1,\alpha ^2)=\underset{\beta ^1,\beta ^2=0,1}{}\stackrel{~}{𝒦}(\beta ^1,\beta ^2;\beta ^1+\alpha ^1(\mathrm{mod}2),\beta ^2+\alpha ^2(\mathrm{mod}2)).$$
(49)
Note that for $`ϵ=1`$ in the untwisted as well as twisted sectors the momenta $`m_i`$ are even integers. This implies that both the untwisted as well as twisted sector contributions $`𝒦(\alpha ^1,\alpha ^2)`$ are non-vanishing - the $`\mathrm{\Omega }R(1)^{F_L}`$ orientifold projection keeps states with zero momenta and arbitrary windings. This implies that all four orientifold planes must have the same R-R charge. This can be seen explicitly by extracting the massless R-R tadpoles in each of these sectors, which are actually identical. (To see this one can perform the Poisson resummation of the corresponding winding sums.) However, for $`ϵ=1`$ the situation is quite different. Thus, in the untwisted sector $`(\alpha ^1,\alpha ^2)=(0,0)`$ the momenta $`m_i`$ are even, so that the corresponding Klein bottle contribution $`𝒦(0,0)`$ is non-vanishing. On the other hand, in the twisted sectors $`(\alpha ^1,\alpha ^2)=(1,0),(0,1),(1,1)`$ we have the momenta $`(m_12𝐙,m_22𝐙+1)`$, $`(m_12𝐙+1,m_22𝐙)`$ and $`(m_12𝐙+1,m_22𝐙+1)`$, respectively. This implies that the corresponding Klein bottle contributions $`𝒦(\alpha ^1,\alpha ^2)`$ vanish, and so do the corresponding tadpoles. This is enough to deduce the R-R charges of the orientifold planes located at the four fixed points. Let $`Q_R(\beta ^1,\beta ^2)`$ be the R-R charge of the O7-plane located at the fixed point labeled by $`(\beta ^1,\beta ^2)`$. Then we have the following conditions:
$`{\displaystyle \underset{\beta ^1,\beta ^2=0,1}{}}Q_R^2(\beta ^1,\beta ^2)=4\times 8^2,`$ (50)
$`{\displaystyle \underset{\beta ^1,\beta ^2=0,1}{}}Q_R(\beta ^1,\beta ^2)Q_R(\beta ^1+\alpha ^1(\mathrm{mod}2),\beta ^2+\alpha ^2(\mathrm{mod}2))=0,(\alpha ^1,\alpha ^2)(0,0).`$ (51)
The first line follows from the fact that the untwisted sector contribution $`𝒦(0,0)`$ is the same as in the case without discrete torsion where all four orientifold 7-planes are of the same type, and carry the R-R charge $`8`$ or $`+8`$ depending on the choice of the orientifold projection. It is not difficult to see that the solutions to the above conditions correspond to having one O7-plane with R-R charge $`\pm 8`$ and the other three O7-planes with R-R charge $`8`$ (where the signs are correlated). The solution that we are interested in is the one where we have one O7<sup>+</sup>-plane and three O7<sup>-</sup>-planes as in this case we can cancel the total R-R charge of the O7-planes, which is $`16`$, by introducing 16 D7-branes<sup>\*†</sup><sup>\*†</sup>\*†For the other solution where we have one O7<sup>-</sup>-plane and three O7<sup>+</sup>-plane the total R-R charge is $`+16`$, so it cannot be canceled by introducing only D7-branes.. This is precisely the result we wished to show.
Note that unlike the case of, say, D9-branes wrapped on a two-torus with $`B`$-field, the “rank reduction” here is not due to non-commuting Wilson lines but rather the fact that we have different types of O7-planes whose total R-R charge adds up to $`16`$ (rather than $`32`$). In contrast, in the case of D9-branes wrapped on a two-torus with $`B`$-flux the O9-plane is of the O9<sup>-</sup> type, and the R-R charge cancellation requires 32 D9-branes. The rank reduction in this case is due to the non-commuting Wilson lines. The difference between these two cases is quite substantial. Thus, in the case of D9-branes the gauge symmetry is $`SO(16)𝐙_2`$ or $`Sp(16)𝐙_2`$. In the case of D7-branes the gauge symmetry is $`SO(16)`$ if all D7-branes are placed at one of the O7<sup>-</sup>-planes, and it is $`Sp(16)`$ if all D7-planes are placed at the O7<sup>+</sup>-plane<sup>\*‡</sup><sup>\*‡</sup>\*‡In these configurations tadpoles are not canceled locally, so one expects a varying dilaton background. Note that this is not the case in analogous configurations involving O$`p`$-planes and D$`p`$-branes with $`p<7`$ as they are non-dilatonic.. Thus, the gauge symmetry in the case of D7-branes does not contain a $`𝐙_2`$ subgroup present in the case of D9-branes.
Before we end this subsection, let us mention the generalization to higher tori. For instance, consider the case of $`T^4`$. For illustrative purposes let us actually concentrate on $`T^4=T^2T^2`$. We can obtain the background with rank $`b=4`$ $`B`$-flux via separately orbifolding the two $`T^2`$’s by the respective $`𝐙_2𝐙_2`$ actions with discrete torsion. Then, if we consider the $`\mathrm{\Omega }R`$ orientifold of Type IIB in this background, where $`R`$ inverts all four coordinates on $`T^4`$, we have 10 O5<sup>-</sup>-planes and 6 O5<sup>+</sup>-planes. Similarly, in the case of the $`\mathrm{\Omega }R(1)^{F_L}`$ orientifold of Type IIB on $`T^6`$ with $`b=6`$, where $`R`$ inverts all six coordinates on $`T^6`$, we have 36 O3<sup>-</sup>-planes and 28 O3<sup>+</sup>-planes. More generally, consider O$`p`$-planes transverse to $`𝐑^{9pd}T^d`$ with $`B`$-flux of rank $`b`$ on $`T^d`$ ($`bd`$). We have total of $`n_f=2^d`$ fixed points. Let the numbers of O$`p^\pm `$-planes be $`n_{f\pm }`$. Then $`n_f=n_{f+}+n_f`$, and $`n_fn_{f+}=2^{db/2}`$. Thus, we have
$$n_{f\pm }=\frac{1}{2}\left(2^d2^{db/2}\right).$$
(52)
It is also straightforward to consider the cases where D-branes and O-planes are wrapped on tori with non-zero $`B`$-field, and, at the same time, the tori transverse to these objects also have non-zero $`B`$-field by combining the results of this and the previous subsections.
### C Mixed Cases
In this subsection we would like to discuss the cases where we have a D$`p`$-D$`p^{}`$ system (with $`pp^{}=0(\mathrm{mod}4)`$ so that we preserve some supersymmetries), where one set of branes is wrapped on a torus with non-zero $`B`$-field, while the other set of branes is transverse to this torus. The most general case is straightforward to treat, however, for illustrative purposes let us consider the case of the D5-D9 system, where D9-branes are wrapped on $`T^4`$, while D5-branes are transverse to $`T^4`$.
Thus, let us assume that we have $`n_5`$ D5-branes and $`n_9`$ D9-branes. In fact, in this subsection we will not worry about tadpole/anomaly cancellation as the point we would like to make here is purely geometric, and the former issues are irrelevant for this discussion. So we will not introduce any orientifold planes. Moreover, here we will focus on the 59 sector states, so that the conclusions we draw in this subsection are unchanged even after introduction of orientifold planes (recall, for instance, that the 59 sector states do not contribute to the Möbius strip amplitude).
To begin with, let us consider the above system with zero $`B`$-flux on $`T^4`$. Then the gauge group is $`U(n_5)U(n_9)`$, and the 59 open string sector hypermultiplets transform in the bifundamental representation $`(n_5,n_9)`$ (we will suppress the $`U(1)`$ charges which are straightforward to restore). Now let us turn on a half-integer $`B`$-flux of rank $`b=2`$. For the sake of simplicity let us consider the case of $`T^4=T^2T^2`$ with non-zero $`B`$-flux in the first $`T^2`$. Then, as in the previous subsections, we can view D9-branes wrapped on $`T^4`$ with $`B`$-flux in terms of turning on non-commuting Wilson lines corresponding to the two cycles of the first $`T^2`$. The 55 gauge group is unchanged as the Wilson lines are turned on in the directions transverse to D5-branes. The 99 gauge group, however, is broken down to $`U(N)𝐙_2`$, where $`Nn_9/2`$. (Recall that the first Wilson line brakes $`U(n_9)`$ to $`U(N)U(N)`$, while the second Wilson line breaks the latter to its diagonal subgroup $`U(N)𝐙_2`$.) Even though the 99 gauge group is broken by the Wilson lines, the 59 sector states are not affected (we will explain this in a moment). In particular, the number of the 59 hypermultiplets is still $`n_5n_9`$. More concretely, we have two hypermultiplets in $`(n_5,N)`$ of $`U(n_5)U(N)`$. Actually, the gauge group is $`U(n_5)U(N)𝐙_2`$, and the 59 hypermultiplets are given by $`(n_5,N)_{+1}`$ and $`(n_5,N)_1`$, where the subscript indicates the corresponding $`𝐙_2`$ charge. This is precisely the phenomenon first observed in \- the 59 open string states come with a non-trivial multiplicity, which depends on the rank of the $`B`$-field. This multiplicity is given by
$$\xi _{59}=2^{b/2}.$$
(53)
Here we can understand this multiplicity completely geometrically - it corresponds to the $`𝐙_2^{(b/2)}`$ discrete gauge symmetry, that is, the $`2^{b/2}`$ states, which have otherwise identical quantum numbers, carry all different quantum numbers under the $`𝐙_2^{(b/2)}`$ discrete gauge symmetry. This answers the question of the vertex operators in the 59 sector. For instance, in the $`b=2`$ case we can write down the vertex operators for the 59 states as follows. The first Wilson line breaks $`U(n_9)`$ to $`U(N)U(N)`$. The original hypermultiplet in the bifundamental $`(n_5,n_9)`$ of $`U(n_5)U(n_9)`$ thus gives rise to the following hypermultiplets charged under $`U(n_5)U(N)U(N)`$: $`(n_5,N,1)`$ and $`(n_5,1,N)`$. Next, the second Wilson line breaks $`U(N)U(N)`$ down to its diagonal subgroup $`U(N)𝐙_2`$. The aforementioned hypermultiplets can now be combined into two linear combinations
$$|(n_5,N)_{\pm 1}=\frac{1}{\sqrt{2}}\left(|(n_5,N,1)\pm |(n_5,1,N)\right),$$
(54)
which carry definite $`U(n_5)U(N)𝐙_2`$ gauge quantum numbersActually, the appearance of the $`𝐙_2^{(b/2)}`$ discrete gauge symmetry in orientifolds with $`B`$-flux was originally pointed out in ..
Now let us explain why the 59 sector states are not affected by turning on Wilson lines even though the 99 sector states are. The point is that the 59 sector states have no momentum (or winding) excitations - the 99 strings have only momenta on $`T^4`$, while the 55 strings have only windings on $`T^4`$ (this is precisely why 55 sector states are unaffected by the Wilson lines which act only on the momenta). Thus, the Wilson lines do not act on the 59 sector states.
Before we end this section we would like to discuss one other point. As we have seen from the previous discussions, the 99 and 55 sectors feel the presence of the $`B`$-flux in qualitatively different ways. Thus, the orientifold 9-planes are unaffected by the presence of the $`B`$-field, while the D9-brane gauge symmetry suffers rank reduction. On the other hand, the structure of orientifold 5-planes is modified in the presence of the $`B`$-flux, while D5-branes are unaffected. More generally, the above conclusions hold for O-planes and D-branes wrapped on tori with $`B`$-flux vs. O-planes and D-branes transverse to such tori. At first this might seem puzzling in the light of T-duality, which one might expect to map the aforementioned two setups into each other. However, as we will see in a moment, such an expectation would be erroneous, and there is no puzzle here.
To understand this, let us consider the simplest case of D9-branes wrapped on $`T^2`$ with the $`B`$-flux $`B_{12}=1/2`$. Let the metric on $`T^2`$ be $`g_{ij}`$. In the following it will be convenient to work with $`G_{ij}2g_{ij}`$, and introduce the following matrix (here we are closely following the discussion in ):
$$E_{ij}G_{ij}+B_{ij}=\left(\begin{array}{cc}G_{11}& G_{12}+B_{12}\\ G_{12}B_{12}& G_{22}\end{array}\right).$$
(55)
The T-duality group in the case of $`T^2`$ is $`SO(2,2,𝐙)`$ whose elements can be described in terms of $`4\times 4`$ matrices
$$\left(\begin{array}{cc}\alpha & \beta \\ \gamma & \delta \end{array}\right),$$
(56)
where $`\alpha ,\beta ,\gamma ,\delta `$ are $`2\times 2`$ matrices with integer entries, and satisfy the following constraints:
$`\gamma ^T\alpha +\alpha ^T\gamma =0,`$ (57)
$`\delta ^T\beta +\beta ^T\delta =0,`$ (58)
$`\gamma ^T\beta +\alpha ^T\delta =I.`$ (59)
Here the superscript $`T`$ stands for transposition, and $`I`$ denotes the $`2\times 2`$ identity matrix. The above T-duality element acts on $`E_{ij}`$ as follows:
$$EE^{}=(\alpha E+\beta )(\gamma E+\delta )^1.$$
(60)
In these notations the familiar $`S`$\- and $`T`$-transformations are described as follows. The $`S`$-transformation corresponds to taking $`\alpha =0`$, $`\beta =I`$, $`\gamma =I`$ and $`\delta =0`$, and amounts to mapping a 2-torus with metric $`G`$ and zero $`B`$-field to another 2-torus (with zero $`B`$-field) whose metric is given by the inverse of $`G`$. This is just the usual “$`R1/R`$” type of T-duality transformation. On the other hand, the $`T`$-transformation corresponds to taking $`\alpha =I`$, $`\beta =\mathrm{\Sigma }`$, $`\gamma =0`$ and $`\delta =I`$, and amounts to unit shifts of the $`B`$-field (but does not affect the metric on $`T^2`$), where $`\mathrm{\Sigma }`$ is the $`2\times 2`$ antisymmetric matrix with $`\mathrm{\Sigma }_{12}=1`$.
Here we can ask what happens to D9-branes wrapped on $`T^2`$ with $`B_{12}=1/2`$ under the $`S`$-transformation. Note that for the $`S`$-transformation the new matrix $`E^{}`$ is simply the inverse of $`E`$, so that we have
$$E^{}=G^{}+B^{}=\frac{1}{det(G)+B_{12}^2}\left(\begin{array}{cc}G_{22}& G_{12}B_{12}\\ G_{12}+B_{12}& G_{11}\end{array}\right).$$
(61)
Thus, the new $`B`$-field has the non-zero component given by
$$B_{12}^{}=\frac{B_{12}}{det(G)+B_{12}^2}.$$
(62)
Since $`B_{12}=1/2`$, and we must require that $`B_{12}^{}`$ must also be half-integer (or else the orientifold in the T-dual picture would not be well defined), it follows that the $`S`$-transformation can only be performed forActually, there are additional solutions to this constraint, but all of them are equivalent to the one we discuss here by T-duality transformations. Thus, for instance, we can start from the point $`det(G)=1/12`$, $`B_{12}=1/2`$, which is mapped by the $`S`$-transformation to the point $`det(G^{})=3/4`$, $`B_{12}^{}=3/2`$. However, the point $`det(G)=1/12`$, $`B_{12}=1/2`$ is equivalent to the point $`det(G)=3/4`$, $`B_{12}=1/2`$ via the T-duality transformation $`STS`$. I would like to thank Ofer Aharony for bringing this point to my attention. $`det(G)=3/4`$. The point in the moduli space of Kähler structures of $`T^2`$ where $`B_{12}=1/2`$ and $`det(G)=3/4`$ is one of the self-dual points. That is, the momentum and winding states at this point are indistinguishable, and, therefore, D9-branes wrapped on such a 2-torus are indistinguishable from D7-branes transverse to such a torus<sup>\*∥</sup><sup>\*∥</sup>\*∥As we will explain in a moment, however, because of half-integer $`B`$-field the precise map between these objects is different from that at the self-dual point without the $`B`$-field.. Thus, the $`S`$-transformation, which maps D9-branes to D7-branes, can only be performed at the self-dual point where the two types of branes are identical. This avoids the aforementioned puzzle with T-duality. Indeed, at a generic point in the Kähler structure moduli space D9-branes wrapped on $`T^2`$ with $`B`$-flux are not T-dual to D7-branes transverse to (another) $`T^2`$ with $`B`$-flux. As to the self-dual point in the Kähler structure moduli space, there are two a priori consistent setups, one of which can be continuously deformed into D9-branes wrapped on a generic $`T^2`$ with half-integer $`B`$-flux, and the other one can be continuously deformed into D7-branes transverse to a generic $`T^2`$ with half-integer $`B`$-flux. The fact that these two setups are indeed different is evident - at generic points in the respective moduli spaces the number of D9-branes is twice the number of D7-branes for a given rank of the gauge group.
Even though the difference between the aforementioned two setups is evident, we would like to understand the origin of these two choices<sup>\***</sup><sup>\***</sup>\***Parts of our discussion here have appeared in a footnote in .. To do this, let us consider D9-branes wrapping $`T^2`$ with $`B_{12}=1/2`$. Following we can assume that the closed strings that couple to D-branes satisfy the “no momentum flow” condition in the directions of $`T^2`$, that is, these states have the left- and right-moving momenta (note that $`p_i^{L,R}2e_iP_{L,R}`$)
$`p_i^L=m_i+E_{ji}n^j,`$ (63)
$`p_i^R=m_iE_{ij}n^j,`$ (64)
which satisfy
$$p_i^L=p_i^R.$$
(65)
This constraint implies that the closed strings coupled to D9-branes have the momenta and windings such that
$$m_iB_{ij}n^j=0,$$
(66)
from which it follows that the windings $`n^i`$ must be even<sup>\*††</sup><sup>\*††</sup>\*††This is in accord with (17). More precisely, (17) gives the loop-channel annulus amplitude. Upon the modular transformation $`t1/t`$, where $`t`$ is the proper time on the cylinder, which involves the appropriate Poisson resummation of the momentum sum in (17), we arrive at the tree-channel annulus amplitude corresponding to the closed string exchanges between D-branes. The latter amplitude is in agreement with the aforementioned conclusion that the closed string states that couple to D-branes have even windings. Note that, as we explained in detail in subsection A, strictly speaking the interpretation corresponding to (17), which arises in the approach of , is somewhat imprecise. However, it suffices for our purposes here. The above analysis can be repeated in the language of the freely acting $`𝐙_2𝐙_2`$ orbifold discussed in the previous subsections, which gives the precise description of D-branes wrapped on $`T^2`$ with half-integer $`B`$-field. Here we have chosen the approach of for illustrative purposes, as it suffices to explain the point we are trying to make here.. This, in particular, leads to the rank reduction for the 99 Chan-Paton gauge group , which we have understood in terms of non-commuting Wilson lines in the previous subsections. Note that in this case we also have a consistent coupling between the D9-branes and the O9-plane. Indeed, the loop-channel Klein bottle amplitude receives contributions from the closed string states with the left- and right-moving momenta satisfying $`p_i^L=p_i^R`$. Thus, in the loop channel we have arbitrary momenta $`m_i`$ and zero windings $`n^i`$. After the modular transformation $`t1/t`$, which maps the loop-channel Klein bottle amplitude to the tree-channel Klein bottle amplitude, we obtain a sum over arbitrary integer windings with the metric $`g_{ij}`$. That is, the closed string states that couple to the O9-plane have the left- and right-moving momenta $`p_i^L=p_i^R=G_{ij}n^j`$, so that $`P_L=P_R=e_in^i`$, and $`P_L^2=P_R^2=g_{ij}n^in^j=\frac{1}{2}G_{ij}n^in^j`$. (Note that these states are the same as in the case without the $`B`$-field.) Thus, the closed string states that couple to D9-branes wrapped on $`T^2`$ with half-integer $`B`$-field are a subset of the closed string states that couple to the O9-plane, so that D9-branes consistently couple to the O9-planes, in particular, the Möbius strip amplitude is consistent<sup>\*‡‡</sup><sup>\*‡‡</sup>\*‡‡Once again, as we explained in subsection A, a more precise description of this coupling is given in terms of the freely acting $`𝐙_2𝐙_2`$ orbifold, but the above description is adequate for our purposes here..
However, a priori there is another choice we can make for the constraint on the left- and right-moving momenta of the closed string states that couple to D-branes. This second choice has been recently discussed in , and is given by<sup>†\*</sup><sup>†\*</sup>†\*There is a misprint in , which amounts to a missing minus sign in (67).
$$p_i^L=_{i}^{}{}_{}{}^{j}p_j^R,$$
(67)
where $`E^TE^1`$. Note that (67) reduces to (65) for $`B_{ij}=0`$, but is different otherwise. It is not difficult to show that the solution of (67) is as follows. The closed string states coupled to D-branes have zero momenta $`m_i`$ and arbitrary windings $`n^i`$ (and in this case we do not expect rank reduction). This implies that these states have the following left- and right-moving momenta:
$$p_i^L=E_{ji}n^j,p_i^R=E_{ij}n^j.$$
(68)
Note that these states have the expected property
$$P_L^2=P_R^2=\frac{1}{2}𝒢_{ij}n^in^j,$$
(69)
where
$$𝒢GBG^1B.$$
(70)
Note, however, that at generic points (for half-integer $`B`$-flux) these states are quite different from the states that couple to the O9-plane for which from the above discussion we have $`P_L^2=P_R^2=\frac{1}{2}G_{ij}n^in^j`$. In fact, these two sets of closed string states coincide only for zero $`B`$-flux. Thus, D9-branes defined via (67) cannot be consistently coupled to the O9-plane for half-integer $`B`$-field.
There is, however, a setup where we can consistently couple such D9-branes to orientifold planes, except that these are not O9- but O7-planes. Thus, consider D7-branes transverse to $`T^2`$ with half-integer $`B`$-flux. Recall that in the case of D9-branes wrapped on such a $`T^2`$ we have imposed the “no momentum flow” condition (65). In the case of D7-branes transverse to such a $`T^2`$ the analogous condition is that of “no winding flow” :
$$p_i^L=p_i^R.$$
(71)
This constraint implies that $`n^i`$ are zero and $`m_i`$ are arbitrary. Thus, the left- and right-moving momenta in this case are given by $`p_i^L=p_i^R=m_i`$, and
$$P_L^2=P_R^2=\frac{1}{2}G^{ij}m_im_j,$$
(72)
where $`G^{ij}`$ is the inverse of $`G_{ij}`$. Let us compare (69) and (72). They are identical at the self-dual point in the moduli space of Kähler structures where $`det(G)=3/4`$ and $`B_{12}=1/2`$. Indeed, note that at a generic point in this moduli space we have
$$𝒢_{ij}=\left(1+\frac{B_{12}^2}{det(G)}\right)G_{ij}.$$
(73)
At the self-dual point we have
$$𝒢_{ij}=\frac{4}{3}G_{ij}=\frac{4}{3}\left(\begin{array}{cc}G_{11}& G_{12}\\ G_{12}& G_{22}\end{array}\right).$$
(74)
On the other hand, for the inverse metric $`G^{ij}`$ at the self-dual point we have
$$G^{ij}=\frac{1}{det(G)}\left(\begin{array}{cc}G_{22}& G_{12}\\ G_{12}& G_{11}\end{array}\right)=\frac{4}{3}\left(\begin{array}{cc}G_{22}& G_{12}\\ G_{12}& G_{11}\end{array}\right).$$
(75)
Even though (74) and (75) generically are different, the corresponding squared momenta $`P_{L,R}^2`$ in (69) and (72) are the same. Moreover, the corresponding momentum states are the same once we make the appropriate identification
$$n^i=ϵ^{ij}m_j,$$
(76)
where $`ϵ^{ij}`$ is a unit antisymmetric $`2\times 2`$ matrix (that is, $`ϵ^{12}=ϵ^{21}=+1`$ or $`1`$). Thus, at the self-dual point (with half-integer $`B`$-field) D7-branes satisfying the “no winding flow” condition have the same spectrum as D9-branes satisfying the condition (67), and not the usual “no momentum flow” condition (65). At first this might seem a bit strange, as we can ask what is the analogous statement for D9-branes that satisfy the usual “no momentum flow condition” (65). The answer to this question, actually, is very simple. It is not difficult to show that D9-branes that satisfy the usual “no momentum flow” condition have the same spectrum as D7-branes satisfying the “no winding flow” condition at the self-dual point in the moduli space of Kähler structures corresponding to $`det(G)=1`$ and $`B_{12}=0`$.
Now everything falls in place, and we arrive at the following consistent picture. D-branes transverse to a torus are the same whether the $`B`$-field on the torus is zero or non-zero. However, D-branes wrapped on a torus are sensitive to whether the $`B`$-field is zero or non-zero. In the former case they are T-dual to the corresponding D-branes transverse to the torus even at generic points in the moduli space of Kähler structures, and at the self-dual point without the $`B`$-field the two types of D-branes are indistinguishable provided that they satisfy the usual “no momentum flow” and “no winding flow” conditions, respectively. In the case with non-zero $`B`$-field there is no T-duality between the two types of D-branes (in the presence of orientifold planes) at generic points of the Kähler structure moduli space as the T-duality $`S`$-transformation would map a torus with half-integer $`B`$-field into a torus with non-zero $`B`$-field which does not take half-integer values. At the self-dual point with half-integer $`B`$-field the $`S`$-transformation is consistent with the orientifold action, but precisely at this point D-branes transverse to the torus and D-branes wrapped on the torus are indistinguishable, except that the former still satisfy the usual “no winding flow” condition, while the latter satisfy the modified constraint (67). Here it is crucial that the latter type of D-branes cannot be consistently coupled to the orientifold planes of the same spatial dimensionality (but, at the self-dual point, couple consistently to the corresponding O-planes transverse to the torus as they are indistinguishable from the corresponding D-branes transverse to the torus). Note that above we arrived at these conclusions in the case of $`T^2`$, where we have two inequivalent self-dual points in the Kähler structure moduli space. The generalization to higher tori should be evident once we observe that even though the number of inequivalent self-dual points grows, the latter always have half-integer (or zero) $`B`$-field, so that different self-dual points correspond to different half-integer $`B`$-field configurations. The above discussion should make it evident that the geometric picture of orientifolds with $`B`$-flux we described in this section is completely consistent with T-duality considerations. In particular, the fact that D-branes transverse to a torus with $`B`$-flux behave quite differently from D-branes wrapped on such a torus is no longer mysterious.
Finally, let us ask to what extent T-duality can be useful in the usual sense. In particular, suppose we have D-branes wrapped on a small volume (in the string units) torus with non-zero $`B`$-field, and we would like to perform a T-duality transformation that maps this torus to a large volume torus. Is there such a T-duality transformation? This question has already been answered in , but we would like to reiterate this point here for the sake of completeness. Here we will be a bit more general, and closely follow the discussion in . Thus, consider D9-branes wrapped on $`T^2`$ with the metric $`G_{ij}`$ and $`B`$-field $`B_{ij}=B_{12}\mathrm{\Sigma }`$, where $`B_{12}=1/k`$, $`k𝐍\{1\}`$. Now consider the T-duality transformation (56) with
$$\alpha =I,\beta =0,\gamma =k\mathrm{\Sigma },\delta =I.$$
(77)
We will denote this T-duality transformation by $`P`$. The corresponding matrix $`E^{}=G^{}+B^{}`$ is given by
$$E^{}=BG^1BB=\frac{B_{12}^2}{det(G)}GB.$$
(78)
Thus, the T-duality transformation $`P`$ amounts to
$$GG^{}=\frac{B_{12}^2}{det(G)}G,BB^{}=B.$$
(79)
In particular, a small volume torus with $`B`$-flux is mapped to a large volume torus with opposite $`B`$-flux. (Note that $`det(G^{})=B_{12}^4/det(G)`$.)
Next, let us see what happens to D9-branes under the T-duality transformation $`P`$. It is not difficult to see that the transformation $`P`$ can be written as
$$P=ST^kS.$$
(80)
Under the first $`S`$-transformation D9-branes are mapped to D7-branes, the subsequent $`T`$-transformations do not affect the dimensionality of the branes, and the last $`S`$-transformation maps D7-branes back to D9-branes. Thus, the T-duality transformation $`P`$ maps D$`p`$-branes wrapped on a small volume $`T^2`$ with $`B`$-flux ($`B_{12}=1/k`$) to D$`p`$-branes wrapped on a large volume $`T^2`$ with opposite $`B`$-flux<sup>††</sup><sup>††</sup>††In fact, this was used in in the discussion of unification via Kaluza-Klein thresholds in gauge theories compactified on non-commutative tori, that is, tori with non-zero $`B`$-flux.. (Generalizations to higher tori should be clear.)
## III K3 Orientifolds with $`B`$-flux
Having understood toroidal orientifolds with $`B`$-flux, we would like to discuss K3 orientifolds with $`B`$-flux next. The setup of this section is mostly<sup>†‡</sup><sup>†‡</sup>†‡In the $`𝐙_3`$ cases we will also consider $`\mathrm{\Omega }R`$ orientifolds, where $`R`$ reverses the sign of all coordinates on K3. In such orientifolds we have D5-branes but no D9-branes. going to be the $`\mathrm{\Omega }`$ orientifold<sup>†§</sup><sup>†§</sup>†§In the $`𝐙_3,𝐙_4,𝐙_6`$ cases $`\mathrm{\Omega }`$ will actually be accompanied by an additional action as in \- see subsections B,C,D for details. of Type IIB on $`𝐑^{1,5}\mathrm{K3}`$, where $`\mathrm{K3}=T^4/𝐙_M`$. Here the orbifold group generator $`g`$ acts as $`gz_1=\omega z_1`$, $`gz_2=\omega ^1z_2`$, where $`z_{1,2}`$ are the complex coordinates parametrizing $`T^4`$, and $`\omega \mathrm{exp}(2\pi i/M)`$. Note that in order for the orbifold action to be crystallographic, $`M`$ must be 2,3,4 or 6. The resulting background has $`𝒩=1`$ supersymmetry in six dimensions, and contains D9-branes for $`M=3`$, and both D9- and D5-branes for $`M=2,4,6`$. K3 orientifolds without $`B`$-flux have been studied in detail in . The cases with rank $`b=2,4`$ $`B`$-flux in the directions of K3 were discussed in . Here we would like to revisit K3 orientifolds with $`B`$-flux using our improved understanding of their underlying geometric structure. In subsections A, B, C, D we will discuss the $`𝐙_2,𝐙_3,𝐙_6,𝐙_4`$ models, respectively. In subsection E we briefly summarize the results of this section.
### A The $`𝐙_2`$ Models
Let us first consider the $`\mathrm{\Omega }`$ orientifold of Type IIB on $`𝐑^{1,5}(T^4/𝐙_2)`$, where the generator $`R`$ of $`𝐙_2`$ acts as $`Rz_{1,2}=z_{1,2}`$ on the complex coordinates parametrizing $`T^4`$. In fact, for our purposes here it will suffice to consider $`T^4=T^2T^2`$, where the first and the second 2-tori are parametrized by $`z_1`$ and $`z_2`$, respectively.
The orientifold group is $`𝒪=\{1,R,\mathrm{\Omega },\mathrm{\Omega }R\}`$. Here and in the following $`\mathrm{\Omega }=\mathrm{\Omega }_{}`$, that is, $`\mathrm{\Omega }`$ induces the $`SO`$ type of orientifold projection on D9-branes. Note that the presence of $`\mathrm{\Omega }`$ among the orientifold group elements implies that we have the O9<sup>-</sup>-plane. Suppose the $`B`$-field in the compact directions is trivial. Then the presence of the $`\mathrm{\Omega }R`$ orientifold group element implies that we have 16 O5<sup>-</sup>-planes as well. This has a non-trivial implication on the action of the $`𝐙_2`$ orbifold group element $`R`$ on the D9- and D5-brane Chan-Paton charges. Let $`\gamma _{R,9}`$ and $`\gamma _{R,5}`$ be the corresponding Chan-Paton matrices. (Here for definiteness we will assume that all D5-branes are placed at the same O5<sup>-</sup>-plane located at the origin $`z_1=z_2=0`$.) Then we can show that (up to equivalent representations)
$$\gamma _{R,9}=\gamma _{R,5}=i\sigma _3I_{16}.$$
(81)
To show this, let us first note that the presence of the O9<sup>-</sup>-plane implies that we must introduce 32 D9-branes to cancel the 10-form R-R charge. Similarly, the presence of 16 O5<sup>-</sup>-planes implies that we must introduce 32 D5-branes to cancel the 6-form R-R charge. Thus, all Chan-Paton matrices will be $`32\times 32`$ dimensional. Next, the matrix $`\gamma _{\mathrm{\Omega },9}`$ is symmetric :
$$\gamma _{\mathrm{\Omega },9}^T=+\gamma _{\mathrm{\Omega },9}.$$
(82)
This follows from the fact that the orientifold projection on D9-branes is of the $`SO`$ type. (Note that in this case we always include the appropriate minus sign in the definition of the Möbius strip amplitude as in (19).) On the other hand, $`\gamma _{\mathrm{\Omega },5}`$ is antisymmetric :
$$\gamma _{\mathrm{\Omega },5}^T=\gamma _{\mathrm{\Omega },5}.$$
(83)
This follows from the fact that the $`\mathrm{\Omega }`$ projection on D5-branes must be of the $`Sp`$ type (which is consistent with ). One way to see this is to consider the action of $`\mathrm{\Omega }^2`$ in the 59 sector, where $`\mathrm{\Omega }^2=1`$ (note that in the 99 and 55 sectors $`\mathrm{\Omega }^2=+1`$) . On the other hand, the orientifold 5-planes are of the O5<sup>-</sup> type. This implies that $`\gamma _{\mathrm{\Omega }R,5}`$ must be symmetric:
$$\gamma _{\mathrm{\Omega }R,5}^T=+\gamma _{\mathrm{\Omega }R,5}.$$
(84)
From (83) and (84) together with the fact that we must have (the choice of the sign is immaterial)
$$\gamma _{\mathrm{\Omega }R,5}=\pm \gamma _{\mathrm{\Omega },5}\gamma _{R,5},$$
(85)
we obtain
$$\gamma _{R,5}^T=\gamma _{\mathrm{\Omega },5}\gamma _{R,5}\gamma _{\mathrm{\Omega },5}.$$
(86)
Now consider the basis where $`\gamma _{R,5}`$ is diagonal. Then $`\gamma _{R,5}^T=+\gamma _{R,5}`$, which together with (86) and the fact that we must have
$$\gamma _{\mathrm{\Omega },5}^2=\gamma _{\mathrm{\Omega }R,5}^2=1,$$
(87)
implies that
$$\gamma _{R,5}^2=1.$$
(88)
This implies that the eigenvalues of the matrix $`\gamma _{R,5}`$ are $`\pm i`$. For the $`𝐙_2`$ orbifold projection in the 59 open string sector to be consistent (that is, so that we get only $`𝐙_2`$ valued phases from the action of the orbifold group on the Chan-Paton charges in the 59 sector), we must then require that the eigenvalues of the $`\gamma _{R,9}`$ matrix must also be $`\pm i`$, so that
$$\gamma _{R,9}^2=1.$$
(89)
Note that this is consistent with the analog of (83) for the D9-branes, namely,
$$\gamma _{\mathrm{\Omega }R,9}^T=\gamma _{\mathrm{\Omega }R,9}.$$
(90)
Finally, the twisted tadpole cancellation condition implies that (in the aforementioned setup) the matrices $`\gamma _{R,9}`$ and $`\gamma _{R,5}`$ are traceless . This then implies (81).
Note that in the above discussion we have assumed that the O5-planes are of the O5<sup>-</sup> type. The physical reason for this is clear - had we assumed that the O5-planes were of the O5<sup>+</sup> type, we would not have been able to cancel all tadpoles by introducing D5-branes. Another important point is that the twisted Chan-Paton matrices $`\gamma _{R,9}`$ and $`\gamma _{R,5}`$ both must have eigenvalues $`\pm i`$ or $`\pm 1`$, which, as we have explained above, follows from the requirement that the $`𝐙_2`$ orbifold projection be consistent in the 59 sector. In the above model (with trivial $`B`$-field) we must choose these matrices to have eigenvalues $`\pm i`$ \- had we chosen $`\gamma _{R,5}`$ with eigenvalues $`\pm 1`$, we would have found that the O5-planes are of the O5<sup>+</sup> type. In fact, the constraint (89) can be alternatively derived by considering the Möbius strip amplitude in this model. Thus, we can organize the latter into four terms according to the Chan-Paton factors that multiply them:
$`=\left({\displaystyle \frac{1}{2}}\right)^2[`$ $`\mathrm{Tr}(\gamma _{\mathrm{\Omega },9}^1\gamma _{\mathrm{\Omega },9}^T)𝒵(\mathrm{\Omega },9)+\mathrm{Tr}(\gamma _{\mathrm{\Omega }R,9}^1\gamma _{\mathrm{\Omega }R,9}^T)𝒵(\mathrm{\Omega }R,9)+`$ (92)
$`\mathrm{Tr}(\gamma _{\mathrm{\Omega }R,5}^1\gamma _{\mathrm{\Omega }R,5}^T)𝒵(\mathrm{\Omega }R,5)+\mathrm{Tr}(\gamma _{\mathrm{\Omega },5}^1\gamma _{\mathrm{\Omega },5}^T)𝒵(\mathrm{\Omega },5)],`$
where the overall factor of $`(1/2)^2`$ arises due to the orientifold and orbifold projections. Let us discuss each term in the above Möbius strip amplitude. The first term containing $`𝒵(\mathrm{\Omega },9)`$ when rewritten in the closed string tree-channel corresponds to the closed string exchange between D9-branes and the O9<sup>-</sup>-plane. In fact, (82) is precisely the statement that the O9-plane is of the O9<sup>-</sup> type (note the overall minus sign in the definition of the Möbius strip amplitude in (92)). Similarly, the term containing $`𝒵(\mathrm{\Omega }R,5)`$ when rewritten in the closed string tree-channel corresponds to the closed string exchange between D5-branes and the O5-planes. Moreover, the characters $`𝒵(\mathrm{\Omega },9)`$ and $`𝒵(\mathrm{\Omega }R,5)`$ are actually identical except for the corresponding momentum respectively winding sums (that is, the string oscillator contributions to these characters are identical). Note that (84) is the statement that the O5-planes are of the O5<sup>-</sup> type. Next, consider the term containing $`𝒵(\mathrm{\Omega }R,9)`$. The latter corresponds to the 99 sector Möbius contribution with the $`𝐙_2`$ orbifold projection inserted on the boundary. Similarly, the term containing $`𝒵(\mathrm{\Omega },5)`$ corresponds to the 55 Möbius contribution with the $`𝐙_2`$ orbifold projection inserted on the boundary. In fact, the characters $`𝒵(\mathrm{\Omega }R,9)`$ and $`𝒵(\mathrm{\Omega },5)`$ are actually identical<sup>†¶</sup><sup>†¶</sup>†¶These characters actually vanish. More precisely, all the characters vanish by supersymmetry. However, the bosonic and fermionic pieces in the aforementioned characters vanish separately. Thus, in the open string loop-channel the NS and R contributions to these characters vanish separately. Moreover, in the closed string tree-channel the NS-NS and R-R contributions to these characters also vanish separately. This, in particular, implies that the corresponding terms in the Möbius strip amplitude do not contribute to massless tadpoles. In fact, these terms are often dropped when discussing this model (as in, e.g., ). However, these terms are important to keep when discussing the spectrum and vertex operators in this model. In particular, one has to make sure that the $`𝐙_2`$ orbifold projection is consistent, which is precisely the constraint we are going to discuss in a moment.. Putting all of the above together, we can now derive a non-trivial constraint on $`\gamma _{R,9}`$. Thus, note that the traces $`\mathrm{Tr}(\gamma _{\mathrm{\Omega },9}^1\gamma _{\mathrm{\Omega },9}^T)=\mathrm{Tr}(\gamma _I)=32`$ and $`\mathrm{Tr}(\gamma _{\mathrm{\Omega }R,5}^1\gamma _{\mathrm{\Omega }R,5}^T)=\mathrm{Tr}(\gamma _I)=32`$ in front of the characters $`𝒵(\mathrm{\Omega },9)`$ respectively $`𝒵(\mathrm{\Omega }R,5)`$ are identical. This implies that the traces $`\mathrm{Tr}(\gamma _{\mathrm{\Omega }R,9}^1\gamma _{\mathrm{\Omega }R,9}^T)=\mathrm{Tr}(\gamma _{R,9}^2)`$ and $`\mathrm{Tr}(\gamma _{\mathrm{\Omega },5}^1\gamma _{\mathrm{\Omega },5}^T)=\mathrm{Tr}(\gamma _I)=32`$ in front of the characters $`𝒵(\mathrm{\Omega }R,9)`$ respectively $`𝒵(\mathrm{\Omega },5)`$ must also be identical for the $`𝐙_2`$ orbifold projection in the 99 sector to be consistent with that in the 55 sector. This then implies (89). As we will see in the following, requiring consistency of the orbifold projection in the Möbius strip amplitude will result in additional non-trivial constraints in the models with non-zero $`B`$-flux.
Before we consider the cases with non-zero $`B`$-flux, for later convenience let us review the $`𝐙_2`$ model of without the $`B`$-field. The closed string sector contains the six dimensional $`𝒩=1`$ supergravity multiplet, one untwisted (self-dual) tensor supermultiplet, 4 untwisted hypermultiplets, and 16 twisted hypermultiplets. Note that the 16 fixed points of the $`𝐙_2`$ orbifold give rise to hypermultiplets but no (anti-self-dual) tensor multiplets as all 16 fixed points are even under the orientifold action. This follows from the fact that all 16 orientifold 5-planes located at the fixed points are of the O5<sup>-</sup> type. Next, let us discuss the open string spectrum. The gauge group is $`U(16)_{99}U(16)_{55}`$, and the massless matter consists of the following hypermultiplets:
$`2\times `$ $`(\mathrm{𝟏𝟐𝟎};\mathrm{𝟏})_{99},`$ (93)
$`2\times `$ $`(\mathrm{𝟏};\mathrm{𝟏𝟐𝟎})_{55},`$ (95)
$`(\mathrm{𝟏𝟔};\mathrm{𝟏𝟔})_{95}.`$
Here semi-colon separates the 99 and 55 gauge quantum numbers. Note that $`U(1)`$’s are actually anomalous, and are broken via the generalized Green-Schwarz mechanism , so that the gauge group is actually $`SU(16)_{99}SU(16)_{55}`$.
Here we would like to stress one important point. In particular, the $`\mathrm{\Omega }`$ projection on D5-branes is of the $`Sp`$ type, while the O5-planes located at the 16 $`𝐙_2`$ fixed points are of the O5<sup>-</sup> type, that is, they induce the $`SO`$ type of orientifold projection on the Chan-Paton charges of D5-branes. The above discussion relates this to the fact that the latter projection is determined by $`\gamma _{\mathrm{\Omega }R,5}`$, and not by $`\gamma _{\mathrm{\Omega },5}`$. Nonetheless, at first it might appear a bit strange that the projection on D5-branes is of the $`SO`$ type - from the arguments of we would expect (subgroups of) symplectic gauge groups coming form D5-branes. In particular, appearance of antisymmetric representations (namely, $`\mathrm{𝟏𝟐𝟎}`$ of $`SU(16)`$) in the 55 sector might seem puzzling in the context of . However, there is no puzzle here as there is a geometric explanation of this point. To arrive at this explanation, however, we will first need to review the difference between the O5<sup>-</sup>\- and O5<sup>+</sup>-planes. Here we will be closely following the discussion in .
Thus, consider Type IIB on $`𝐑^{1,9}`$ in the presence of an O5-plane. The O5-plane is located at the fixed point at the origin of $`𝐑^4/𝐙_2`$, where the $`𝐙_2`$ action simultaneously reflects all four coordinates of $`𝐑^4`$ transverse to the O5-plane. The orientifold replaces a 3-sphere $`𝐒^3`$ around the origin of $`𝐑^4`$ by $`\mathrm{𝐑𝐏}^3=𝐒^3/𝐙_2`$. (Recall that the real projective $`n`$-space $`\mathrm{𝐑𝐏}^n`$ is defined as the quotient $`𝐒^n/𝐙_2`$ of the $`n`$-sphere $`𝐒^n`$ defined via $`_{i=1}^{n+1}x_i^2=\rho ^2`$, where the action of $`𝐙_2`$ on the coordinates $`x_i`$, $`i=1,\mathrm{},n+1`$, is given by $`x_ix_i`$, and $`\rho `$ is the radius of the $`n`$-sphere $`𝐒^n`$.) Now consider unorientable closed world-sheets $`\mathrm{\Sigma }=\mathrm{𝐑𝐏}^2`$. Such a world-sheet is embeddable in $`\mathrm{𝐑𝐏}^3`$. Thus, we can define a $`𝐙_2`$ charge for the O5-plane as follows. Consider a constant NS-NS $`B`$-flux. Then the world-sheets $`\mathrm{\Sigma }=\mathrm{𝐑𝐏}^2`$ contribute to the path integral with an extra phase
$$\mathrm{exp}\left(i_\mathrm{\Sigma }B\right),$$
(96)
which is $`+1`$ for the trivial $`B`$-flux, and $`1`$ for the half-integer $`B`$-flux (recall that the $`B`$-flux is quantized in the presence of an orientifold plane). The aforementioned $`𝐙_2`$ charge assignments are then as follows. In the former case we assign charge 0, while in the latter case we assign charge 1, and the $`𝐙_2`$ charge is defined modulo 2. One can then show that the O5-plane with the $`𝐙_2`$ charge 0 is of the O5<sup>-</sup> type, while the one with the charge 1 is of the O5<sup>+</sup> type . This, for instance, can be seen by considering the following BPS configuration with eight supercharges. Let the O5-plane fill the coordinates $`x_0,x_1,x_2,x_3,x_4,x_5`$ of $`𝐑^{1,9}`$, and a $`\frac{1}{2}`$NS5-brane<sup>†∥</sup><sup>†∥</sup>†∥In our conventions a $`\frac{1}{2}`$NS5-brane is the S-dual of a D5-brane, whose R-R charge is $`+1`$, while O5<sup>±</sup>-planes have the R-R charges $`\pm 2`$, respectively. Note that an NS5-brane is S-dual of a pair of D5-branes, which combine into a dynamical 5-brane - in the presence of an O5-plane D5-branes always move in pairs. fill the coordinates $`x_0,x_1,x_2,x_3,x_4,x_6`$. That is, the O5-plane and the $`\frac{1}{2}`$NS5-plane intersect at 90 degrees in the 56-plane. Now consider the origin of the 789 space transverse to both the O5-plane and the $`\frac{1}{2}`$NS5-brane. The orientifold replaces a 2-sphere around the origin of this space by $`\mathrm{𝐑𝐏}^2`$. The NS-NS $`B`$-flux couples magnetically to NS5-branes. Thus, with the appropriate normalization $`_{\mathrm{𝐑𝐏}^2}B`$ counts the number of $`\frac{1}{2}`$NS5-branes modulo 2. This implies that the aforementioned $`𝐙_2`$ charge of the O5-plane, which with this normalization can be identified with $`_{\mathrm{𝐑𝐏}^2}B`$, changes by 1 every time the O5-plane crosses the $`\frac{1}{2}`$NS5-brane<sup>†\**</sup><sup>†\**</sup>†\**Here we note that for an O$`p`$-plane with $`p5`$ one can define another $`𝐙_2`$ charge (in the appropriate normalization) via $`_{\mathrm{𝐑𝐏}^{5p}}C^{(5p)}`$ , where $`C^{(5p)}`$ is a Ramond-Ramond form. The relevant brane configuration here is that of an O$`p`$-plane intersecting with a D$`(p+2)`$-brane such that we have 8 unbroken supercharges. The aforementioned $`𝐙_2`$ charge then changes by 1 every time the O$`p`$-plane crosses the D$`(p+2)`$-brane. This additional $`𝐙_2`$ charge, however, will not be important in the subsequent discussions.. To see that the O5<sup>-</sup>-plane has the $`𝐙_2`$ charge 0, consider the T-dual version of the above setup, where we have an O3-plane. Then from Montonen-Olive self-duality of the $`SO(2k)`$ gauge theories we conclude that the O3<sup>-</sup>-plane must have zero $`𝐙_2`$ charge or else it would not be invariant under the $`SL(2,𝐙)`$ symmetry of Type IIB.
Now we can explain why the O5-planes are of the O5<sup>-</sup> type in the aforementioned model of without the $`B`$-field. Thus, naively we expect that the O5-planes are of the O5<sup>+</sup> type as the orientifold projection is of the $`SO`$ type on D9-branes implying that the orientifold projection is of the $`Sp`$ type on the D5-branes. This would require that the corresponding $`𝐙_2`$ charge related to the $`B`$-field for the orientifold planes is 1, that is, we have an odd-half-integer $`B`$-flux in $`\mathrm{𝐑𝐏}^2`$. However, the O5-planes are located at the $`𝐙_2`$ orbifold fixed points, and each $`𝐙_2`$ orbifold fixed point corresponds to a collapsed 2-sphere $`𝐏^1`$. As was pointed out in , in the conformal field theory $`𝐙_2`$ orbifold there is an odd-half-integer twisted $`B`$-field<sup>†††</sup><sup>†††</sup>†††The twisted $`B`$-field plays an important role in the context of orientifolds in a number of setups - see, e.g., and . stuck inside of each collapsed $`𝐏^1`$. The orientifold replaces these $`𝐏^1`$’s by $`\mathrm{𝐑𝐏}^2`$’s, so that we have an odd-half-integer twisted $`B`$-flux stuck inside of each collapsed $`\mathrm{𝐑𝐏}^2`$. This additional twisted $`B`$-flux converts the would-be O5<sup>+</sup>-planes into O5<sup>-</sup>-planes, which is the result we wished to explain.
Next, let us consider the $`\mathrm{\Omega }`$ orientifold of Type IIB on $`T^4/𝐙_2`$ in the presence of the $`B`$-field. Here we will mainly focus on the case of $`T^4=T^2T^2`$ with $`B`$-field of rank $`b=2`$ turned on in the directions of the first $`T^2`$ (while the $`B`$-field in the directions of the second $`T^2`$ is trivial). Generalizations to generic $`T^4`$’s with $`B`$-field of rank $`b=2`$ as well as $`b=4`$ should be evident.
In the case of $`B`$-field of rank $`b`$ we have $`n_f=8(1+1/2^{b/2})`$ O5<sup>-</sup>-planes and $`n_{f+}=8(11/2^{b/2})`$ O5<sup>+</sup>-planes. In the closed string $`𝐙_2`$ twisted sector the orientifold projection at the $`𝐙_2`$ orbifold fixed points where the O5<sup>-</sup>-planes are located gives rise to hypermultiplets, while at the fixed points where the O5<sup>+</sup>-planes are located it gives rise to (anti-self-dual) tensor multiplets. On the other hand, the untwisted closed string sectors of the models with $`b=2,4`$ are the same as that of the $`b=0`$ model. Thus, the closed string sector of the model with $`B`$-field of rank $`b`$ contains the six dimensional $`𝒩=1`$ supergravity multiplet, one untwisted (self-dual) tensor supermultiplet, 4 untwisted hypermultiplets, $`n_f`$ twisted hypermultiplets and $`n_{f+}`$ twisted (anti-self-dual) tensor multiplets.
Next, let us discuss the open string sector. As we have already mentioned, we will specialize to the case of $`b=2`$. First, let us understand the geometric structure of the orientifold planes. Let the vielbeins on the first $`T^2`$ (where we have non-zero $`B`$-flux) be $`e_i`$, $`i=1,2`$, while the vielbeins on the second $`T^2`$ (where the $`B`$-flux is trivial) be $`d_i`$, $`i=1,2`$. Let us define $`e_3e_1e_2`$, and $`d_3d_1d_2`$. The 16 $`𝐙_2`$ fixed points are located at $`(0,0),(e_a/2,0),(0,d_a/2),(e_a/2,d_b/2)`$, $`a,b=1,2,3`$. Without loss of generality we can choose the following distribution for the O5-planes. At $`(0,0),(0,d_a/2),(e_i/2,0),(e_i/2,d_a/2)`$ we have 12 O5<sup>-</sup>-planes, while at $`(e_3/2,0),(e_3/2,d_a/2)`$ we have 4 O5<sup>+</sup>-planes.
At first it might seem that the above setup is completely consistent, in particular, that there is no difficulty with having O5<sup>-</sup>\- and O5<sup>+</sup>-planes at the same time. Thus, this is certainly consistent in the toroidal case, so it might seem that in the $`T^4/𝐙_2`$ orbifold case this should also be consistent. However, there is a subtlety here. In particular, note that the orientifold projection is of the $`SO`$ type on D9-branes, and, therefore, it is of the $`Sp`$ type on D5-branes. The odd-half-integer twisted $`B`$-flux is present at all 16 fixed points of the $`T^4/𝐙_2`$ orbifold. Thus, repeating the above argument we would conclude that all 16 O5-planes must be of the O5<sup>-</sup> type. So there seems to be a puzzle here. Let us, therefore, try to understand this point better.
To begin with, let us note that there are two separate issues here. First, by studying the orientifold action in the closed string sector we unambiguously arrive at the conclusion that we have 12 O5<sup>-</sup>\- and 4 O5<sup>+</sup>-planes as in the toroidal case. However, this does not guarantee that, once we introduce both 32 D9-branes and 16 D5-branes, the orientifold action in the open string sector is indeed consistent. This statement might seem surprising at first as the orientifold action in the 99 and 55 sectors certainly seems to be consistent. This follows from our analyses of the corresponding toroidal orientifolds with $`B`$-flux (and we do not expect any obstruction in consistently orbifolding the 99 and 55 sectors). However, the subtlety here can (and, as we will point out in a moment, does) arise in the 59 sector. In particular, it is not always sufficient to consider, say, the Klein bottle amplitude (and the corresponding tadpoles) to conclude that we have some numbers of O$`p^{}`$\- and O$`p^+`$-planes. Rather, we must also make sure that in any given setup these objects indeed induce the $`SO`$ and $`Sp`$ type of orientifold projections on D$`p`$-branes (that is, we must make sure that the couplings between the O$`p`$-planes and D$`p`$-branes are consistent).
Since the issue we are discussing here appears to be subtle, we would like to proceed step-by-step. Thus, first let us make sure that the orientifold projection in the closed string sector is indeed such that we have $`n_f`$ twisted hypermultiplets and $`n_{f+}`$ twisted tensor multiplets. Thus, we must show that the orientifold action on $`n_{f+}`$ fixed points has an extra minus sign compared with that on the other $`n_f`$ fixed points. One way to see this explicitly was already discussed in . Thus, let us consider a special point in the moduli space of $`T^4`$’s corresponding to the $`SO(8)`$ symmetry. At this point the $`B`$-field has rank $`b=2`$ , and in the conformal field theory we have the $`[SO(8)]_L[SO(8)]_R`$ current algebra at level 1 on the closed string world-sheet. The $`𝐙_2`$ orbifold action reduces this current algebra to $`[SU(2)^4]_L[SU(2)^4]_R`$ at level 1. The vertex operators for the 16 fixed points are especially simple at this point. In particular, they carry the following quantum numbers under $`[SU(2)^4]_L[SU(2)^4]_R`$: $`(\mathrm{𝟐},\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏}||\mathrm{𝟐},\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})`$, plus states obtained by simultaneous permutations of the $`SU(2)`$ factors for both left- and right-movers. The orientifold action reduces the $`[SU(2)^4]_L[SU(2)^4]_R`$ current algebra to $`[SU(2)^4]_{\mathrm{diag}}`$, and the fixed points now carry the following quantum numbers: $`(\mathrm{𝟑}_s\mathrm{𝟏}_a,\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})`$, plus states obtained by permutations of the $`SU(2)`$ factors. Note that the states transforming in $`\mathrm{𝟑}_s`$’s are symmetric under $`\mathrm{\Omega }`$, while the states transforming in $`\mathrm{𝟏}_a`$’s are antisymmetric. Thus, we indeed have 12 twisted hypermultiplets and 4 twisted tensor multiplets in this case<sup>†‡‡</sup><sup>†‡‡</sup>†‡‡It is not difficult to see that in the $`b=4`$ case we then have 10 twisted hypermultiplets and 6 twisted tensor multiplets .. In particular, this is consistent with the fact that we have $`n_f`$ O5<sup>-</sup>-planes located at the fixed points giving rise to twisted hypermultiplets, and $`n_{f+}`$ O5<sup>+</sup>-planes located at the fixed points giving rise to twisted tensor multiplets (as can be explicitly seen by considering the Klein bottle amplitude in this model).
Next, let us discuss the open string sector. In the 99 sector we have Chan-Paton matrices $`\gamma _{S_a,9}`$ corresponding to non-commuting Wilson lines<sup>‡\*</sup><sup>‡\*</sup>‡\*As in the previous section, here we are going to view half-integer $`B`$-flux on the first $`T^2`$ with vielbeins $`e_i`$ in terms of the $`𝐙_2𝐙_2`$ freely acting orbifold (with discrete torsion) of the torus $`\stackrel{~}{T}^2`$ with vielbeins $`E_i=2e_i`$ and zero $`B`$-flux. The non-commuting Wilson lines then correspond to half-lattice shifts $`S_iX_i=X_i+E_i/2`$ on $`\stackrel{~}{T}^2`$ acting on the Chan-Paton degrees of freedom., and also the twisted Chan-Paton matrix $`\gamma _{R,9}`$. We can write the corresponding Chan-Paton matrices as follows:
$`\gamma _{S_a,9}=\gamma _aI_2I_8,`$ (97)
$`\gamma _{R,9}=I_2\nu \sigma _3I_8,`$ (98)
where $`\gamma _a`$ are the $`2\times 2`$ matrices corresponding to the $`D_4`$ or $`D_4^{}`$ type of Wilson lines given in (30) respectively (31). On the other hand, for the twisted Chan-Paton matrix $`\gamma _{R,9}`$ a priori we have two inequivalent choices: $`\nu ^2=+1`$ and $`\nu ^2=1`$. However, as we will show in a moment, these choices are correlated with the choices of the non-commuting Wilson lines as follows:
$`D_4:\nu ^2=+1,`$ (99)
$`D_4^{}:\nu ^2=1.`$ (100)
To show this, let us consider the Möbius strip amplitude:
$`=\left({\displaystyle \frac{1}{2}}\right)^2\left[𝒴(\mathrm{\Omega },9)+𝒴(\mathrm{\Omega }R,9)+𝒴(\mathrm{\Omega }R,5)+𝒴(\mathrm{\Omega },5)\right].`$ (101)
Here the characters $`𝒴`$, which contain the corresponding Chan-Paton factors, are defined as follows. The character $`𝒴(\mathrm{\Omega },9)`$ corresponds to the 99 sector contribution with the identity element of the $`𝐙_2`$ orbifold group inserted on the boundary of the Möbius strip; the character $`𝒴(\mathrm{\Omega }R,9)`$ corresponds to the 99 sector contribution with element $`R`$ of the $`𝐙_2`$ orbifold group inserted on the boundary. Similarly, the character $`𝒴(\mathrm{\Omega }R,5)`$ corresponds to the 55 sector contribution with the identity element inserted on the boundary; the character $`𝒴(\mathrm{\Omega },5)`$ corresponds to the 55 sector contribution with the element $`R`$ inserted on the boundary. Starting from (19), it is not difficult to see that the character $`𝒴(\mathrm{\Omega }R,9)`$ is given by:
$`𝒴(\mathrm{\Omega }R,9)`$ $`={\displaystyle \frac{1}{4}}𝒵(\mathrm{\Omega }R,9)\left[\mathrm{Tr}(\gamma _{\mathrm{\Omega }R,9}^1\gamma _{\mathrm{\Omega }R,9}^T)+{\displaystyle \underset{a=1}{\overset{3}{}}}\mathrm{Tr}(\gamma _{\mathrm{\Omega }RS_a,9}^1\gamma _{\mathrm{\Omega }RS_a,9}^T)\right]`$ (103)
$`={\displaystyle \frac{1}{4}}𝒵(\mathrm{\Omega }R,9)\mathrm{Tr}(\gamma _{R,9}^2)\left[1+{\displaystyle \underset{a=1}{\overset{3}{}}}\eta _{aa}\right],`$
where the character $`𝒵(\mathrm{\Omega }R,9)`$ here is the same as in (92). Next, let us consider the character $`𝒴(\mathrm{\Omega },5)`$. Since there are no Wilson lines in the 55 sector, we have
$$𝒴(\mathrm{\Omega },5)=\mathrm{Tr}(\gamma _{\mathrm{\Omega },5}^1\gamma _{\mathrm{\Omega },5}^T)𝒵(\mathrm{\Omega },5),$$
(104)
where the character $`𝒵(\mathrm{\Omega },5)`$ is the same as in (92). Now repeating the argument after (92) for the above model, we conclude that in the Möbius strip amplitude for the $`𝐙_2`$ orbifold projection in the 99 sector to be consistent with that in the 55 sector, the following constraint must be satisfied:
$$\frac{1}{4}\mathrm{Tr}(\gamma _{R,9}^2)[1+\underset{a=1}{\overset{3}{}}\eta _{aa}]=\mathrm{Tr}(\gamma _{\mathrm{\Omega },5}^1\gamma _{\mathrm{\Omega },5}^T).$$
(105)
We can simplify this constraint as follows. First, note that $`\mathrm{Tr}(\gamma _{\mathrm{\Omega },5}^1\gamma _{\mathrm{\Omega },5}^T)=16`$. Also, $`\mathrm{Tr}(\gamma _{R,9}^2)=32\nu ^2`$. Finally, $`_{a=1}^3\eta _{aa}=+1`$ for the $`D_4^{}`$ type of Wilson lines (31), and $`_{a=1}^3\eta _{aa}=3`$ for the $`D_4`$ type of Wilson lines (30). Putting all of this together, we conclude that for the $`D_4`$ type of Wilson lines we must have $`\nu ^2=+1`$, whereas for the $`D_4^{}`$ type of Wilson lines we must have $`\nu ^2=1`$.
The constraint we have just derived is the analogue of the corresponding constraint in the case without $`B`$-flux. In the latter case, as we have already discussed above, relaxing this constraint would result in a model where tadpoles/anomalies cannot be canceled completely. In the case with non-zero $`B`$-field relaxing the corresponding constraint would also result in inconsistent models. Thus, for instance, consider taking the $`D_4`$ type of Wilson lines with $`\nu ^2=1`$. The resulting model has the following massless spectrum. We have already discussed the closed string sector (which is independent of the choice of the Chan-Paton matrices) - it contains 16 hypermultiplets and 5 tensor multiplets (for $`b=2`$). In the open string sector the gauge group is $`[U(8)𝐙_2]_{99}U(8)_{55}`$, and the massless matter is given by the following hypermultiplets:
$`2\times `$ $`(\mathrm{𝟑𝟔};\mathrm{𝟏})_{99},`$ (106)
$`2\times `$ $`(\mathrm{𝟏};\mathrm{𝟑𝟔})_{55},`$ (109)
$`(\mathrm{𝟖}_+;\mathrm{𝟖})_{95},`$
$`(\mathrm{𝟖}_{};\mathrm{𝟖})_{59},`$
where the subscript $`\pm `$ in the 95 sector refers to the 99 $`𝐙_2`$ charge. Note that this spectrum is anomalous. In particular, the irreducible $`R^4`$ gravitational anomaly does not cancel. A similar conclusion applies to the case where the Wilson lines are of the $`D_4^{}`$ type, while the $`𝐙_2`$ twisted Chan-Paton matrix is chosen such that $`\nu ^2=+1`$.
Next, we would like to study the models arising for the aforementioned two inequivalent choices satisfying the above consistency condition. To begin with, let us consider the following choice:
$`\gamma _{S_a,9}=\gamma _aI_2I_8,`$ (110)
$`\gamma _{R,9}=I_2i\sigma _3I_8,`$ (111)
where the $`\gamma _a`$ matrices correspond to the $`D_4^{}`$ type of Wilson lines given in (31). In the 55 sector we must also specify the choice of $`\gamma _{R,5}`$, which is a $`16\times 16`$ matrix (and not a $`32\times 32`$ matrix as we have only 16 D5-branes). The choice of this matrix must be consistent with the orientifold action. In particular, let us consider an O5<sup>-</sup>-plane, say, that located at the fixed point at the origin $`(0,0)`$. Then, if we place 16 D5-branes on top of this orientifold plane, according to the above arguments we must choose $`\gamma _{R,5}`$ such that its eigenvalues are $`\pm i`$. On the other hand, suppose we consider an O5<sup>+</sup>-plane, say, that located at the fixed point $`(e_3/2,0)`$. Then, if we placed D5-branes on top of this O5<sup>+</sup>-plane, we would encounter an inconsistency. To see this let us repeat the argument given in the beginning of this subsection for the case where the O5-plane is of the O5<sup>+</sup> type. Thus, in this case we have
$$\gamma _{\mathrm{\Omega }R,5}^T=\gamma _{\mathrm{\Omega }R,5}.$$
(112)
Together with (83) and (85) this implies that
$$\gamma _{R,5}^T=+\gamma _{\mathrm{\Omega },5}\gamma _{R,5}\gamma _{\mathrm{\Omega },5}.$$
(113)
Now consider the basis where $`\gamma _{R,5}`$ is diagonal. Then $`\gamma _{R,5}^T=+\gamma _{R,5}`$, which together with (113) and (87) implies that
$$\gamma _{R,5}^2=+1.$$
(114)
This implies that the eigenvalues of the matrix $`\gamma _{R,5}`$ are $`\pm 1`$. Note, however, that the eigenvalues of the matrix $`\gamma _{R,9}`$ are $`\pm i`$. That is, in the 59 sector the $`𝐙_2`$ orbifold action is inconsistent as the 59 states would have phases $`\pm i`$ (instead of $`\pm 1`$). At first it might seem that this difficulty can be simply avoided by concluding that in this background it is inconsistent to place D5-branes at the O5<sup>+</sup>-planes, albeit it is consistent to place them at the O5<sup>-</sup>-planes. However, this naive resolution fails due to the following simple argument<sup>‡†</sup><sup>‡†</sup>‡†This effective field theory argument was pointed out to me by Alex Buchel. \- there is no obstruction to moving D5-branes from an O5<sup>-</sup>-plane to an O5<sup>+</sup>-plane in this model.
Thus, let us consider the $`T^4/𝐙_2`$ model with $`B`$-field of rank $`b=2`$. Let us place all 16 D5-branes at the O5<sup>-</sup>-plane located at the origin $`(0,0)`$. Then the consistent choice of $`\gamma _{R,5}`$ is (up to equivalent representations) given by
$$\gamma _{R,5}=i\sigma _3I_8.$$
(115)
The gauge group of this model is $`[U(8)𝐙_2]_{99}U(8)_{55}`$. The open string massless matter consists of the following hypermultiplets
$`2\times `$ $`(\mathrm{𝟐𝟖};\mathrm{𝟏})_{99},`$ (116)
$`2\times `$ $`(\mathrm{𝟏};\mathrm{𝟐𝟖})_{55},`$ (119)
$`(\mathrm{𝟖}_+;\mathrm{𝟖})_{95},`$
$`(\mathrm{𝟖}_{};\mathrm{𝟖})_{95}.`$
Here the subscript $`\pm `$ in the 95 sector refers to the 99 $`𝐙_2`$ discrete gauge charge of the corresponding state. Note that $`U(1)`$’s are actually anomalous, and are broken by the generalized Green-Schwarz mechanism . In particular, the states that participate in the generalized Green-Schwarz mechanism are (certain linear combinations of) the R-R scalars in the twisted hypermultiplets. Thus, the gauge group is actually $`[SU(8)𝐙_2]_{99}SU(8)_{55}`$. Moreover, the corresponding two twisted hypermultiplets are eaten in the process of Higgsing $`U(1)`$’s. Note that the above spectrum is free of the irreducible $`R^4`$ and $`F^4`$ anomalies.
Note that we can Higgs the 55 gauge group by giving VEVs to the hypermultiplets $`2\times (\mathrm{𝟏};\mathrm{𝟐𝟖})_{55}`$. Thus, for instance, we can break the 55 gauge group $`SU(8)_{55}`$ down to $`Sp(8)_{55}`$ by giving appropriate VEVs to the aforementioned hypermultiplets - under the breaking $`SU(8)Sp(8)`$, the antisymmetric representation $`\mathrm{𝟐𝟖}`$ of $`SU(8)`$ decomposes as $`\mathrm{𝟐𝟖}=\mathrm{𝟏}+\mathrm{𝟐𝟕}`$ in terms of the $`Sp(8)`$ representations. Actually, we can start from the $`U(8)_{55}`$ gauge group with the anomalous $`U(1)_{55}`$ factor, and break $`U(8)_{55}`$ down to $`Sp(8)_{55}`$. In this process 28 out of the 56 hypermultiplets $`2\times (\mathrm{𝟏};\mathrm{𝟐𝟖})_{55}`$ are eaten in the Higgs mechanism<sup>‡‡</sup><sup>‡‡</sup>‡‡Note that a singlet of $`Sp(8)_{55}`$ is eaten in Higgsing the $`U(1)_{55}`$ factor. This implies that only one (instead of two) of the closed twisted hypermultiplets is eaten, namely, in the process of Higgsing the anomalous $`U(1)_{99}`$ factor.. The leftover 28 hypermultiplets transform in $`\mathrm{𝟏}\mathrm{𝟐𝟕}`$ of $`Sp(8)_{55}`$. Note that this Higgsing corresponds to nothing but moving together all 16 D5-branes off the O5<sup>-</sup>-plane into the bulk. The VEV of the leftover singlet hypermultiplet then corresponds to the location of D5-branes in the bulk (note that there are 4 real scalars in a hypermultiplet, and those in the singlet hypermultiplet parametrize the location of D5-branes in four real dimensions of K3). On the other hand, we can further break the $`Sp(8)`$ gauge group by giving a VEV to the leftover hypermultiplet which is in $`\mathrm{𝟐𝟕}`$ of $`Sp(8)`$. This Higgsing corresponds to pulling D5-branes apart from each other in groups of 4 (or multiples thereof) - each dynamical 5-brane, which consists of 4 D5-branes<sup>‡§</sup><sup>‡§</sup>‡§Here two pairings take place - one due to the orientifold projection, and the other one due to the orbifold projection., separately gives rise to an $`Sp(2)`$ gauge group. On the other hand, $`k`$ coincident dynamical 5-branes (corresponding to $`4k`$ coincident D5-branes) give rise to $`Sp(2k)`$ gauge group, with all four dynamical 5-branes coincident giving rise to the $`Sp(8)`$ gauge symmetry.
Even though the general case is straightforward to discuss, from now on it will suffice for our purposes here to consider moving together all 16 D5-branes off the O5<sup>-</sup>-plane, so that in the bulk they give rise to the $`Sp(8)`$ gauge symmetry. In this subspace of the moduli space the gauge group is $`[SU(8)𝐙_2]_{99}Sp(8)_{55}`$, and the open string massless matter consists of the following hypermultiplets:
$`2\times `$ $`(\mathrm{𝟐𝟖};\mathrm{𝟏})_{99},`$ (123)
$`(\mathrm{𝟏};\mathrm{𝟏}\mathrm{𝟐𝟕})_{55},`$
$`(\mathrm{𝟖}_+;\mathrm{𝟖})_{95},`$
$`(\mathrm{𝟖}_{};\mathrm{𝟖})_{95}.`$
In fact, this spectrum can be derived directly in the orientifold language if we consider placing 16 D5-branes in the bulk (that is, away from the $`𝐙_2`$ orbifold fixed points). In particular, let us place 8 D5-branes at a generic point $`(z_1,z_2)`$ in the bulk. Then we must place the other 8 D5-branes at the point $`(z_1,z_2)`$, so that the entire background is invariant under the action of the $`\mathrm{\Omega }R`$ element of the orientifold group. If we did not have to perform the further $`𝐙_2`$ orbifold projection with respect to the element $`R`$, the 55 gauge group at such a generic point would be $`U(8)_{55}`$. However, the $`𝐙_2`$ orbifold projection reduces $`U(8)_{55}`$ to its subgroup $`Sp(8)_{55}`$. Note that the reason why the rank of the unbroken gauge group $`Sp(8)_{55}`$ is halved compared with that of $`U(8)_{55}`$ is that $`\gamma _{\mathrm{\Omega }R,5}`$ and $`\gamma _{R,5}`$ do not commute, in fact, they anticommute as can be seen from the relation $`\gamma _{R,5}=\gamma _{\mathrm{\Omega },5}\gamma _{R,5}\gamma _{\mathrm{\Omega },5}`$, which holds in the basis where $`\gamma _{R,5}`$ is diagonal. Furthermore, the 55 gauge bosons are in the symmetric representation (so that the gauge group is symplectic), while the 55 hypermultiplets are in the antisymmetric representation (which is reducible for symplectic gauge groups). The latter fact is due to the extra minus sign that the $`𝐙_2`$ twist $`R`$ has when acting on the hypermultiplets compared with when it acts on the gauge bosons.
Next, we would like to ask what would happen if we bring D5-branes to one of the O5<sup>+</sup>-planes. Here we expect that the 55 gauge symmetry should be enhanced, and, moreover, the 55 gauge group must be a subgroup of $`Sp(16)`$. The latter is the gauge group we would obtain in the toroidal case, and in the orbifold case the 55 gauge group would have to be determined by the choice of $`\gamma _{R,5}`$. Thus, if we choose $`\gamma _{R,5}`$ to be
$$\gamma _{R,5}=\sigma _3I_8,$$
(124)
then the gauge group coming from 16 D5-branes on top of an O5<sup>+</sup>-plane is $`Sp(8)Sp(8)`$. Thus, the gauge group of the model at this point in the moduli space is $`[SU(8)𝐙_2]_{99}[Sp(8)Sp(8)]_{55}`$, and the open string massless matter is given by:
$`2\times `$ $`(\mathrm{𝟐𝟖};\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (126)
$`(\mathrm{𝟏};\mathrm{𝟖},\mathrm{𝟖})_{55},`$
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_+;\mathrm{𝟖},\mathrm{𝟏})_{95},`$ (127)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_+;\mathrm{𝟏},\mathrm{𝟖})_{95},`$ (128)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_{};\mathrm{𝟖},\mathrm{𝟏})_{95},`$ (129)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_{};\mathrm{𝟏},\mathrm{𝟖})_{95}.`$ (130)
Note that the 99 spectrum is the same as before, and the 55 spectrum can be deduced by keeping the states in the toroidal compactification invariant under the $`𝐙_2`$ orbifold action. The states in the 95 sector, however, cannot be deduced in this way as the $`𝐙_2`$ action in this sector is inconsistent - recall that the eigenvalues of $`\gamma _{R,9}`$ are $`\pm i`$, while the eigenvalues of $`\gamma _{R,5}`$ are $`\pm 1`$. The above 95 spectrum has been written down as follows. First, the number of 95 hypermultiplets must be the same whether D5-branes are on top of the O5<sup>+</sup>-plane or in the bulk. Second, the two $`Sp(8)`$ subgroups in the 55 sector are on the equal footing, so that the spectrum should possess a symmetry under the permutation of these two subgroups. This then fixes the spectrum as above. This spectrum, however, is anomalous. Indeed, it contains half-hypermultiplets in complex representations. Note that had we not distinguished the 99 $`𝐙_2`$ discrete gauge quantum numbers, naively we might have thought that we have one hypermultiplet in $`(\mathrm{𝟖};\mathrm{𝟖},\mathrm{𝟏})`$, and one hypermultiplet in $`(\mathrm{𝟖};\mathrm{𝟏},\mathrm{𝟖})`$ of $`SU(8)_{99}[Sp(8)Sp(8)]_{55}`$. This is, however, not the case, and, as we see, the 99 $`𝐙_2`$ discrete gauge symmetry does indeed play an important role.
Thus, we have arrived at an inconsistency - we started from a seemingly consistent setup where we had all 16 D5-branes on top of an O5<sup>-</sup>-plane, and then moved them to one of the O5<sup>+</sup>-planes. Even though in the bulk the 55 gauge theory appears to be consistent, at the O5<sup>+</sup>-plane it is anomalous. We have seen this in the effective field theory language, and in the orientifold language this inconsistency is translated to that in the $`𝐙_2`$ orbifold action in the 59 sector due to the fact that $`\gamma _{R,5}`$ is given by (124), while $`\gamma _{R,9}`$ is given by (111). Here we can ask if we could possibly have chosen $`\gamma _{R,5}`$ as in (115) even if D5-branes are on top of an O5<sup>+</sup>-plane. It is, however, not difficult to see that this choice is inconsistent. Thus, we know from our discussion which led to (114) that the choice consistent with having D5-branes on top of an O5<sup>+</sup>-plane is that given in (124), and not in (115). However, we can readily see what goes wrong if we make this inconsistent choice in the language of the effective field theory as well. Thus, let us assume for a moment that $`\gamma _{R,5}`$ is given by (115) even though D5-branes are placed on top of an O5<sup>+</sup>-plane. Then the gauge group of the model would be $`[U(8)𝐙_2]_{99}U(8)_{55}`$, and the massless open string spectrum would read (here we have kept anomalous $`U(1)`$’s for the convenience reasons that will become clear in a moment):
$`2\times `$ $`(\mathrm{𝟐𝟖};\mathrm{𝟏})_{99},`$ (131)
$`2\times `$ $`(\mathrm{𝟏};\mathrm{𝟑𝟔})_{55},`$ (134)
$`(\mathrm{𝟖}_+;\mathrm{𝟖})_{95},`$
$`(\mathrm{𝟖}_{};\mathrm{𝟖})_{95}.`$
Note that the 95 part of this spectrum now looks consistent. However, this spectrum is actually anomalous once we take into account the massless content of the closed string sector, which consists of the six dimensional $`𝒩=1`$ supergravity multiplet, 5 tensor supermultiplets, and 16 hypermultiplets. Thus, for instance, the $`R^4`$ gravitational anomaly does not cancel in this model. This is due to the fact that we have 55 hypermultiplets in $`\mathrm{𝟑𝟔}`$ (symmetric) representation of $`SU(8)`$ instead of $`\mathrm{𝟐𝟖}`$ (antisymmetric), the reason being that D5-branes now are at an O5<sup>+</sup>-plane instead of an O5<sup>-</sup>-plane. Let us mention that we have kept the anomalous $`U(1)`$ factors merely for the counting convenience - eventually they are broken, so we can drop them, but then instead of 16 closed string hypermultiplets we would have only 14 as two twisted hypermultiplets (or, more precisely, certain linear combinations thereof) are eaten in the corresponding Higgs mechanism.
Thus, the assumption that we can have both O5<sup>-</sup>\- and O5<sup>+</sup>-planes at various fixed points of the conformal field theory orbifold $`T^4/𝐙_2`$ does not seem to be self-consistent. This in accord with our discussion of the role of the twisted $`B`$-flux in the collapsed $`𝐏^1`$’s at the orbifold fixed points. In particular, according to this discussion in the case of the conformal field theory orbifold $`T^4/𝐙_2`$ we expect that only O5<sup>-</sup>-planes can be placed at the orbifold fixed points. The obstruction for placing O5<sup>+</sup>-planes at the fixed points comes precisely from the fact that we have twisted $`B`$-flux. This, in turn, gives us a hint of how to possibly remedy the situation in the case of the orientifold of Type IIB on $`T^2/𝐙_2`$ with $`B`$-flux. More precisely, we can attempt to guess what the correct background for such an orientifold might be (the orientifold of the conformal field theory orbifold $`T^4/𝐙_2`$ does not appear to be the correct background).
The key observation here is that to have an O5<sup>+</sup>-plane at the orbifold fixed point we must have trivial twisted $`B`$-flux in the corresponding collapsed $`𝐏^1`$. However, if we simply turn off the twisted $`B`$-flux, the corresponding background could no longer be described within the world-sheet (that is, the conformal field theory) approach. Indeed, in this case we have a true geometric $`𝐀_1`$ singularity at each of the fixed points with the twisted $`B`$-field turned off. To avoid this, we would have to blow up the orbifold singularity by giving a VEV to the corresponding twisted hypermultiplet. Note, however, that if we first consider the $`T^4/𝐙_2`$ orbifold with collapsed $`𝐏^1`$’s and then orientifold, the fixed points where we have O5<sup>+</sup>-planes would give rise to twisted tensor multiplets only, so such a blow-up would not be possible. However, we can proceed as follows. Consider Type IIB on K3, where K3 is defined as follows. Consider the $`T^4/𝐙_2`$ orbifold with $`B`$-flux of rank $`b=2`$. At the 12 fixed points where we expect O5<sup>-</sup>-planes we have collapsed $`𝐏^1`$’s with half-integer twisted $`B`$-flux, and these points locally can be described in the conformal field theory. At the other 4 fixed points where we expect O5<sup>+</sup>-planes we have $`𝐏^1`$’s of non-zero size but with trivial twisted $`B`$-flux. Moreover, let us assume that the size of K3 is large compared with the size of these blow-ups. Then we can indeed locally describe the other 12 fixed points in the exactly solvable conformal field theory language of the $`𝐂^2/𝐙_2`$ orbifold. On the other hand, the 4 blown-up fixed points without the twisted $`B`$-flux can no longer be described in terms of an exactly solvable orbifold conformal field theory (albeit there should exist some conformal field theory description of such a K3 which is not exactly solvable but corresponds to some non-trivial sigma-model). Next, start from Type IIB on the K3 surface we have just described, and consider its $`\mathrm{\Omega }`$ orientifold. At the 12 fixed points with the twisted $`B`$-flux we have O5<sup>-</sup>-planes, and these fixed points give rise to twisted hypermultiplets, while at the 4 fixed points without the twisted $`B`$-flux we have O5<sup>+</sup>-planes, and these fixed points give rise to twisted tensor multiplets. Since there are no twisted hypermultiplets at these 4 fixed points, we cannot blow down the corresponding $`𝐏^1`$’s. In this sense, the orientifolding procedure does not commute with the blowing-up procedure. This, in turn, might signal that there could be a caveat in the above discussion. In particular, it is not completely evident that the boundary states at the 4 blown-up fixed points, if such are at all present, carry the correct R-R charges to be interpreted as O5<sup>+</sup>-planes. In fact, technical issues in conformal field theories with boundaries might, at least partially, be responsible for difficulties in making this point quantitatively more precise. Nonetheless, we can attempt to proceed further in understanding the underlying qualitative picture assuming that the boundary states at the blown-up fixed points indeed correspond to the O5<sup>+</sup>-planes.
What can we say about this orientifold? With the aforementioned assumption, it is reasonable to assume that if we place 16 D5-branes at one of the O5<sup>-</sup>-planes, then we get the spectrum described above, which (modulo the missing 99 $`𝐙_2`$ discrete symmetry) is the same as that given in . This spectrum is consistent, and with the aforementioned interpretation of the K3 background might adequately describe physics at the corresponding point in the moduli space. Once we move D5-branes off the O5<sup>-</sup>-plane, the 55 gauge symmetry in the bulk is $`Sp(8)`$ (or an appropriate subgroup thereof). The key question, however, is what happens when we approach an O5<sup>+</sup>-plane - after all this was where we have encountered trouble in the above discussion in the context of the conformal field theory orbifold. Note, however, that in the case of the K3 surface under consideration the moduli space corresponding to the motion of D5-branes is no longer flat as it was in the case of the conformal field theory orbifold (more precisely, in the latter case it is flat everywhere except for the fixed points). Because of the non-zero size of the corresponding $`𝐏^1`$’s, D5-branes actually might not be able to come on top of the O5<sup>+</sup>-planes (whose locations in K3 are given by points inside of the blown-up $`𝐏^1`$’s). If so, this would avoid the contradiction we have encountered in the conformal field theory orbifold case. That is, the 55 gauge symmetry would never be enhanced to $`Sp(8)Sp(8)`$ \- in the bulk it is at most $`Sp(8)`$, at an O5<sup>-</sup>-plane it is $`SU(8)`$, while the points corresponding to D5-branes being on top of O5<sup>+</sup>-planes might possibly be thought of as being at infinite distance in the moduli space. If so, the aforementioned K3 surface might indeed be the correct consistent background for the corresponding orientifold with $`B`$-flux. Once again, however, it is not completely clear how to make the above discussion quantitatively more precise due to the fact that the conformal field theory corresponding to such a K3 would not be exactly solvable<sup>‡¶</sup><sup>‡¶</sup>‡¶In the next section we will discuss four dimensional orientifolds with $`𝒩=2`$ and $`𝒩=1`$ supersymmetry where the aforementioned difficulties are avoided within exactly solvable conformal field theory compactifications..
Before we summarize the findings of this subsection, we would like to give two further pieces of evidence that in considering the $`\mathrm{\Omega }`$ orientifold of Type IIB on the conformal field theory $`T^4/𝐙_2`$ orbifold with $`B`$-flux one indeed runs into various subtle inconsistencies. The first piece of such evidence comes from studying the 99 sector moduli space in the above model. The second piece of evidence comes from considering the second seemingly consistent setup, namely, that with the $`D_4`$ type of Wilson lines accompanied with the twisted Chan-Paton matrix $`\gamma _{R,9}`$ with $`\nu ^2=+1`$.
Thus, so far we have focused on the 55 sector moduli space. Similar considerations apply to the 99 sector. Thus, suppose we start from the point in the moduli space where all 16 D5-branes are placed at an O5<sup>-</sup>-plane. At this point we have $`[SU(8)𝐙_2]_{99}SU(8)_{55}`$ gauge symmetry. We can Higgs the $`SU(8)_{99}`$ gauge group down to $`Sp(8)_{99}`$ or its appropriate subgroups, and this Higgsing now corresponds to turning on Wilson lines. We can describe these Wilson lines as in (32), except that now we also have the twisted Chan-Paton matrix $`\gamma _{R,9}`$. Thus, consider the following choice of $`\gamma _{S_a,9}`$ and $`\gamma _{R,9}`$:
$`\gamma _{S_1,9}=\sigma _3\rho (a)I_8,`$ (135)
$`\gamma _{S_2,9}=\sigma _1\rho (a^1)I_8,`$ (136)
$`\gamma _{S_3,9}=i\sigma _2I_2I_8,`$ (137)
$`\gamma _{R,9}=I_2i\sigma _1I_8.`$ (138)
Here $`\rho (a)\mathrm{diag}(a,a^1)`$, where $`a`$ is a complex phase. For definiteness we have chosen $`\gamma _{S_3,9}=+\gamma _{S_1,9}\gamma _{S_2,9}`$. Note that $`\gamma _{R,9}`$ commutes with $`\gamma _{S_3,9}`$. However, for $`\gamma _{S_i,9}`$, $`i=1,2`$, we have:
$$\gamma _{R,9}\gamma _{S_i,9}=\gamma _{S_i,9}^1\gamma _{R,9}.$$
(139)
Note that for $`a^2=1`$ $`\gamma _{R,9}`$ actually commutes with $`\gamma _{S_i,9}`$.
Before we proceed further, the following remarks are in order. First, for $`a^2=1`$ (as well as $`a^2=1`$) the aforementioned Wilson lines can be thought of in terms of the freely acting $`𝐙_2𝐙_2`$ orbifold (with discrete torsion). However, at generic points they correspond to more general Wilson lines that do not possess a simple (freely acting) orbifold description. However, at rational values of the phase, say, $`a=\mathrm{exp}(2\pi i/k)`$, $`k𝐙`$, such an orbifold description does exist. Thus, for instance, let $`k2𝐍+1`$. Then $`\gamma _{S_1,9}^2`$ generates a $`𝐙_k`$ discrete group. In the language of a freely acting orbifold this can be understood as having $`𝐙_k`$ valued shifts of the torus lattice. Note that such a shift is mapped to its inverse by the reflection of the torus coordinates described by the twist $`R`$. This is precisely the reason why we have chosen the Chan-Paton matrices $`\gamma _{S_a,9}`$ and $`\gamma _{R,9}`$ so that they satisfy the conjugation relation (139). In fact, this conjugation relation should also hold for generic (that is, irrational) phases as well. To see this, consider a non-trivial Wilson line (which we schematically write as a diagonal $`N\times N`$ matrix)
$$W=\mathrm{exp}\left(i_CA\right),$$
(140)
where $`A=\mathrm{diag}(\theta _1,\mathrm{},\theta _N)`$ is a constant gauge field, and $`C`$ is a non-trivial 1-cycle. Now consider a $`𝐙_2`$ orbifold action such that it reverses the coordinate parametrizing $`C`$. Then the action of the $`𝐙_2`$ orbifold generator $`R`$ on the Wilson line is given by (note that $`R^2=1`$)
$$RWR=W^1,$$
(141)
which follows from the fact that $`A`$ reverses its sign under the action of $`R`$. This then implies that the conjugation relation (139) also holds for generic values of the phase $`a`$.
Next, note that at, say, $`a=1`$ the above choice of $`\gamma _{S_a,9}`$ and $`\gamma _{R,9}`$ is equivalent to that in (110) and (111). In particular, the 99 gauge group is $`[SU(8)𝐙_2]_{99}`$, while the 55 gauge group is $`SU(8)_{55}`$, and all the D5-branes are placed at an O5<sup>-</sup>-plane. Now let us continuously deform the above Wilson lines from the point $`a=1`$ to, say, the point $`a=i`$. At this point the Wilson lines $`\gamma _{S_a,9}`$ are of the $`D_4`$ type (while the Wilson lines are of the $`D_4^{}`$ type at $`a=1`$) - the eigenvalues of all three matrices $`\gamma _{S_a,9}`$ are $`\pm i`$. So at this point the 99 gauge group would be $`Sp(16)𝐙_2`$ if we did not have a further $`𝐙_2`$ orbifold action. Let us see what the gauge group is once we perform the $`𝐙_2`$ orbifold projection with respect to $`\gamma _{R,9}`$. Actually, $`\gamma _{R,9}`$ does not commute with $`\gamma _{S_i,9}`$, $`i=1,2`$, at $`a=i`$, in fact, they anticommute. Thus, we should find a basis where the twisted Chan-Paton matrix commutes with the Wilson lines. Such a basis is given by $`\gamma _{S_a,9}`$ together with $`\gamma _{\stackrel{~}{R},9}`$, where $`\stackrel{~}{R}RS_3`$, and we can choose
$$\gamma _{\stackrel{~}{R},9}=\gamma _{R,9}\gamma _{S_3,9}=i\sigma _2i\sigma _1I_8.$$
(142)
Note that the eigenvalues of $`\gamma _{\stackrel{~}{R},9}`$ are no longer $`\pm i`$ but $`\pm 1`$, which is consistent with the fact that for the $`D_4`$ type of Wilson lines we must have $`\nu ^2=+1`$. This then implies that the 99 gauge group is actually $`[Sp(8)Sp(8)𝐙_2]_{99}`$. On the other hand, the 55 gauge group is unchanged - it is still $`SU(8)_{55}`$. It is then not difficult to see that here we are running into a problem similar to that in the case where we attempted to place D5-branes on top of an O5<sup>+</sup>-plane - the resulting spectrum in the 59 sector is anomalous. The reason for this is that the $`𝐙_2`$ projection in the 99 sector corresponds to the twisted Chan-Paton matrix $`\gamma _{\stackrel{~}{R},9}`$ with eigenvalues $`\pm 1`$, while that in the 55 sector corresponds to $`\gamma _{R,5}`$, whose eigenvalues are $`\pm i`$. That is, here we have the same type of inconsistency in the 59 sector as that encountered in the case of D5-branes sitting on top of an O5<sup>+</sup>-plane. Note that if we attempt to interpret $`\gamma _{R,9}`$ (instead of $`\gamma _{\stackrel{~}{R},9}`$) as the twisted Chan-Paton matrix with respect to which we must perform the $`𝐙_2`$ orbifold projection, then the latter might naively appear to be consistent in the 59 sector. However, it is not difficult to see that having performed the $`𝐙_2`$ orbifold projection in the 59 sector with respect to $`\gamma _{R,9}`$ (together with $`\gamma _{R,5}`$), the resulting 59 sector states would not transform in representations of the $`[Sp(8)Sp(8)𝐙_2]_{55}`$ gauge group, which is due to the fact that $`\gamma _{R,9}`$ and $`\gamma _{S_i,9}`$ do not commute at this point in the moduli space<sup>‡∥</sup><sup>‡∥</sup>‡∥We will encounter similar situations in other models in the following subsections, where we will discuss this point in more detail..
Note that the aforementioned inconsistency arises if we choose the background to be the $`\mathrm{\Omega }`$ orientifold of Type IIB on the conformal field theory orbifold $`T^4/𝐙_2`$ as in this case there is no obstruction to continuously deforming the above Wilson lines from the point $`a=1`$ to the point $`a=i`$. However, if we consider the partially blown-up K3 surface described above as the consistent orientifold background, such a continuous deformation of Wilson lines might no longer be possible as the corresponding moduli space is no longer flat. As in the discussion of the 55 moduli space, however, it is not clear how to make this point quantitatively precise. At any rate, if the aforementioned K3 surface is indeed a consistent background for the above orientifold, there would have to exist an obstruction to continuously deforming the Wilson lines from the points with $`a^2=+1`$ to those with $`a^2=1`$, as well as an obstruction to placing D5-branes on top of an O5<sup>+</sup>-plane. If this is indeed the case, then the above $`b=2`$ model (where the Wilson lines are of the $`D_4^{}`$ type, $`\nu ^2=1`$, and all 16 D5-branes are placed at an O5<sup>-</sup>-plane) would be consistent. In fact, then we could also consider its $`b=4`$ counterpart as follows. Consider the $`\mathrm{\Omega }`$ orientifold of Type IIB on K3, where K3 is a $`T^4/𝐙_2`$ orbifold with $`b=4`$ $`B`$-flux, and 10 of the $`𝐙_2`$ orbifold fixed points locally can be described in the language of the conformal field theory orbifold $`𝐂^2/𝐙_2`$, while the other 6 fixed points are blown-up, and the corresponding twisted $`B`$-flux is trivial. The former 10 fixed points are those at which we have O5<sup>-</sup>-planes, while the latter 6 fixed points are those at which we have O5<sup>+</sup>-planes. The $`b=4`$ $`B`$-flux can be described as follows. For the sake of simplicity consider $`T^4=T^2T^2`$. On the first $`T^2`$ we have non-commuting Wilson lines $`\gamma _{S_i,9}`$, while on the second $`T^2`$ we have non-commuting Wilson lines $`\gamma _{T_i,9}`$ (these two sets of Wilson lines, however, commute). Now consider the case where both sets of Wilson lines are of the $`D_4^{}`$ or $`D_4`$ type. Then the consistent choice for the twisted Chan-Paton matrix $`\gamma _{R,9}`$ is that with $`\nu ^2=1`$. The twisted Chan-Paton matrix $`\gamma _{R,5}`$ (which is an $`8\times 8`$ matrix as we have 8 D5-branes in this case) is then also fixed. If we place all D5-branes at an O5<sup>-</sup>-plane, then the gauge group is $`[U(4)𝐙_2𝐙_2]_{99}U(4)_{55}`$, and the open string sector massless hypermultiplets are given by:
$`2\times `$ $`(\mathrm{𝟔};\mathrm{𝟏})_{99},`$ (143)
$`2\times `$ $`(\mathrm{𝟏};\mathrm{𝟔})_{55},`$ (148)
$`(\mathrm{𝟒}_{++};\mathrm{𝟒})_{95},`$
$`(\mathrm{𝟒}_+;\mathrm{𝟒})_{95},`$
$`(\mathrm{𝟒}_+;\mathrm{𝟒})_{95},`$
$`(\mathrm{𝟒}_{};\mathrm{𝟒})_{95},`$
where the subscript $`\pm \pm `$ in the 95 sector refers to the 99 $`𝐙_2𝐙_2`$ discrete gauge charges. This spectrum (together with that from the closed string sector) is free of the irreducible $`R^4`$ and $`F^4`$ anomalies. The $`U(1)`$ factors are anomalous as usual, and are broken via the generalized Green-Schwarz mechanism. Note that (modulo the missing 99 $`𝐙_2𝐙_2`$ discrete gauge symmetry) this is the spectrum of the $`b=4`$ $`𝐙_2`$ model in .
Next, in the $`b=2`$ case, we would like to discuss the second a priori consistent choice for the Wilson lines and the twisted Chan-Paton matrix, namely, where the Wilson lines are of the $`D_4`$ type, and the twisted Chan-Paton matrix is chosen such that $`\nu ^2=+1`$. Note that in this case we must have $`\gamma _{R,5}^2=+1`$, from which it follows that D5-branes must be placed at an O5<sup>+</sup>-plane. Note that just as in the previous case the conformal field theory $`T^4/𝐙_2`$ orbifold does not appear to be the correct background for such an orientifold. In particular, in the conformal field theory $`T^4/𝐙_2`$ orbifold case we would have all the troubles we have encountered for the previous choice of the Wilson lines and the twisted Chan-Paton matrix. Thus, for instance, there would be no obstruction to moving D5-branes from an O5<sup>+</sup>-plane to an O5<sup>-</sup>-plane, and at the latter point in the moduli space the spectrum of the model would be anomalous. However, in the present case (that is, where the Wilson lines are of the $`D_4`$ type, and the twisted Chan-Paton matrix is chosen such that $`\nu ^2=+1`$) there is an additional puzzling issue. In particular, from our discussion of the role of the twisted $`B`$-flux it actually follows that we could not have an O5<sup>+</sup>-plane at a fixed point (locally) corresponding to the conformal field theory orbifold $`𝐂^2/𝐙_2`$. In fact, if we can at all have an O5<sup>+</sup>-plane at a $`𝐙_2`$ fixed point, we expect that the latter would have to be blown-up, and the twisted $`B`$-field would have to be trivial. Moreover, in this case it would be impossible to place D5-branes on top of such an O5<sup>+</sup>-plane.
In the light of the above discussion, we would like to see whether we can find any inconsistency in the model where the Wilson lines are of the $`D_4`$ type, the twisted Chan-Paton matrix is chosen such that $`\nu ^2=+1`$, and all D5-branes are placed at an O5<sup>+</sup>-plane. It is not difficult to see that the gauge group of this model is $`[Sp(8)Sp(8)𝐙_2]_{99}[Sp(8)Sp(8)]_{55}`$, and the massless open string hypermultiplets are given by:
$`(\mathrm{𝟖},\mathrm{𝟖};\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (150)
$`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟖},\mathrm{𝟖})_{55},`$
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_+,\mathrm{𝟏};\mathrm{𝟖},\mathrm{𝟏})_{95},`$ (151)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟖}_{},\mathrm{𝟏};\mathrm{𝟖},\mathrm{𝟏})_{95},`$ (152)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟏},\mathrm{𝟖}_+;\mathrm{𝟏},\mathrm{𝟖})_{95},`$ (153)
$`{\displaystyle \frac{1}{2}}`$ $`(\mathrm{𝟏},\mathrm{𝟖}_{};\mathrm{𝟏},\mathrm{𝟖})_{95},`$ (154)
where the subscript $`\pm `$ in the 95 sector refers to the 99 $`𝐙_2`$ discrete gauge charge of the corresponding state. Note that the $`R^4`$ gravitational anomaly cancels in this model. However, the above spectrum is not completely anomaly free - it contains half-hypermultiplets in real representations<sup>‡\**</sup><sup>‡\**</sup>‡\**Note that the fundamental representation of a symplectic gauge group is pseudoreal. However, here we have bifundamental representations of a product symplectic group, and the former are real.. Note that had we not distinguished the 99 $`𝐙_2`$ discrete gauge quantum numbers, naively we might have thought that in the 95 sector we have one hypermultiplet in $`(\mathrm{𝟖},\mathrm{𝟏};\mathrm{𝟖},\mathrm{𝟏})`$, and one hypermultiplet in $`(\mathrm{𝟏},\mathrm{𝟖};\mathrm{𝟏},\mathrm{𝟖})`$, which would give a consistent spectrum. This is, however, not the case here, and, once again, the 99 $`𝐙_2`$ discrete gauge symmetry indeed plays an important role. Thus, the above spectrum with the $`[Sp(8)Sp(8)𝐙_2]_{99}[Sp(8)Sp(8)]_{55}`$ gauge symmetry does not appear to be completely consistent<sup>‡††</sup><sup>‡††</sup>‡††The $`Sp(8)^4`$ model with the above spectrum but with missing 99 $`𝐙_2`$ discrete gauge symmetry was originally constructed in via the “rational construction” equivalent to the conformal field theory orbifold $`T^4/𝐙_2`$ at the special point in the moduli space of $`T^4`$’s corresponding to the $`SO(8)`$ symmetry. This model was discussed in the context of general $`T^4/𝐙_2`$ compactifications with $`b=2`$ $`B`$-flux in the second reference in , and more recently in . However, all of these references missed the importance of the 99 $`𝐙_2`$ discrete gauge symmetry, whose presence, as we have just pointed out, leads to a subtle inconsistency in the model., and, as we have discussed above, this gives additional evidence for the conclusion that we cannot have an O5<sup>+</sup>-plane at the conformal field theory $`𝐂^2/𝐙_2`$ orbifold fixed point<sup>‡‡‡</sup><sup>‡‡‡</sup>‡‡‡Note that the above conclusions also apply to the analogous $`b=4`$ model, which is constructed as follows. Consider the $`(T^2T^2)/𝐙_2`$ orbifold. The $`b=4`$ $`B`$-field can be described in terms of the two sets of Wilson lines $`\gamma _{S_i}`$ and $`\gamma _{T_i}`$. Choose one of these sets to be of the $`D_4`$ type, while the other set to be of the $`D_4^{}`$ type. Then it is not difficult to show that the consistent choice for the twisted Chan-Paton matrix is such that $`\nu ^2=+1`$. Consequently, we must place D5-branes at an O5<sup>+</sup>-plane in this model. The gauge group of this model is $`[Sp(4)Sp(4)𝐙_2𝐙_2]_{99}[Sp(4)Sp(4)]_{55}`$. In the 95 sector we again have half-hypermultiplets in real representations, which are charged non-trivially under the 99 $`𝐙_2𝐙_2`$ discrete gauge symmetry..
Let us summarize the results of this subsection. We have considered the $`\mathrm{\Omega }`$ orientifold of Type IIB on $`T^4/𝐙_2`$ with non-zero $`B`$-flux. The latter requires that $`n_{f+}`$ of the 16 O5-planes be of the O5<sup>+</sup> type. However, the latter cannot be placed at conformal field theory orbifold fixed points due to the non-zero twisted $`B`$-flux inside of the corresponding collapsed $`𝐏^1`$’s (while it is precisely an O5<sup>-</sup>-plane that can be consistently placed at such a fixed point). A possible way around this would be to blow up the orbifold fixed point and turn off the twisted $`B`$-flux. The price one would have to pay for this, however, is that the corresponding K3 no longer corresponds to an exactly solvable conformal field theory. Because of this, it is not entirely clear whether the corresponding models are completely consistent, albeit their massless spectra appear to be.
We have presented various pieces of evidence that attempts to interpret the aforementioned orientifolds in the context of the conformal field theory $`T^4/𝐙_2`$ orbifold with $`B`$-flux run into various inconsistencies visible already at the massless level, in particular, in the 59 sector. In fact, here we would like to suggest a simple geometric interpretation of these inconsistencies. Note that a peculiar feature of the aforementioned orientifolds is that we have D9-branes wrapping a torus (or, more precisely, an orbifold thereof) with $`B`$-flux together with D5-branes transverse to this torus. On the one hand, the gauge bundles of branes wrapped on such tori lack vector structure as the corresponding Stieffel-Whitney class is non-vanishing. On the other hand, the presence of D9- and D5-branes implies that we have the 59 sector, where the states transform in the bifundamental representations of the gauge group. The latter require non-trivial vector structure which is in conflict with the lack of vector structure for the 99 gauge bundles. However, in the next section we will be able to avoid this difficulty in non-trivial four dimensional $`𝒩=2`$ and $`𝒩=1`$ supersymmetric orientifolds with $`B`$-flux.
### B The $`𝐙_3`$ Models
In this subsection we will consider orientifolds of Type IIB on $`𝐑^{1,5}(T^4/𝐙_3)`$ with non-zero $`B`$-flux. We will denote the generator of the $`𝐙_3`$ orbifold group via $`\theta `$, whose action on the complex coordinates $`z_1,z_2`$ parametrizing $`T^4`$ is given by $`\theta z_1=\omega z_1`$, $`\theta z_2=\omega ^1z_2`$, where $`\omega \mathrm{exp}(2\pi i/3)`$. In fact, for our purposes here it will suffice to consider $`T^4=T^2T^2`$, where the first and the second 2-tori are parametrized by $`z_1`$ and $`z_2`$, respectively.
The orientifolds we will discuss here are defined as follows. The orientifold action is given by $`\mathrm{\Omega }J`$, where $`\mathrm{\Omega }`$ interchanges the left- and right-movers (and its action is the same as in the smooth K3 case), while $`J=J^{}`$, or $`J=RJ^{}`$. Here $`J^{}`$ acts as follows. Its action on the untwisted sector fields is trivial, however, it interchanges the $`\theta `$ twisted sector with its inverse $`\theta ^1`$ twisted sector. Thus, as explained in , the $`\mathrm{\Omega }J^{}`$ orientifold is precisely that discussed in . The geometric meaning of the $`J^{}`$ action was discussed in detail in . Next, $`R`$ is the simultaneous reflection of the coordinates on $`T^4`$: $`Rz_{1,2}=z_{1,2}`$. Note that in the $`\mathrm{\Omega }J^{}`$ orientifold we expect 32 D9-branes but no D5-branes, while in the $`\mathrm{\Omega }RJ^{}`$ orientifold we expect $`32/2^{b/2}`$ D5-branes but no D9-branes, where, as before, $`b`$ is the rank of the untwisted NS-NS $`B`$-field.
First, let us discuss the $`\mathrm{\Omega }J^{}`$ orientifold with $`b=2`$ (for definiteness we will assume that the $`B`$-flux is turned on in the direction of the first $`T^2`$). In this case we have 32 D9-branes but no D5-branes. The $`B`$-flux can be described in terms of the freely acting $`𝐙_2𝐙_2`$ orbifold with discrete torsion acting on the first $`T^2`$, whose action on the D9-brane Chan-Paton charges is given by the matrices $`\gamma _{S_a}`$. In fact, we are now going to show that these matrices must be of the $`D_4`$ type, that is, it would be inconsistent to choose them of the $`D_4^{}`$ type. This can be seen as follows. First, the first $`T^2`$ (as well as the second $`T^2`$) must have $`𝐙_3`$ symmetry. This implies that the vielbeins $`e_i`$ of the first $`T^2`$ are rotated by the action of $`\theta `$ as follows:
$$\theta e_1=e_2,\theta e_2=e_3,\theta e_3=e_1,$$
(155)
where $`e_3e_1e_2`$. Next, note that $`S_a`$ are half-lattice shifts in the directions of $`e_a`$. This implies that we must have the following relations between the $`𝐙_2𝐙_2`$ orbifold groups elements $`S_a`$ and the $`𝐙_3`$ orbifold group element $`\theta `$:
$$\theta S_1\theta ^1=S_2,\theta S_2\theta ^1=S_3,\theta S_3\theta ^1=S_1.$$
(156)
That is, the orbifold group elements $`\theta `$ and $`S_a`$ do not commute, and, in fact, they generate a non-Abelian tetrahedral subgroup of $`SU(2)`$ (or a double cover thereof)<sup>§\*</sup><sup>§\*</sup>§\*Note that the discrete torsion between the $`𝐙_2𝐙_2`$ elements is compatible with the $`𝐙_3`$ orbifold action.. This then implies that the corresponding Chan-Paton matrices must satisfy the following relations:
$$\gamma _\theta \gamma _{S_1}\gamma _\theta ^1=\gamma _{S_2},\gamma _\theta \gamma _{S_2}\gamma _\theta ^1=\gamma _{S_3},\gamma _\theta \gamma _{S_3}\gamma _\theta ^1=\gamma _{S_1}.$$
(157)
This, in particular, implies that
$$\gamma _{S_1}^2=\gamma _{S_2}^2=\gamma _{S_3}^2.$$
(158)
It then follows that the Wilson lines must be of the $`D_4`$ (and not $`D_4^{}`$) type<sup>§†</sup><sup>§†</sup>§†Recall that $`\gamma _{S_a}^2=\eta _{aa}I_{32}`$, where all three $`\eta _{aa}`$ equal $`1`$ for the $`D_4`$ type of Wilson lines, while two of them equal $`+1`$ and the third one equals $`1`$ for the $`D_4^{}`$ type of Wilson lines..
Next, let us discuss solutions to the above conditions. Up to equivalent representations we can write them as follows (here we are using the fact that $`\gamma _{S_3}=\eta _{12}\gamma _{S_1}\gamma _{S_2}`$):
$`\gamma _{S_1}=i\sigma _3I_{16},`$ (159)
$`\gamma _{S_2}=i\sigma _2I_{16},`$ (160)
$`\gamma _{S_3}=i\eta _{12}\sigma _1I_{16},`$ (161)
$`\gamma _\theta =\xi _\theta \mathrm{\Gamma }_\theta ,`$ (162)
where
$$\xi _\theta \left(\frac{1}{2}\right)\left[I_2+i\sigma _1+i\eta _{12}\sigma _2+i\eta _{12}\sigma _3\right],$$
(163)
and the $`16\times 16`$ matrix $`\mathrm{\Gamma }_\theta `$ is a diagonal matrix with non-zero entries taking values $`1,\omega ,\omega ^1`$. Note that the $`2\times 2`$ matrix $`\xi _\theta `$ has eigenvalues $`\omega `$ and $`\omega ^1`$, so that the matrix $`\gamma _\theta `$ has eigenvalues taking values $`1,\omega ,\omega ^1`$.
Note that in the above solution we still have to fix the form of the matrix $`\mathrm{\Gamma }_\theta `$. It is uniquely determined (up to equivalent representations) once we impose twisted tadpole cancellation conditions. The latter can be deduced as follows. Note that the orientifold with $`b=2`$ $`B`$-flux is the $`𝐙_2𝐙_2`$ orbifold of the orientifold without $`B`$-flux. The trace of the twisted Chan-Paton matrix $`\gamma _\theta `$ is then fixed as in the $`𝐙_3`$ model of :
$$\mathrm{Tr}(\gamma _\theta )=8.$$
(164)
Traces of all the other twisted matrices such as $`\gamma _{\theta S_a}`$ are then fixed unambiguously by the above relations. In particular:
$$\mathrm{Tr}(\gamma _{\theta S_a})=\mathrm{Tr}(\gamma _\theta \gamma _{S_a})=\eta _{12}\mathrm{Tr}(\gamma _\theta )=\eta _{12}\mathrm{Tr}(\mathrm{\Gamma }_\theta ).$$
(165)
Note that $`\mathrm{Tr}(\mathrm{\Gamma }_\theta )=\mathrm{Tr}(\gamma _\theta )=8`$. This implies that up to equivalent representations we have
$$\mathrm{\Gamma }_\theta =\mathrm{diag}(\omega ,\omega ^1)I_8.$$
(166)
Note that in the diagonal basis $`\gamma _\theta `$ can be written as $`\gamma _\theta =\mathrm{diag}(1,1,\omega ,\omega ^1)I_8`$.
Next, let us determine the massless spectrum of this model. First, let us discuss the closed string spectrum. It contains the six dimensional $`𝒩=1`$ supergravity multiplet, one untwisted tensor multiplet, 2 untwisted hypermultiplets, 9 twisted tensor multiplets and 9 twisted hypermultiplets. The open string spectrum can be determined as follows. First, note that the $`2\times 2`$ matrices
$`\gamma _1i\sigma _3,\gamma _2i\sigma _2,\gamma _3i\eta _{12}\sigma _1,\gamma _\theta ^{(k)}\omega ^k\xi _\theta `$ (167)
define three irreducible two dimensional representations of the tetrahedral subgroup $`𝒯`$ of $`SU(2)`$ labeled by the integer $`k=0,1,2`$. The aforementioned set of matrices $`\gamma _{S_a},\gamma _\theta `$ with $`\mathrm{\Gamma }_\theta `$ given by (166) corresponds to taking 8 copies of the two dimensional representation labeled by $`k=1`$ together with 8 copies of the two dimensional representation labeled by $`k=2`$. The $`𝐙_2𝐙_2`$ orbifold action (that is, the action of the Chan-Paton matrices $`\gamma _{S_a}`$) breaks the original $`SO(32)`$ gauge group down to its $`Sp(16)𝐙_2`$ subgroup. The further $`𝐙_3`$ orbifold action (that is, the action of the twisted Chan-Paton matrix $`\gamma _\theta `$) breaks this gauge group down to its $`U(8)`$ subgroup<sup>§‡</sup><sup>§‡</sup>§‡Here we note that the $`𝐙_3`$ twist breaks the $`𝐙_2`$ discrete subgroup in the product $`Sp(16)𝐙_2`$. This point will become important in the next subsection, and we will discuss it there in more detail. However, the fate of the $`𝐙_2`$ discrete gauge symmetry will not be important for our purposes here as no massless states carry non-trivial $`𝐙_2`$ charges.. In fact, the open string massless spectrum is given by the $`𝒩=1`$ gauge supermultiplet in the adjoint of $`U(8)`$ plus one hypermultiplet in $`\mathrm{𝟑𝟔}`$ of $`U(8)`$. (Note that the adjoint of $`Sp(16)`$ decomposes in terms of the $`U(8)`$ representations as follows: $`\mathrm{𝟏𝟑𝟔}=\mathrm{𝟔𝟒}(0)\mathrm{𝟑𝟔}(+2)\overline{\mathrm{𝟑𝟔}}(2)`$, where we have given the $`U(1)`$ charges in parentheses, and the latter are normalized so that the fundamental of $`SU(8)`$ has the $`U(1)`$ charge $`+1`$.) One can directly derive<sup>§§</sup><sup>§§</sup>§§We will give the details of this derivation in the next subsection. this spectrum by keeping the states invariant under the action of the non-Abelian tetrahedral group $`𝒯`$. In particular, note that the action of the $`\gamma _\theta `$ matrix on the $`Sp(16)`$ part of the gauge group, which is given by the $`16\times 16`$ matrix $`\mathrm{\Gamma }_\theta `$, breaks $`Sp(16)`$ down to $`U(8)`$ as $`\mathrm{\Gamma }_\theta =\mathrm{diag}(\omega ,\omega ^1)I_8`$. Alternatively, we can use the following trick. The open string partition function $`𝒵[𝒯]`$ of the full $`𝒯`$ orbifold model can be expressed in terms of the partition functions $`𝒵[𝐙_2𝐙_2]`$, $`𝒵[𝐙_3]`$ and $`𝒵[1]`$ as follows:
$$𝒵[𝒯]=𝒵[𝐙_3]+\frac{1}{3}𝒵[𝐙_2𝐙_2]\frac{1}{3}𝒵[1],$$
(168)
where $`𝒵[1]`$ is the partition function of the model corresponding to the toroidal compactification without the $`B`$-flux (this model has $`𝒩=2`$ supersymmetry and $`SO(32)`$ gauge group), $`𝒵[𝐙_2𝐙_2]`$ is the partition function of the model corresponding to the toroidal compactification with $`b=2`$ $`B`$-flux, that is, the $`𝐙_2𝐙_2`$ freely acting orbifold model (this model has $`𝒩=2`$ supersymmetry and $`Sp(16)𝐙_2`$ gauge group), and $`𝒵[𝐙_3]`$ is the partition function of the model corresponding to the $`T^4/𝐙_3`$ compactification without the $`B`$-flux (this is the $`𝐙_3`$ model of with $`𝒩=1`$ supersymmetry, $`SO(16)U(8)`$ gauge group and massless hypermultiplets in $`(\mathrm{𝟏𝟔},\mathrm{𝟖})(\mathrm{𝟏},\mathrm{𝟐𝟖})`$ of $`SO(16)U(8)`$). If we now write the spectra of the $`𝒩=2`$ models in the $`𝒩=1`$ language, we can then read off the numbers of gauge bosons and massless hypermultiplets in the full $`𝒯`$ orbifold model from $`𝒵[𝒯]`$ defined as above.
Here we note that the above closed plus open string spectrum, which is the same as that given in , is free of the irreducible $`R^4`$ and $`F^4`$ anomalies. The $`U(1)`$ factor is anomalous as usual, and is broken via the generalized Green-Schwarz mechanism in a way similar to the $`𝐙_3`$ model of without the $`B`$-flux.
Next, let us discuss the $`\mathrm{\Omega }J^{}`$ orientifold model with $`b=4`$ $`B`$-flux. The latter can be described in terms of two $`𝐙_2𝐙_2`$ freely acting orbifolds acting on the first respectively second $`T^2`$. We can choose the corresponding Chan-Paton matrices as follows:
$`\gamma _{S_1}=i\sigma _3I_2I_8,`$ (169)
$`\gamma _{S_2}=i\sigma _2I_2I_8,`$ (170)
$`\gamma _{S_3}=i\eta _{12}\sigma _1I_2I_8,`$ (171)
$`\gamma _{T_1}=I_2i\sigma _3I_8,`$ (172)
$`\gamma _{T_2}=I_2i\sigma _2I_8,`$ (173)
$`\gamma _{T_3}=I_2i\eta _{12}^{}\sigma _1I_8,`$ (174)
$`\gamma _\theta =\xi _\theta \xi _\theta ^{}\stackrel{~}{\mathrm{\Gamma }}_\theta ,`$ (175)
where $`\xi _\theta ^{}`$ is given by the the same expression (163) with $`\eta _{12}^{}`$ instead of $`\eta _{12}`$. Note the $`4\times 4`$ matrix $`\xi _\theta \xi _\theta ^{}`$ has eigenvalues $`1,1,\omega ,\omega ^1`$. This then implies that to satisfy the twisted tadpole cancellation condition $`\mathrm{Tr}(\gamma _\theta )=8`$, we must take the $`8\times 8`$ matrix $`\stackrel{~}{\mathrm{\Gamma }}_\theta `$ to be the identity matrix $`I_8`$. It is then not difficult to see that the gauge group of this model is $`SO(8)`$, and we have no massless hypermultiplets in the open string spectrum<sup>§¶</sup><sup>§¶</sup>§¶Note that in the diagonal basis the $`16\times 16`$ matrix $`\xi _\theta ^{}\stackrel{~}{\mathrm{\Gamma }}_\theta `$ can be written as $`\mathrm{diag}(\omega ,\omega ^1)I_8`$, which is consistent with $`\mathrm{\Gamma }_\theta `$ given in (166).. In particular, note that the Wilson lines $`\gamma _{S_a}`$ and $`\gamma _{T_a}`$ break the original $`SO(32)`$ gauge group down to $`SO(8)(𝐙_2)^2`$. The action of the $`𝐙_3`$ orbifold group on the $`SO(8)`$ part of the gauge group is trivial<sup>§∥</sup><sup>§∥</sup>§∥As in the $`b=2`$ case, however, it breaks the $`(𝐙_2)^2`$ discrete gauge symmetry. as it is given by $`\stackrel{~}{\mathrm{\Gamma }}_\theta =I_8`$. The closed string spectrum is the same as in the $`b=2`$ model. Note that the spectrum of the $`b=4`$ model, which is the same as that given in , is free of the irreducible $`R^4`$ and $`F^4`$ anomalies.
Now we would like to discuss the $`\mathrm{\Omega }RJ^{}`$ orientifolds with $`B`$-flux. Let us start with the $`b=2`$ case. As before, let us assume that the $`B`$-field is turned on in the direction of the first $`T^2`$. Let $`e_i`$ be the vielbeins of the first $`T^2`$, and $`d_i`$ be the vielbeins of the second $`T^2`$. As before, we will use the notation $`e_3e_1e_2`$, $`d_3d_1d_2`$. In this model we have 16 O5-planes located at the 16 points fixed under the action of the reflection $`R`$. These fixed points are located at $`(0,0),(0,d_a/2),(e_a/2,0),(e_a/2,d_a/2)`$. Note that 12 out of these O5-planes must be of the O5<sup>-</sup> type, while 4 must be of the O5<sup>+</sup> type. Together with the requirement that the entire background be $`𝐙_3`$ symmetric (so that the further $`𝐙_3`$ orbifold is consistent), this uniquely fixes the allowed distribution of O5-planes. Thus, the O5-planes located at the fixed points $`(0,0),(0,d_a/2)`$ must be of the O5<sup>+</sup> type, while the rest of the O5-planes are of the O5<sup>-</sup> type.
In this orientifold we have 16 D5-branes. The latter must be distributed in a $`𝐙_3`$ symmetric fashion. First, let us consider placing all 16 D5-branes on top of the O5<sup>+</sup>-plane at the origin $`(0,0)`$ of $`T^4/𝐙_3`$. Before the $`𝐙_3`$ orbifold projection the gauge group is $`Sp(16)`$. The action of the $`𝐙_3`$ orbifold on the Chan-Paton charges is given by the twisted $`16\times 16`$ Chan-Paton matrix $`\gamma _\theta `$. The only constraint on this matrix comes from the twisted tadpole cancellation condition, which in this case reads
$$\mathrm{Tr}(\gamma _\theta )=8.$$
(176)
Note that this is the same twisted tadpole cancellation condition as in the case of D9-branes wrapped on $`T^4/𝐙_3`$ without $`B`$-field except for the extra minus sign<sup>§\**</sup><sup>§\**</sup>§\**Here we note that generally twisted tadpole cancellation conditions for D5-branes transverse to the $`𝐂^2/𝐙_M`$ orbifold are different from those for D9-branes with the $`𝐂^2/𝐙_M`$ orbifold in their world-volumes. However, for a particular case of the $`𝐙_3`$ orbifold group they actually coincide with the overall sign depending on the type of the corresponding orientifold plane.. This minus sign is, in fact, due to the O5-plane here being of the O5<sup>+</sup> type (while the twisted tadpole cancellation condition in the case of D9-branes was derived in for the configuration involving the O9<sup>-</sup>-plane). The above tadpole cancellation condition uniquely fixes the twisted Chan-Paton matrix (up to equivalent representations):
$$\gamma _\theta =\mathrm{diag}(\omega ,\omega ^1)I_8.$$
(177)
Thus, just as in the case of $`b=2`$ $`\mathrm{\Omega }J^{}`$ orientifold, this model with D5-branes has $`U(8)`$ gauge group and one massless hypermultiplet in $`\mathrm{𝟑𝟔}`$ of $`U(8)`$. (The closed string sector is the same as in the $`\mathrm{\Omega }J^{}`$ orientifold.) Note that if we attempted to place some D5-branes in the bulk or at other O5-planes, we would not have been able to cancel twisted tadpoles. Indeed, D5-branes away from the origin must be placed in a $`𝐙_3`$ symmetric fashion, that is, the number of D5-branes away from the origin must be a multiple of 3. These branes are then permuted by the action of the $`𝐙_3`$ orbifold, and the corresponding part of the twisted Chan-Paton matrix $`\gamma _\theta `$ is traceless<sup>§††</sup><sup>§††</sup>§††Note that this is the case even if we place D5-branes (in a fashion compatible with the action of the $`\mathrm{\Omega }RJ^{}`$ orientifold) at $`𝐙_3`$ orbifold fixed points not fixed under $`R`$. That is, the corresponding part of the twisted Chan-Paton matrix $`\gamma _\theta `$ must be traceless to satisfy tadpole cancellation conditions.. The part corresponding to the D5-branes at the origin then cannot have trace equal $`8`$, so that the twisted tadpole cancellation conditions cannot be satisfied. The point we have just discussed can be understood in terms of the effective field theory language as follows. Note that moving D5-branes off the O5<sup>+</sup>-plane (located at the origin) into the bulk would correspond to giving the appropriate VEV to a massless hypermultiplet. The only massless hypermultiplet in this model (with all D5-branes at the origin) is in $`\mathrm{𝟑𝟔}`$ of $`U(8)`$. If we could give a VEV to this hypermultiplet, it would break $`U(8)`$ down to $`SO(8)`$. (Note that under the breaking $`SU(8)SO(8)`$ we have $`\mathrm{𝟑𝟔}=\mathrm{𝟏}+\mathrm{𝟑𝟓}`$.) However, to satisfy the D-flatness conditions we must have at least two hypermultiplets in $`\mathrm{𝟑𝟔}`$ of $`U(8)`$, so that the aforementioned Higgsing is not possible<sup>§‡‡</sup><sup>§‡‡</sup>§‡‡One way to see this is as follows. Note that the $`U(1)`$ subgroup of the $`U(8)`$ gauge group is anomalous. When all D5-branes are on top of the O5<sup>+</sup>-plane at the origin of $`T^4/𝐙_3`$, this $`U(1)`$ factor is broken via the generalized Green-Schwarz mechanism involving the twisted closed string sector hypermultiplet coming from the $`𝐙_3`$ fixed point at the origin. However, if we move D5-branes away from the origin, the $`U(1)`$ breaking can no longer involve this twisted hypermultiplet - the latter is localized at the corresponding fixed point. In fact, the singlet of $`SO(8)`$ in the decomposition $`\mathrm{𝟑𝟔}=\mathrm{𝟏}+\mathrm{𝟑𝟓}`$ under $`SU(8)SO(8)`$ carries a non-zero $`U(1)`$ charge (which is equal $`+2`$ in the aforementioned normalization). This is precisely the singlet whose VEV would measure the separation between the O5<sup>+</sup>-plane at the origin and the D5-branes in the bulk had the Higgsing been possible. However, in this case this singlet would also have to be the one eaten in Higgsing the $`U(1)`$ factor, so that we would have only the $`SO(8)`$ gauge bosons but no massless hypermultiplets coming from D5-branes in the bulk. This, however, would mean that there is no modulus corresponding to the separation between the O5<sup>+</sup>-plane and D5-branes. We, therefore, conclude that the aforementioned Higgsing is indeed impossible.. This is in accord with the fact that no distribution of D5-branes in the bulk giving rise to the $`SO(8)`$ gauge symmetry could possibly be $`𝐙_3`$ symmetric. Thus, D5-branes are stuck at the O5<sup>+</sup>-plane located at the origin of $`T^4/𝐙_3`$ in this model.
Finally, let us discuss the $`b=4`$ $`\mathrm{\Omega }RJ^{}`$ orientifold. In this case we have 10 O5<sup>-</sup>-planes and 6 O5<sup>+</sup>-branes, and the unique distribution of O5-planes compatible with the $`𝐙_3`$ symmetry is the following: those at the fixed points $`(0,0),(e_a/2,d_a/2)`$ are of the O5<sup>-</sup> type, while those at the fixed points $`(0,d_a/2),(e_a/2,0)`$ are of the O5<sup>+</sup> type. In this case we have only 8 D5-branes all of which must be placed at the origin, so that the twisted tadpole cancellation condition
$$\mathrm{Tr}(\gamma _\theta )=8$$
(178)
can be satisfied. Note that the sign in this case is plus instead of minus as in the $`b=2`$ case as here the O5-plane at which the D5-planes are placed is of the O5<sup>-</sup> type. The unique solution to the above tadpole condition is $`\gamma _\theta =I_8`$, so that the gauge group of this model is $`SO(8)`$ with no massless hypermultiplets in the open string sector. The closed string sector spectrum is the same as in the $`b=2`$ case. In fact, the massless spectrum of this model is the same as that of the $`\mathrm{\Omega }J^{}`$ orientifold with $`b=4`$. Note that, just as in the $`b=2`$ case, in the $`b=4`$ model D5-branes are also stuck at the origin of $`T^4/𝐙_3`$ \- there are no massless open string hypermultiplets in this model.
Before we end this subsection let us note that in the case of $`\mathrm{\Omega }J^{}`$ orientifolds the inability to Higgs the gauge group is interpreted as impossibility of turning on Wilson lines compatible with the $`𝐙_3`$ symmetry in such a way that all tadpoles are canceled.
### C The $`𝐙_6`$ Models
In this subsection we will discuss the $`\mathrm{\Omega }J^{}`$ orientifolds of Type IIB on $`T^4/𝐙_6`$ with $`b=2,4`$. As before, we will assume that $`T^4=T^2T^2`$. Let $`g`$ be the generator of $`𝐙_6`$. Then, since $`𝐙_6𝐙_2𝐙_3`$, we can write $`g=R\theta `$, where $`R`$ and $`\theta `$ are the generators of the $`𝐙_2`$ and $`𝐙_3`$ subgroups, respectively, with the following actions on the coordinates $`z_1,z_2`$ parametrizing the two 2-tori: $`Rz_{1,2}=z_{1,2}`$, $`\theta z_1=\omega z_1`$, $`\theta z_2=\omega ^1z_2`$ ($`\omega \mathrm{exp}(2\pi i/3)`$). The action of $`J^{}`$ on the closed string untwisted as well as $`R`$ twisted sector fields is trivial, while $`J^{}`$ interchanges the $`\theta `$ twisted sector with the $`\theta ^1`$ twisted sector, as well as the $`g`$ twisted sector with the $`g^1`$ twisted sector.
To begin with let us discuss the closed sting sector of the above orientifold. The untwisted sector gives rise to the six dimensional $`𝒩=1`$ supergravity multiplet plus one tensor multiplet and 2 hypermultiplets. The $`g`$ and $`g^1`$ twisted sectors together give rise to one tensor multiplet and one hypermultiplet. The $`\theta `$ and $`\theta ^1`$ twisted sectors together give rise to 5 tensor multiplets and 5 hypermultiplets. As to the $`R`$ twisted sector, we must consider $`b=0,2,4`$ cases separately as the $`\mathrm{\Omega }`$ projection acts differently on the corresponding fixed points for different values of $`b`$.
In the $`b=0`$ case before the $`𝐙_3`$ projection in the $`R`$ twisted sector we have 16 fixed points. At all of these fixed points we have O5<sup>-</sup>-planes, so that each of them gives rise to a hypermultiplet (but no tensor multiplets). Note that the fixed point at the origin is invariant under the $`𝐙_3`$ twist $`\theta `$. The other 15 fixed points fall into 5 distinct groups, each group containing 3 fixed points which are permuted by the action of $`\theta `$. In each of these 5 groups we can form one linear combination of the corresponding 3 fixed points which is invariant under $`\theta `$. Thus, we have total of 6 hypermultiplets coming from the $`R`$ twisted sector - one from the origin, and the other 5 from the rest of the fixed points.
Next, let us consider the $`b=2`$ case. Here we have 12 O5<sup>-</sup>-planes and 4 O5<sup>+</sup>-planes. At the fixed point at the origin we have an O5<sup>+</sup>-plane. Thus, this fixed point gives rise to a tensor multiplet. The other 3 fixed points at which we have O5<sup>+</sup>-planes together give rise to another tensor multiplet. Finally, the 12 fixed points at which we have O5<sup>-</sup>-planes together give rise to 4 hypermultiplets. Thus, the $`R`$ twisted sector gives rise to 4 hypermultiplets and 2 tensor multiplets in the $`b=2`$ case.
In the $`b=4`$ case we have 10 O5<sup>-</sup>-planes and 6 O5<sup>+</sup>-planes. At the fixed point at the origin we have an O5<sup>-</sup>-plane. This fixed point, therefore, gives rise to a hypermultiplet. The other 9 fixed points at which we have O5<sup>-</sup>-planes together give rise to 3 additional hypermultiplets. Finally, the 6 fixed points at which we have O5<sup>+</sup>-planes together give rise to 2 tensor multiplets. Thus, the $`R`$ twisted sector gives rise to 4 hypermultiplets and 2 tensor multiplets in the $`b=4`$ case, which is the same as in the $`b=2`$ case<sup>¶\*</sup><sup>¶\*</sup>¶\*This corrects the error in , where the $`R`$ twisted sector in the $`b=2,4`$ $`𝐙_6`$ models was thought to give rise to 6 hypermultiplets and no tensor multiplets..
Next, let us discuss the open string sector in the $`b=2`$ model. In this model we have 32 D9-branes and 16 D5-branes. Note that from our discussion of the corresponding $`𝐙_3`$ model it follows that all D5-branes in this models must be placed at the O5<sup>+</sup>-plane at the origin of $`T^4/𝐙_6`$. This then, following our discussion in subsection A, implies that the twisted Chan-Paton matrix $`\gamma _{R,5}`$ must have eigenvalues $`\pm 1`$, and so must the matrix $`\gamma _{R,9}`$. From this it follows that the Wilson lines in the 99 sector must be of the $`D_4`$ type, which is consistent with our discussion in subsection B<sup>¶†</sup><sup>¶†</sup>¶†These points were missed in in the discussion of this model. There it was erroneously assumed that the O5-plane at the origin, at which all D5-branes were placed, is of the O5<sup>-</sup> type. Consequently, the matrices $`\gamma _{R,5}`$ and $`\gamma _{R,9}`$ were assumed to have eigenvalues $`\pm i`$, and, in the language we are using here, the Wilson lines in the 99 sector were assumed to be of the $`D_4^{}`$ type. From the above discussions it should be clear that such a setup would be inconsistent as it is not even $`𝐙_3`$ symmetric (so that the $`𝐙_3`$ orbifolding procedure would be inconsistent). In fact, it is not difficult to show that with these assumptions the corresponding massless spectrum would have to be anomalous (as some of the tadpoles would not be canceled). The corresponding spectrum given in , however, is free of, say, the $`R^4`$ anomaly. One of the errors made in that had lead to this seemingly consistent spectrum, as we have already mentioned, was the incorrect computation of the number of twisted tensor multiplets in the closed string spectrum. Another error, related to the multiplicity of states, was made in the discussion of the 59 sector. In fact, this point is rather non-trivial, and we will discuss it in more detail in a moment. Finally, the $`\theta `$ projection was carried out erroneously in , which was already noticed in . All these errors added up to give the erroneous spectrum reported in ..
Here we can ask whether the $`b=2`$ $`𝐙_6`$ model is consistent once we make the aforementioned choices. First, recall that the $`b=2`$ $`𝐙_2`$ model with the Wilson lines of the $`D_4`$ type (and the twisted Chan-Paton matrices with $`\nu ^2=+1`$ \- see subsection A) suffers from the presence of half-hypermultiplets in real representations in the 59 sector. The $`b=2`$ $`𝐙_6`$ model, therefore, is expected to have a similar problem as well. However, as we have already mentioned, the 99 $`𝐙_2`$ discrete gauge symmetry is broken by the $`𝐙_3`$ twist. This might at first seem to imply that in the $`𝐙_6`$ model unlike the $`𝐙_2`$ model we might be able to avoid the difficulty with the 59 half-hypermultiplets. In fact, if this were so, then this model at first might seem to be consistent even for the conformal field theory orbifold - recall from subsection B that in the $`b=2`$ $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ model we cannot move D5-branes away from the O5<sup>+</sup>-plane at the origin of K3 (and, similarly, in the $`b=2`$ $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ model we cannot Higgs the 99 gauge group by turning on Wilson lines). This then implies that in the $`b=2`$ $`𝐙_6`$ model a priori we do not have one of the problems we have encountered in the corresponding $`𝐙_2`$ model, in particular, that related to the inconsistencies arising once we move D5-branes from an O5<sup>+</sup>-plane to an O5<sup>-</sup>-plane, or vice-versa. However, such a conclusion would immediately run into a puzzle with our discussion of the role of the twisted $`B`$-flux - recall that we do not expect to be able to consistently have O5<sup>+</sup>-planes within the conformal field theory orbifold if the orbifold group contains the $`𝐙_2`$ generator $`R`$ (albeit, O5<sup>+</sup>-planes are perfectly consistent with conformal field theory orbifolds of odd order). In fact, as we will see in a moment, in the $`b=2`$ $`𝐙_6`$ model there is indeed a subtle inconsistency in the 59 sector<sup>¶‡</sup><sup>¶‡</sup>¶‡It is then not difficult to show that similar conclusions hold for the $`b=4`$ $`𝐙_6`$ model as well..
Thus, let us understand the 59 sector in this model. In fact, to understand the point we would like to make here it suffices to consider the 59 sector before the $`𝐙_2`$ orbifold projection. We can alternatively view this as introducing D5-brane probes in the $`b=2`$ $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ model. Note that the 59 sector states are in bifundamental representations of the 55 and 99 gauge groups. In fact, for our purposes here the precise 55 quantum numbers are not going to be relevant. This is related to the fact that in the 55 sector we have just the $`𝐙_3`$ orbifold projection, which is straightforward to carry out. However, in the 99 sector we have the additional projections coming from the $`𝐙_2𝐙_2`$ freely acting orbifold. Thus, we would like to understand how $`\mathrm{𝟑𝟐}`$ of $`SO(32)`$ decomposes under the gauge group left unbroken after the $`𝐙_2𝐙_2`$ freely acting orbifold as well as the $`𝐙_3`$ orbifold projections<sup>¶§</sup><sup>¶§</sup>¶§Note that before any of the orbifold projections the 95 sector states are in the following half-hypermultiplet of $`SO(32)_{99}Sp(16)_{55}`$: $`\frac{1}{2}(\mathrm{𝟑𝟐};\mathrm{𝟏𝟔})`$.. Before we do this, however, it is instructive to consider the analogous decomposition for the adjoint of $`SO(32)`$. Note that the Wilson line $`\gamma _{S_1,9}`$ breaks $`SO(32)`$ down to $`U(16)`$. On the other hand, the twisted Chan-Paton matrix $`\gamma _{\theta ,9}`$ breaks $`SO(32)`$ down to $`SO(16)U(8)`$. Now, the maximal common subgroup of $`U(16)`$ and $`SO(16)U(8)`$ is $`U(8)U(8)`$. Under the breaking $`SO(32)U(8)U(8)`$ the adjoint of $`SO(32)`$ decomposes as follows:
$`\mathrm{𝟒𝟗𝟔}=`$ $`(\mathrm{𝟔𝟒},\mathrm{𝟏})_{1,1}(\mathrm{𝟏},\mathrm{𝟔𝟒})_{1,1}(\mathrm{𝟖},\overline{\mathrm{𝟖}})_{1,\omega ^1}(\overline{\mathrm{𝟖}},\mathrm{𝟖})_{1,\omega }`$ (181)
$`(\mathrm{𝟐𝟖},\mathrm{𝟏})_{1,1}(\mathrm{𝟏},\mathrm{𝟐𝟖})_{1,1}(\mathrm{𝟖},\mathrm{𝟖})_{1,\omega }`$
$`(\overline{\mathrm{𝟐𝟖}},\mathrm{𝟏})_{1,1}(\mathrm{𝟏},\overline{\mathrm{𝟐𝟖}})_{1,1}(\overline{\mathrm{𝟖}},\overline{\mathrm{𝟖}})_{1,\omega ^1},`$
where the subscript indicates the $`𝐙_2`$ valued phase due to the $`\gamma _{S_1,9}`$ projection as well as the $`𝐙_3`$ valued phase due to the $`\gamma _{\theta ,9}`$ projection. The states with the $`𝐙_2`$ phase $`1`$ are all heavy, so that the massless states all have the $`𝐙_2`$ phase $`+1`$. Such states with the $`𝐙_3`$ phase 1 are the gauge bosons of $`U(8)U(8)`$, while the states with the $`𝐙_3`$ phases $`\omega `$ and $`\omega ^1`$ combine into one massless hypermultiplet in $`(\mathrm{𝟖},\overline{\mathrm{𝟖}})`$ of $`U(8)U(8)`$.
Next, the gauge group left unbroken after the full $`𝒯`$ orbifold projection can be determined as follows. The action of the second Wilson line $`\gamma _{S_2,9}`$ amounts to permuting the two $`U(8)`$ subgroups in $`U(8)U(8)`$ (left unbroken by $`\gamma _{S_1,9}`$ and $`\gamma _{\theta ,9}`$) accompanied by the complex conjugation. Thus, for instance, $`(\mathrm{𝟖},\mathrm{𝟏})`$ of $`U(8)U(8)`$ is mapped to $`(\mathrm{𝟏},\overline{\mathrm{𝟖}})`$ by the action of $`\gamma _{S_2,9}`$. This implies that the final unbroken gauge group is $`U(8)`$, and the massless matter consists of one hypermultiplet in $`\mathrm{𝟑𝟔}`$ of $`U(8)`$. Note that normally we would expect the appearance of the $`𝐙_2`$ discrete gauge subgroup in the breaking $`U(8)U(8)U(8)_{\mathrm{diag}}𝐙_2`$. However, as we will show in a moment, this $`𝐙_2`$ discrete gauge group is actually broken in the case under consideration. Note that it is the 59 sector massless states that are expected to carry non-trivial 99 $`𝐙_2`$ discrete gauge quantum numbers. However, as we will see momentarily, the 59 sector states in this model do not carry well defined gauge quantum numbers at all.
To see this, let us discuss the decomposition of $`\mathrm{𝟑𝟐}`$ of $`SO(32)`$ under $`SO(32)U(8)U(8)U(8)`$. Under the first breaking we have:
$$\mathrm{𝟑𝟐}=(\mathrm{𝟖},\mathrm{𝟏})_1(\overline{\mathrm{𝟖}},\mathrm{𝟏})_1(\mathrm{𝟏},\mathrm{𝟖})_\omega (\mathrm{𝟏},\overline{\mathrm{𝟖}})_{\omega ^1}.$$
(182)
Here we have only shown the $`𝐙_3`$ valued phases due to the $`\gamma _{\theta ,9}`$ projection. In fact, we did not give the phases (which are actually $`𝐙_4`$ valued) due to the $`\gamma _{S_1,9}`$ projection for the reason that the Wilson lines (more precisely, the freely acting $`𝐙_2𝐙_2`$ orbifold generators) do not act in the 59 sector (just as they do not act in the 55 sector). One way to see this is to note that the Wilson lines can only act on states with Kaluza-Klein momenta (but not windings) coming from the compact directions. This can also be seen by noting that continuously turning on Wilson lines (which is equivalent to Higgsing the 99 gauge group by giving VEVs to the 99 hypermultiplets) in, say, the $`b=0`$ $`𝐙_2`$ model does not change the number of 59 hypermultiplets. In fact, this is precisely the key reason for the problem we are going to point out next. Note that under the action of $`\gamma _{S_2,9}`$ the state $`(\mathrm{𝟖},\mathrm{𝟏})_1`$ is mapped to the state $`(\mathrm{𝟏},\overline{\mathrm{𝟖}})_{\omega ^1}`$, and, similarly, the state $`(\overline{\mathrm{𝟖}},\mathrm{𝟏})_1`$ is mapped to the state $`(\mathrm{𝟏},\mathrm{𝟖})_\omega `$. Thus, the linear combinations that carry well defined gauge quantum numbers under the unbroken $`U(8)`$ gauge group are given by:
$`|\mathrm{𝟖}_\pm {\displaystyle \frac{1}{\sqrt{2}}}\left(|(\mathrm{𝟖},\mathrm{𝟏})_1\pm |(\mathrm{𝟏},\overline{\mathrm{𝟖}})_{\omega ^1}\right),`$ (183)
$`|\overline{\mathrm{𝟖}}_\pm {\displaystyle \frac{1}{\sqrt{2}}}\left(|(\overline{\mathrm{𝟖}},\mathrm{𝟏})_1\pm |(\mathrm{𝟏},\mathrm{𝟖})_\omega \right),`$ (184)
where the subscript on the left hand side indicates the $`𝐙_2`$ gauge quantum numbers with the $`𝐙_2`$ subgroup arising in the breaking $`U(8)U(8)U(8)_{\mathrm{diag}}𝐙_2`$. Note, however, that the states $`|\mathrm{𝟖}_\pm `$ and $`|\overline{\mathrm{𝟖}}_\pm `$ do not carry well defined $`𝐙_3`$ quantum numbers. This implies that if in the 59 sector we perform the $`𝐙_3`$ orbifold projection by keeping the $`𝐙_3`$ invariant states, then the latter will not have consistent couplings to the $`U(8)`$ gauge bosons. If instead we keep the above linear combinations with consistent couplings to the $`U(8)`$ gauge bosons, then these states will be incompatible with the $`𝐙_3`$ orbifold projection. Either way we have an inconsistency in the 59 sector of this model, which, in turn, is consistent with our discussions in subsection A<sup>¶¶</sup><sup>¶¶</sup>¶¶Here the following remark is in order. Consider probe D1-branes in the $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ models with $`B`$-flux. Then in the 19 open string sector we would encounter the same problem as that we have just discussed for D5-brane probes. This might signal that there could be a non-perturbative inconsistency in these $`𝐙_3`$ models, which would have to be visible in the dual heterotic compactification. In particular, on the heterotic side this inconsistency might manifest itself via the world-sheet theory of the fundamental heterotic string being inconsistent. Note, however, that such a problem is absent in the $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ orientifolds with $`B`$-flux which could, therefore, be consistent even non-perturbatively. Here we note that the $`\mathrm{\Omega }J^{}`$ and $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ models with $`B`$-flux are different at the massive (which are non-BPS) levels, which might be the reason why one set of these models could have different non-perturbative behavior compared with the other..
Before we end this subsection, we would like to discuss the massless spectra of the $`𝐙_6`$ models with $`B`$-flux. More precisely, here we can discuss the 99 and 55 as well as closed string sectors. The closed string sector in both $`b=2`$ and $`b=4`$ models contains (together with the six dimensional $`𝒩=1`$ supergravity multiplet) 9 tensor multiplets and 12 hypermultiplets. The gauge group of the $`b=2`$ model is $`[U(4)U(4)]_{99}[U(4)U(4)]_{55}`$. The massless hypermultiplets in the 99 and 55 sectors are given by:
$`(\mathrm{𝟒},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (185)
$`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟒},\mathrm{𝟒})_{55}.`$ (186)
As to the 95 sector, the irreducible $`R^4`$ and $`F^4`$ anomaly cancellation would require that we have the following massless hypermultiplets:
$`(\mathrm{𝟒},\mathrm{𝟏};\mathrm{𝟒},\mathrm{𝟏})_{95},`$ (187)
$`(\mathrm{𝟏},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟒})_{95}.`$ (188)
However, as we discussed above, the 59 sector states in this model do not carry well defined gauge quantum numbers, hence inconsistency in this model.
Finally, in the $`b=4`$ model the gauge group is $`U(4)_{99}U(4)_{55}`$, and there are no massless hypermultiplets in the 99 and 55 sectors<sup>¶∥</sup><sup>¶∥</sup>¶∥This corrects the error in the corresponding spectrum given in .. As to the 95 sector, the irreducible $`R^4`$ and $`F^4`$ anomaly cancellation would require that we have two massless hypermultiplets in $`(\mathrm{𝟒};\mathrm{𝟒})_{95}`$. Note, however, that as in the $`b=2`$ model the 95 sector states in this model do not carry well defined gauge charges. This is due to the fact that the 99 $`(𝐙_2)^2`$ discrete gauge symmetry is completely broken by the $`𝐙_3`$ twist. This, in turn, is consistent with the fact that the 95 sector states do not carry well defined gauge quantum numbers - it would otherwise be difficult to understand how we can have two copies of the aforementioned 95 hypermultiplets as there is no discrete gauge symmetry to distinguish the corresponding vertex operators<sup>¶\**</sup><sup>¶\**</sup>¶\**Here we note that before the $`𝐙_3`$ orbifold projection, that is, in the corresponding $`b=4`$ $`𝐙_2`$ model we have four distinct vertex operators in the 59 sector distinguished by their charges under the 99 $`(𝐙_2)^2`$ discrete symmetry. To have an anomaly free model, however, only two of these vertex operators would have to survive the $`𝐙_3`$ projection, while the other two would have to be projected out. Thus, the $`𝐙_3`$ twist would have to act non-trivially on the corresponding quantum numbers. In fact, it indeed does, except that its action is incompatible with the 99 gauge quantum numbers - as we have already explained, the $`𝐙_3`$ invariant states in the 59 sector do not possess well defined gauge quantum numbers. The reason for this is that the $`𝐙_3`$ twist does not commute with the already non-commuting Wilson lines. The situation in the $`b=2`$ $`𝐙_6`$ model is analogous to that we have just described for the $`b=4`$ model, except that there are additional subtleties due to the fact that before the $`𝐙_3`$ projection we have half-hypermultiplets in the corresponding $`𝐙_2`$ model.. Thus, we conclude that the $`𝐙_6`$ models with $`B`$-flux cannot be made completely consistent within this framework<sup>¶††</sup><sup>¶††</sup>¶††The above massless spectra (with the guessed 95 matter content) are free of the irreducible anomalies, so that there might exist consistent string constructions giving rise to these spectra. However, the $`\mathrm{\Omega }J^{}`$ orientifolds of Type IIB on $`T^4/𝐙_6`$ with $`B`$-flux do not seem to be the corresponding constructions..
### D The $`𝐙_4`$ Models
In this section we will discuss the $`\mathrm{\Omega }J^{}`$ orientifolds of Type IIB on $`T^4/𝐙_4`$ (for simplicity we will assume that $`T^4=T^2T^2`$) with $`b=2,4`$. Let $`g`$ be the generator of $`𝐙_4`$. Its action on the complex coordinates $`z_1,z_2`$ parametrizing the two 2-tori is given by: $`gz_1=iz_1,gz_2=iz_2`$. Note that $`g^2R`$ is the generator of the $`𝐙_2`$ subgroup of $`𝐙_4`$ ($`Rz_{1,2}=z_{1,2}`$). The action of $`J^{}`$ on the closed string untwisted as well as $`R`$ twisted sector fields is trivial, while $`J^{}`$ interchanges the $`g`$ twisted sector with the $`g^1`$ twisted sector.
Lets us first discuss the closed string sector of the above orientifold. The untwisted sector gives rise to the six dimensional $`𝒩=1`$ supergravity multiplet plus one tensor multiplet and 2 hypermultiplets. The $`g`$ and $`g^1`$ twisted sectors together give rise to 4 tensor multiplets and 4 hypermultiplets. As to the $`R`$ twisted sector, as in the $`𝐙_6`$ case, we must consider $`b=0,2,4`$ cases separately.
In the $`b=0`$ case we have 16 fixed points in the $`R`$ twisted sector before the $`𝐙_4`$ projection. At all of these fixed points we have O5<sup>-</sup>-planes, so each of them gives rise to a hypermultiplet (but no tensor multiplets). Let $`e_a`$ and $`d_a`$ be the vielbeins corresponding to the two 2-tori. Then the 4 fixed points at $`(0,0),(e_3/2,0),(0,d_3/4),(e_3/2,d_3/2)`$ are invariant under the $`𝐙_4`$ twist $`g`$. The other 12 fixed points fall into 6 distinct groups, each group containing 2 fixed points which are permuted by the action of $`g`$ (note that $`g`$ acts as a $`𝐙_2`$ twist on these fixed points which follows from the fact that by definition $`g^2=R`$ must be 1 on all points fixed under $`R`$). In each of these 6 groups we can form one linear combination of the corresponding 2 fixed points which is invariant under $`g`$. Thus, we have total of 10 hypermultiplets coming from the $`R`$ twisted sector - 4 from the points fixed under both $`R`$ and $`g`$, and 6 from the points fixed under $`R`$ but not $`g`$.
Next, let us consider the $`b=2`$ case. Here we have 12 O5<sup>-</sup>-planes and 4 O5<sup>+</sup>-planes. For definiteness let us assume that the $`B`$-flux is turned on on the first $`T^2`$. Then the $`𝐙_4`$ symmetry requires one of the following two distributions of the O5-planes. We can have 4 O5<sup>+</sup>-planes at the fixed points $`(0,0),(0,d_a/2)`$, or we can have 4 O5<sup>+</sup>-planes at the fixed points $`(e_3/2,0),(e_3/2,d_a/2)`$. Both setups give equivalent models, so we will focus on the latter setup for definiteness. Note that the fixed points $`(e_3/2,0)`$ and $`(e_3/2,d_3/2)`$ are also fixed under $`g`$, so that these fixed points give rise to one tensor multiplet each. On the other hand, the two fixed points $`(e_3/2,d_1/2)`$ and $`(e_3/2,d_2/2)`$ are permuted by the action of $`g`$, so that they together give rise to one tensor multiplet. Finally, it is not difficult to see that the rest of the fixed points (at which we have O5<sup>-</sup>-planes) give rise to 7 hypermultiplets. Thus, in the $`b=2`$ case we have 3 tensor multiplets and 7 hypermultiplets coming from the $`R`$ twisted sector<sup>¶‡‡</sup><sup>¶‡‡</sup>¶‡‡This corrects the error in , where the $`R`$ twisted sector in the $`b=2`$ $`𝐙_4`$ model was thought to give rise to 2 tensor multiplets and 8 hypermultiplets..
In the $`b=4`$ case we have 10 O5<sup>-</sup>-planes and 6 O5<sup>+</sup>-planes. One of the four equivalent distributions of O5-planes consistent with the $`𝐙_4`$ symmetry is the following. At the fixed points $`(0,0)`$ and $`(e_a/2,d_b/2)`$ we have O5<sup>-</sup>-planes, while at the fixed points $`(0,d_a/2)`$ and $`(e_a/2,0)`$ we have O5<sup>+</sup>-planes. After the $`g`$ projection, the latter give rise to 4 tensor multiplets, while the former give rise to 6 hypermultiplets. Thus, in the $`b=4`$ case we have 4 tensor multiplets and 6 hypermultiplets in the $`R`$ twisted sector<sup>∥\*</sup><sup>∥\*</sup>∥\*This corrects the error in , where the $`R`$ twisted sector in the $`b=4`$ $`𝐙_4`$ model was thought to give rise to 3 tensor multiplets and 7 hypermultiplets..
Next, let us discuss the open string sector. First let us focus on the $`b=2`$ model. Note that we have the $`B`$-flux in the direction of the first $`T^2`$, so that we have the corresponding non-commuting Wilson lines in the directions $`e_1`$ and $`e_2`$. Note that the action of $`g`$ on the corresponding shifts $`S_a`$ is given by (here $`S_3S_1S_2`$):
$$gS_1g^1=S_2,gS_2g^1=S_1^1=S_1,gS_3g^1=S_1^1S_2=S_3,$$
(189)
where we have used the fact that the shifts $`S_i`$ are $`𝐙_2`$ valued, that is, $`S_i^2=1`$. This implies that similar relations must also hold for the corresponding Chan-Paton matrices $`\gamma _{S_a,9}`$ and $`\gamma _{g,9}`$. There are, however, immaterial sign ambiguities in these relations such as whether to require $`\gamma _{g,9}\gamma _{S_3,9}\gamma _{g,9}^1=+\gamma _{S_3,9}`$ or $`\gamma _{S_3,9}`$. This ambiguity is related to the fact that even though the $`S_a`$ shifts are $`𝐙_2`$ valued, the corresponding $`\gamma _{S_a}`$ matrices can be $`𝐙_4`$ valued in the sense that $`\gamma _{S_a}^2`$ need not be the identity matrix $`I_{32}`$ but can also be equal $`I_{32}`$ (in the latter case $`\gamma _{S_a}=\gamma _{S_a}^1`$). In the following it will be convenient to use the following choices for these signs:
$$\gamma _{g,9}\gamma _{S_1,9}\gamma _{g,9}^1=\gamma _{S_2,9},\gamma _{g,9}\gamma _{S_2,9}\gamma _{g,9}^1=\gamma _{S_1,9},\gamma _{g,9}\gamma _{S_3,9}\gamma _{g,9}^1=\gamma _{S_3,9}.$$
(190)
A solution to these conditions can be written in terms of 16 copies of the corresponding 2 dimensional representations:
$`\gamma _{S_1,9}=\kappa \sigma _3I_{16},`$ (191)
$`\gamma _{S_2,9}=\kappa \sigma _1I_{16},`$ (192)
$`\gamma _{S_3,9}=i\sigma _2I_{16},`$ (193)
$`\gamma _{g,9}=\zeta _g\mathrm{\Gamma }_g,`$ (194)
where $`\kappa ^2=1`$ corresponds to the $`D_4^{}`$ type of Wilson lines, while $`\kappa ^2=1`$ corresponds to the $`D_4`$ type of Wilson lines. Note that both of these choices are a priori allowed as (190) implies that
$$\gamma _{S_1,9}^2=\gamma _{S_2,9}^2,$$
(195)
but does not relate $`\gamma _{S_3,9}^2`$ to $`\gamma _{S_i,9}^2`$.
Note that in the above solution for $`\gamma _{S_a,9}`$ and $`\gamma _{g,9}`$ we have assumed (for definiteness) that $`\eta _{12}=\kappa ^2`$, albeit this can be relaxed. The $`2\times 2`$ matrix $`\zeta _g`$, whose eigenvalues are $`1,1`$, is given by:
$$\zeta _g\frac{1}{\sqrt{2}}\left(\sigma _3+\sigma _1\right).$$
(196)
Finally, the matrix $`\mathrm{\Gamma }_g`$ is determined as follows. First, the twisted tadpole cancellation conditions imply that $`\gamma _{g,9}`$ must be traceless . It then follows that $`\mathrm{\Gamma }_g`$ must be traceless as well. Next, we have
$$\gamma _{R,9}=\gamma _{g,9}^2=I_2\mathrm{\Gamma }_g^2.$$
(197)
Note that for $`\kappa ^2=+1`$ (that is, for the $`D_4^{}`$ type of Wilson lines) we must have $`\gamma _{R,9}`$ (which is also traceless) with eigenvalues $`\pm i`$, while for $`\kappa ^2=1`$ (that is, for the $`D_4`$ type of Wilson lines) we must have $`\gamma _{R,9}`$ with eigenvalues $`\pm 1`$. This fixes $`\mathrm{\Gamma }_g`$ (up to equivalent representations) as follows:
$`\kappa ^2=+1:\mathrm{\Gamma }_g=\mathrm{diag}(\alpha ,\alpha ^1,\alpha ,\alpha ^1)I_4,`$ (198)
$`\kappa ^2=1:\mathrm{\Gamma }_g=\mathrm{diag}(1,1,i,i)I_4,`$ (199)
where $`\alpha \mathrm{exp}(2\pi i/8)`$.
The above discussion can be generalized to the $`b=4`$ case as well. Here we have two sets of Wilson lines $`\gamma _{S_a,9}`$ and $`\gamma _{T_a,9}`$ corresponding to the first and the second $`T^2`$, respectively. A solution for the Wilson lines and the twisted Chan-Paton matrix $`\gamma _{g,9}`$ satisfying all the required consistency conditions is given by:
$`\gamma _{S_1,9}=\kappa \sigma _3I_2I_8,`$ (200)
$`\gamma _{S_2,9}=\kappa \sigma _1I_2I_8,`$ (201)
$`\gamma _{S_3,9}=i\sigma _2I_2I_8,`$ (202)
$`\gamma _{T_1,9}=I_2\sigma _3I_8,`$ (203)
$`\gamma _{T_2,9}=I_2\sigma _1I_8,`$ (204)
$`\gamma _{T_3,9}=I_2i\sigma _2I_8,`$ (205)
$`\gamma _{g,9}=\zeta _g\zeta _g\stackrel{~}{\mathrm{\Gamma }}_g,`$ (206)
where the $`8\times 8`$ matrix $`\stackrel{~}{\mathrm{\Gamma }}_g`$ is given by
$`\kappa ^2=+1:\stackrel{~}{\mathrm{\Gamma }}_g=\mathrm{diag}(\alpha ,\alpha ^1,\alpha ,\alpha ^1)I_2,`$ (207)
$`\kappa ^2=1:\stackrel{~}{\mathrm{\Gamma }}_g=\mathrm{diag}(1,1,i,i)I_2.`$ (208)
Here we note that in the $`b=4`$ case for $`\kappa ^2=+1`$ the 99 gauge group before the $`\gamma _{g,9}`$ projection is $`SO(8)(𝐙_2)^2`$, while for $`\kappa ^2=1`$ it is $`Sp(8)(𝐙_2)^2`$. In the $`b=2`$ case with the aforementioned choice of Wilson lines the 99 gauge group before the $`\gamma _{g,9}`$ projection is $`SO(16)𝐙_2`$ for $`\kappa ^2=+1`$, while for $`\kappa ^2=1`$ it is $`Sp(16)𝐙_2`$.
Note that the twisted Chan-Paton matrix $`\gamma _{g,5}`$ is fixed (up to equivalent representations) once $`\gamma _{g,9}`$ is fixed. In fact, without loss of generality we can choose it to be given by<sup>∥†</sup><sup>∥†</sup>∥†Here we are assuming that all D5-branes are placed at an O5-plane located at a $`𝐙_4`$ fixed point. As we will show in a moment, other a priori allowed configurations would be continuously connected to these ones had the $`𝐙_4`$ models with $`B`$-flux been consistent, so this assumption is not restrictive.
$`b=2:\gamma _{g,5}=\mathrm{\Gamma }_g,`$ (209)
$`b=4:\gamma _{g,5}=\stackrel{~}{\mathrm{\Gamma }}_g.`$ (210)
This implies that before the $`\gamma _{g,5}`$ projection (assuming that we place all D5-branes at the same O5-plane) the 55 gauge group is $`SO(32/2^{b/2})`$ for $`\kappa ^2=+1`$, and $`Sp(32/2^{b/2})`$ for $`\kappa ^2=1`$ ($`b=2,4`$). That is, in the former case we must place D5-branes at an O5<sup>-</sup>-plane, while in the latter case we must place D5-branes at an O5<sup>+</sup>-plane. This is in complete parallel with our discussion of the corresponding $`𝐙_2`$ models.
In fact, here we run into the same problem as in the $`𝐙_2`$ case if we attempt to interpret the above orientifold in the context of the conformal field theory orbifold. More precisely, there are two separate issues here. As we will point out in a moment, in the $`𝐙_4`$ models with $`B`$-flux we appear to have a problem with the 59 sector states analogous to that in the $`𝐙_6`$ models. That is, the 59 vertex operators are not well defined. However, the issue that we would like to discuss first depends only on the structure of the 55 (and 99) sector states, whose vertex operators are well defined. So for a moment we will (erroneously) assume that the $`𝐙_4`$ orbifold action yields consistent 59 vertex operators corresponding to a discrete gauge group which is $`(𝐙_2)^{(b/2)}`$ or a subgroup thereof. With this assumption we can derive the spectra of the corresponding models.
First, let us start with the $`b=2`$ model with $`\kappa ^2=+1`$. The gauge group is $`[U(4)U(4)𝐃]_{99}[U(4)U(4)]_{55}`$, and the massless hypermultiplets are given by:
$`(\mathrm{𝟔},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏})_{99},(\mathrm{𝟏},\mathrm{𝟔};\mathrm{𝟏},\mathrm{𝟏})_{99},(\mathrm{𝟒},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (211)
$`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟔},\mathrm{𝟏})_{55},(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟔})_{55},(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟒},\mathrm{𝟒})_{55},`$ (212)
$`(\mathrm{𝟒}_D,\mathrm{𝟏};\mathrm{𝟒},\mathrm{𝟏})_{95},(\mathrm{𝟏},\mathrm{𝟒}_D;\mathrm{𝟏},\mathrm{𝟒})_{95},`$ (213)
where the subscript $`D`$ in the 95 sector indicates the 99 $`𝐃`$ discrete gauge quantum numbers. Here we encounter the first signs of trouble. In particular, note that for no choice of the discrete gauge symmetry $`𝐃`$ can we cancel, say, the $`R^4`$ irreducible anomaly (recall that the closed string sector contains 13 hypermultiplets and 8 tensor multiplets in this model).
Next, consider the $`b=4`$ model with $`\kappa ^2=+1`$. The gauge group is $`[U(2)U(2)𝐃]_{99}[U(2)U(2)]_{55}`$, and the massless hypermultiplets are given by:
$`(\mathrm{𝟏}_a,\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏})_{99},(\mathrm{𝟏},\mathrm{𝟏}_a;\mathrm{𝟏},\mathrm{𝟏})_{99},(\mathrm{𝟐},\mathrm{𝟐};\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (214)
$`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏}_a,\mathrm{𝟏})_{55},(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏}_a)_{55},(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟐},\mathrm{𝟐})_{55},`$ (215)
$`(\mathrm{𝟐}_D,\mathrm{𝟏};\mathrm{𝟐},\mathrm{𝟏})_{95},(\mathrm{𝟏},\mathrm{𝟐}_D;\mathrm{𝟏},\mathrm{𝟐})_{95},`$ (216)
where $`\mathrm{𝟏}_a`$ is the antisymmetric representation of $`U(2)`$ (it is a singlet of $`SU(2)`$ but is charged under the $`U(1)`$ factor). Note that this spectrum, just as in the $`b=2`$ case, cannot be made free of, say, the $`R^4`$ irreducible anomaly (the closed string sector contains 12 hypermultiplets and 9 tensor multiplets in this model).
The gauge group of the $`\kappa ^2=1`$ model with rank $`b`$ $`B`$-flux is $`[U(K)Sp(K)Sp(K)𝐃]_{99}[U(K)Sp(K)Sp(K)]_{55}`$, where $`K8/2^{b/2}`$ ($`b=2,4`$). The massless hypermultiplets are given by:
$`(𝐊,𝐊,\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{99},(𝐊,\mathrm{𝟏},𝐊;\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{99},`$ (217)
$`(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};𝐊,𝐊,\mathrm{𝟏})_{55},(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};𝐊,\mathrm{𝟏},𝐊)_{55},`$ (218)
$`(𝐊_D,\mathrm{𝟏},\mathrm{𝟏};𝐊,\mathrm{𝟏},\mathrm{𝟏})_{95},{\displaystyle \frac{1}{2}}(\mathrm{𝟏},𝐊_D,\mathrm{𝟏};\mathrm{𝟏},𝐊,\mathrm{𝟏})_{95},{\displaystyle \frac{1}{2}}(\mathrm{𝟏},\mathrm{𝟏},𝐊_D;\mathrm{𝟏},\mathrm{𝟏},𝐊)_{95}.`$ (219)
Just as for the $`\kappa ^2=+1`$ models, these spectra cannot be made free of the $`R^4`$ anomaly (the closed string matter contents in the $`\kappa ^2=1`$ models are the same as in the corresponding $`\kappa ^2=+1`$ models). Moreover, the 95 sector spectrum is actually anomalous as it contains half-hypermultiplets in real representations<sup>∥‡</sup><sup>∥‡</sup>∥‡Here we note that the above spectrum differs in the 95 sector from that given for these models in the second reference in . However, neither spectra are completely consistent due to the problem with the 95 vertex operators we are going to point out in a moment. In fact, this problem is at least partially responsible for an ambiguity in determining the 95 sector spectrum in these models, which had led to that reported in the second reference in .. This is similar to what happens in the corresponding $`𝐙_2`$ models.
In fact, the main reason why we have given the above spectra for both $`\kappa ^2=\pm 1`$ models is that, if we assume that they are realized as orientifolds of the conformal field theory $`T^4/𝐙_4`$ orbifold, then they are continuously connected. This is in complete parallel with the corresponding discussion in the $`𝐙_2`$ models. Thus, for instance, in the conformal field theory orbifold setup there is no obstruction to moving D5-branes in, say, the $`b=2`$ model with $`\kappa ^2=+1`$ from an O5<sup>-</sup>-plane to an O5<sup>+</sup>-plane. This then leads to the inconsistency in the 59 sector spectrum where half-hypermultiplets arise in various inappropriate representations. To see that the Higgsing corresponding to such motion of D5-branes is allowed, note that in this model (that is, the $`b=2`$ model with $`\kappa ^2=+1`$) we can simultaneously turn on VEVs of the hypermultiplets $`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟔},\mathrm{𝟏})_{55}`$, $`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟔})_{55}`$ and $`(\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟒},\mathrm{𝟒})_{55}`$ in such a way that the D-flatness conditions are satisfied. The maximal unbroken 55 gauge subgroup then is $`Sp(4)`$. This corresponds to moving together 2 dynamical 5-branes off the O5<sup>-</sup>-plane. Note that each dynamical 5-brane is made of 8 D5-branes - one pairing is due to the orientifold projection, while another grouping in multiples of 4 is due to the $`𝐙_4`$ orbifold projection. In this subspace of the moduli space, which corresponds to these two dynamical 5-branes sitting on top of each other in the bulk, we have one 55 hypermultiplet in $`\mathrm{𝟏}\mathrm{𝟓}`$ of $`Sp(4)_{55}`$. In fact, the VEV of the singlet measures the separation between the D5-branes and the O5<sup>-</sup>-plane. We can split the two dynamical 5-branes from each other by giving a VEV to the hypermultiplet in $`\mathrm{𝟓}`$ of $`Sp(4)`$, which breaks the gauge group down to $`Sp(2)Sp(2)`$. Thus, each dynamical 5-brane, as expected, gives rise to one $`Sp(2)`$ subgroup, and the leftover two singlet hypermultiplets measure the individual locations of the two dynamical 5-branes in the bulk. Note that once all 16 D5-branes approach an O5<sup>+</sup>-plane, the 55 gauge group is enhanced to $`[U(4)Sp(4)Sp(4)]_{55}`$. In fact, the matter content of the 55 sector in the $`b=2`$ model with $`\kappa ^2=1`$ is just right to Higgs the latter gauge group down to $`Sp(4)_{55}`$ plus the hypermultiplet in $`\mathrm{𝟏}\mathrm{𝟓}`$ (or its $`Sp(2)Sp(2)`$ subgroup with two singlet hypermultiplets). This then implies that we will have inconsistencies arising in the 59 sector once we, say, move D5-branes from an O5<sup>-</sup>-plane to an O5<sup>+</sup>-plane in the $`b=2`$ model with $`\kappa ^2=+1`$. In fact, these inconsistencies are of the same type as those we have encountered in the corresponding $`𝐙_2`$ models in subsection A.
The above discussion, just as in the corresponding $`𝐙_2`$ cases, leads us to the conclusion, which is consistent with our discussions in subsection A, that we cannot consistently have O5<sup>+</sup>-planes in the context of the conformal field theory $`T^4/𝐙_4`$ orbifold. This is not surprising as $`𝐙_4`$ contains a $`𝐙_2`$ subgroup, and we have already come to such a conclusion in the $`T^4/𝐙_2`$ case. In the latter models we also pointed out a possibility of deforming the orbifold by partially blowing-up the collapsed $`𝐏^1`$’s and turning off the twisted $`B`$-flux at the $`𝐙_2`$ fixed points where we expect the O5<sup>+</sup>-planes. The resulting K3 compactification could then a priori be the correct framework for considering the corresponding orientifolds with $`B`$-flux, at least, in the $`𝐙_2`$ case we did not find any obvious inconsistencies with such a possibility. So here we can ask whether a similar possibility exists in the $`𝐙_4`$ cases as well. The answer to this question appears to be negative for at least two reasons. First, the massless spectra of the $`\kappa ^2=+1`$ (as well as $`\kappa ^2=1`$) models cannot be made anomaly free. We would like to discuss the second reason next.
The point is that, just as in the $`𝐙_6`$ models with $`B`$-flux (and for essentially the same reason), the 99 $`(𝐙_2)^{(b/2)}`$ discrete gauge symmetry is broken by the $`𝐙_4`$ orbifold action. This, in turn, implies that the above spectra, in particular, in the 59 sector, are not completely correct. More concretely, in the 59 sector the states invariant under the $`\gamma _{g,9}`$ (together with $`\gamma _{g,5}`$) $`𝐙_4`$ orbifold projection do not carry well defined gauge quantum numbers at all. Just as in the $`𝐙_6`$ cases, this is due to the fact that $`\gamma _{g,9}`$ does not commute with already non-commuting Wilson lines $`\gamma _{S_i,9}`$.
Let us illustrate this point for the $`b=2`$ model with $`\kappa ^2=+1`$ (all other cases can be treated similarly). Note that in this case $`\gamma _{S_a,9}=\gamma _aI_{16}`$ (where $`\gamma _a`$ are the corresponding $`2\times 2`$ matrices), and $`\gamma _{g,9}=\zeta _g\mathrm{\Gamma }_g`$. The non-commutativity between $`\gamma _{g,9}`$ and $`\gamma _{S_a,9}`$ is in the $`2\times 2`$ block corresponding to $`\gamma _a`$ and $`\zeta _a`$. So, we can ignore the other $`16\times 16`$ block structure for our purposes here. Moreover, just as in the $`𝐙_6`$ cases, the important point is how $`\mathrm{𝟑𝟐}`$ of $`SO(32)`$ (that is, the 99 gauge group before the $`𝐙_4`$ orbifold as well as $`𝐙_2𝐙_2`$ freely acting orbifold projections) is decomposed under the unbroken 99 gauge group - the 55 part of the corresponding vertex operators cannot possibly give any trouble here. So, for simplicity we will discuss the following Chan-Paton matrices: $`\gamma _{S_a,9}=\gamma _aI_{16}`$, $`\gamma _{g,9}=\zeta _gI_{16}`$. (This choice does not satisfy the twisted tadpole cancellation conditions, but this is not going to be relevant here.) Moreover, for definiteness we will choose $`\kappa =1`$, so that $`\gamma _1=\sigma _3`$, $`\gamma _2=\sigma _1`$, $`\gamma _3=i\sigma _2`$.
The unbroken gauge group with the aforementioned choice of Chan-Paton matrices is $`SO(16)`$. This gauge group arises as follows. First, $`\gamma _{S_1,9}`$ breaks $`SO(32)`$ down to $`SO(16)SO(16)`$. Then $`\gamma _{S_2,9}`$ breaks $`SO(16)SO(16)`$ down to $`SO(16)_{\mathrm{diag}}𝐙_2`$. Finally, $`\gamma _{g,9}`$ breaks the $`𝐙_2`$ subgroup, but does not affect the $`SO(16)_{\mathrm{diag}}`$ subgroup. To understand this, let us see how $`\mathrm{𝟑𝟐}`$ of $`SO(32)`$ decomposes under these breakings. First, under $`\gamma _{S_1,9}`$ we have
$$\mathrm{𝟑𝟐}=(\mathrm{𝟏𝟔},\mathrm{𝟏})(\mathrm{𝟏},\mathrm{𝟏𝟔}).$$
(220)
Next, under $`\gamma _{S_2,9}`$ the two $`SO(16)`$ subgroups are permuted, so that we have
$$\mathrm{𝟑𝟐}=\mathrm{𝟏𝟔}_+\mathrm{𝟏𝟔}_{},$$
(221)
where the subscript indicates the 99 $`𝐙_2`$ discrete gauge charges. The latter can be quantified as follows. Note that the eigenvectors of the matrix $`\gamma _1=\sigma _3`$ are 2-component column-vectors $`\psi _\pm `$ with the components $`\psi _+^{}=1`$, $`\psi _+^{}=0`$, $`\psi _{}^{}=0`$, $`\psi _{}^{}=1`$. Note that the vertex operators in (220) for $`(\mathrm{𝟏𝟔},\mathrm{𝟏})`$ and $`(\mathrm{𝟏},\mathrm{𝟏𝟔})`$ are proportional to $`\psi _+`$ and $`\psi _{}`$, respectively. However, the eigenvectors of the matrix $`\gamma _2=\sigma _1`$ are
$$\chi _\pm =\frac{1}{\sqrt{2}}\left(\psi _+\pm \psi _{}\right).$$
(222)
In fact, the vertex operators for $`\mathrm{𝟏𝟔}_+`$ and $`\mathrm{𝟏𝟔}_{}`$ in (221) are proportional to $`\chi _+`$ and $`\chi _{}`$, respectively. Thus, $`\chi _\pm `$ provide precisely the part of the vertex operators corresponding to the 99 $`𝐙_2`$ discrete gauge quantum numbers. Note, however, that $`\chi _\pm `$ are not eigenvectors of $`\zeta _g`$ as $`\zeta _g`$ and $`\gamma _2`$ do not commute. In particular, we have
$$\zeta _g\chi _\pm =\psi _\pm ,$$
(223)
so that the action of $`\gamma _{g,9}`$ on the 59 sector states mixes states with different gauge quantum numbers. That is, the 59 states with distinct gauge quantum numbers under the unbroken $`SO(16)`$ subgroup are incompatible with the $`𝐙_4`$ orbifold projection. On the other hand, the 59 states invariant under the $`𝐙_4`$ orbifold projection do not possess well defined gauge quantum numbers. We therefore conclude that we have an inconsistency at the level of the 59 vertex operators in this and other $`𝐙_4`$ models with $`B`$-flux.
### E Summary
Let us briefly summarize the results of this section. In subsection A we have argued that the $`\mathrm{\Omega }`$ orientifold of Type IIB on the conformal field theory $`T^4/𝐙_2`$ orbifold with $`B`$-flux is inconsistent. This can be seen by noting that in the toroidal orbifold with $`B`$-flux we would have to introduce O5<sup>+</sup>-planes, which cannot be placed at the conformal field theory $`T^4/𝐙_2`$ orbifold fixed points. The obstruction here is due to the twisted half-integer $`B`$-flux (which makes the conformal field theory orbifold non-singular) stuck in the corresponding collapsed $`𝐏^1`$’s. We have also pointed out that perhaps by deforming the orbifold from the conformal field theory point (such a deformation would involve blowing up the fixed points where we expect the O5<sup>+</sup>-planes and turning off the corresponding twisted $`B`$-flux) we could make the corresponding orientifold background consistent. Such a K3 compactification, however, would not correspond to an exactly solvable conformal field theory, and it is unclear at present how to make the corresponding statements quantitatively precise. In particular, the moduli space corresponding to the motion of D5-branes as well as that corresponding to turning on continuous Wilson lines in the 99 sector would have to have some funny properties which we do not know how to argue for (or, for that sake, against). Thus, for instance, D5-branes should not be able to come on top of the O5<sup>+</sup>-planes - this could in principle be a property of compactifications on such a K3 as the corresponding moduli space is no longer flat, but it is unclear whether it really is. If, however, this is indeed the case, then the $`𝐙_2`$ models with $`B`$-flux could be consistent in the context of such K3 compactifications if we place D5-branes on top of an O5<sup>-</sup>-plane or in the bulk with the gauge bundle corresponding to the $`𝐙_2`$ orbifold twist “without vector structure” in the language of . However, attempts to construct models with $`𝐙_2`$ gauge bundles “with vector structure” run into various inconsistencies. In fact, these inconsistencies can be seen in the language of the effective field theory where they manifest themselves via anomalies. In particular, the fact that the conformal field theory $`T^4/𝐙_2`$ orbifold cannot be the correct background for the corresponding orientifold with $`B`$-flux can also be seen in the effective field theory language. Thus, for instance, in the conformal field theory orbifold case there is no obstruction to moving D5-branes from an O5<sup>-</sup>-plane to an O5<sup>+</sup>-plane, and at the latter points in the moduli space the massless spectra in the 59 sector are anomalous.
In fact, it is always the 59 sector where the trouble shows up. This is, in turn, not an accident. Thus, there is no reason for any inconsistencies to arise in the 99 or 55 sectors (at least perturbatively). However, the 59 sector by definition (since it contains hypermultiplets in the bifundamental representations of the 99 and 55 gauge groups) requires certain vector structure contrary to the lack of vector structure required by the fact that once we turn on the $`B`$-flux the generalized second Stieffel-Whitney class is non-vanishing. This appears to be the key reason for the aforementioned inconsistencies arising in these models.
Another (perhaps indirect) hint of this is what we have found in the $`𝐙_6`$ and $`𝐙_4`$ models with $`B`$-flux, namely, that the 59 sector states, if properly projected by the orbifold action, do not carry well defined gauge quantum numbers. Even though in the previous subsections we have only used examples to illustrate this point, the corresponding underlying reason can be stated quite generally.
Thus, consider a general setup with Type IIB compactified on $`T^d`$, where for definiteness we will choose $`d`$ to be even. Let us introduce some number $`N`$ of D9-branes wrapping $`T^d`$. Note that to cancel tadpoles we would have to introduce the O9<sup>-</sup>-plane and set $`N=32`$. However, we will not do so here, that is, we will not introduce any orientifold planes at all, and we will let $`N`$ be arbitrary - the point we would like to make here is independent of the tadpole cancellation requirements. Next, let us introduce some number $`N^{}`$ of D$`p`$-branes transverse to $`T^d`$. Even though this is not particularly important for our discussion here, we will assume that $`9p`$ is a multiple of 4 so that supersymmetry is not completely broken in this background. Note that in the $`9p`$ open string sector we have no momenta or windings (but only oscillator excitations) in the directions of $`T^d`$.
Next, we would like to turn on non-zero $`B`$-flux on $`T^d`$. We can do this by considering a freely acting orbifold whose generators $`S_i`$, $`i=1,\mathrm{},d`$, are shifts in the corresponding directions of $`T^d`$ (and, therefore, they commute). Note that here we can have discrete torsion $`\mathrm{\Delta }_{ij}`$ between different $`S_i`$ generators. In particular, we must have $`\mathrm{\Delta }_{ii}=+1`$, but $`\mathrm{\Delta }_{ij}`$ can be equal $`1`$ for $`ij`$ as long as both $`S_i`$ and $`S_j`$ have even orders. The matrix $`B_{ij}`$ corresponding to the $`B`$-field can then be compactly written as
$$B_{ij}=\frac{1}{2\pi i}\mathrm{ln}\left(\mathrm{\Delta }_{ij}\right).$$
(224)
Note that $`B_{ij}B_{ij}+1`$.
The action of the shifts $`S_i`$ on the 99 sector Chan-Paton charges is described by $`N\times N`$ matrices $`\gamma _{S_i,9}`$, and corresponds to turning on Wilson lines. The string consistency requires that
$$\gamma _{S_i,9}\gamma _{S_j,9}=\mathrm{\Delta }_{ij}\gamma _{S_j,9}\gamma _{S_i,9},$$
(225)
so that if we have non-trivial discrete torsion, then the corresponding Wilson lines are non-commuting. Here we note that the action of the shifts on the $`pp`$ open string sector is trivial. In the 99 sector, however, it breaks the gauge group $`U(N)`$ to its subgroup $`G`$. Suppose now that we have non-trivial discrete torsion between some generators $`S_i`$. That is, let the rank $`b`$ of the matrix $`B_{ij}`$ be non-zero. Then the rank of the unbroken gauge group $`r(G)<N`$. In fact, $`r(G)=N/2^{b/2}`$. The coset $`U(N)/G`$ is given by the discrete gauge group $`D(𝐙_2)^{(b/2)}`$. Let $`d_\alpha `$ denote the corresponding discrete charges, $`\alpha =1,\mathrm{},|D|`$ ($`|D|=2^{b/2}`$). Then it is not difficult to see that the fundamental $`𝐍`$ of $`U(N)`$ decomposes as follows:
$$𝐍=\underset{\mathrm{},\alpha }{}(𝐍_{\mathrm{}},d_\alpha ),$$
(226)
where $`𝐍_{\mathrm{}}`$ are the fundamental representations of the corresponding subgroups of $`G=_{\mathrm{}}U(N_{\mathrm{}})`$. (Note that $`r(G)=_{\mathrm{}}N_{\mathrm{}}`$.) On the other hand, the adjoint of $`U(N)`$ decomposes as follows:
$$\mathrm{𝐀𝐝𝐣}=𝐍\overline{𝐍}=\underset{\mathrm{},\mathrm{}^{},\alpha ,\beta }{}(𝐍_{\mathrm{}}\overline{𝐍}_{\mathrm{}^{}},d_\alpha \overline{d}_\beta ).$$
(227)
Note that the gauge bosons of the unbroken gauge group $`G`$ come from the subset with $`\mathrm{}=\mathrm{}^{}`$ and $`\beta =\alpha `$. This subset reads:
$$\underset{\mathrm{},\alpha }{}(\mathrm{𝐀𝐝𝐣}_{\mathrm{}},d_\alpha \overline{d}_\alpha ).$$
(228)
In fact, this subset contains exactly $`|D|`$ copies of the adjoint of $`G`$ as $`\left|_\alpha d_\alpha \overline{d}_\alpha \right|=|D|`$. Note that in the 99 sector only one of these copies survives the full freely acting orbifold projection with respect to the action of $`\gamma _{S_i,9}`$. It is then not difficult to see that the remaining states are given by<sup>∥§</sup><sup>∥§</sup>∥§Thus, for instance, consider the case where $`b=d`$, and $`\mathrm{\Delta }_{12}=\mathrm{\Delta }_{34}=\mathrm{}=\mathrm{\Delta }_{b1,b}=1`$ with all other $`\mathrm{\Delta }_{ij}=+1`$. Then we can group the $`b`$ independent projections $`\gamma _{S_i,9}`$ as follows. First, $`b/2`$ projections $`\gamma _{S_1,9},\gamma _{S_3,9},\mathrm{},\gamma _{S_{b1},9}`$ break the original $`U(N)`$ gauge group down to its subgroup of rank $`N`$. Then, the rest of the projections $`\gamma _{S_2,9},\gamma _{S_4,9},\mathrm{},\gamma _{S_b,9}`$ break this gauge group to its subgroup $`GD`$. Note that the number of gauge bosons in (228) is $`|G||D|=2^{b/2}|G|`$. This corresponds to keeping the states with trivial discrete gauge charges (indeed, the representations $`_\alpha d_\alpha \overline{d}_\alpha `$ correspond to precisely such discrete gauge charges). These states are invariant under half of the $`\gamma _{S_i,9}`$ projections, say, the $`\gamma _{S_2,9},\gamma _{S_4,9},\mathrm{},\gamma _{S_b,9}`$ projections. On the hand, the number of states in (228) which are also invariant under the remaining projections $`\gamma _{S_1,9},\gamma _{S_3,9},\mathrm{},\gamma _{S_{b1},9}`$ is $`|G|`$. These states are given by (229).
$$\underset{\mathrm{}}{}(\mathrm{𝐀𝐝𝐣}_{\mathrm{}},D_{\mathrm{inv}}),$$
(229)
where the vertex operator corresponding to the discrete gauge charge $`D_{\mathrm{inv}}`$ is given by
$$|D_{\mathrm{inv}}=\frac{1}{\sqrt{D}}\underset{\alpha }{}|d_\alpha \overline{d}_\alpha .$$
(230)
In contrast, in the $`9p`$ sector no states are projected out by the action of this freely acting orbifold - it actually does not act in the $`9p`$ sector (neither does it act in the $`pp`$ sector). So instead the fundamental $`𝐍`$ of $`U(N)`$ in the $`9p`$ sector simply decomposes under the unbroken gauge group as in (226).
Now suppose we orbifold $`T^d`$ with the $`B`$-field turned on by the action of the orbifold point group $`\mathrm{\Gamma }`$ such that some of the elements of $`\mathrm{\Gamma }`$ do not commute with the shifts $`S_i`$. It is necessary for the consistency of the theory that the shifts $`S_i`$ and the twists in $`\mathrm{\Gamma }`$ form a larger orbifold group $`\stackrel{~}{\mathrm{\Gamma }}`$, which is non-Abelian. Let us consider one non-trivial element $`g`$ (such that $`g^21`$) that does not commute with some of the shifts $`S_i`$. Then this implies that the corresponding Chan-Paton matrices $`\gamma _{S_i,9}`$ and $`\gamma _{g,9}`$ also do not commute. It is, however, important to note that the action of the twisted Chan-Paton matrix $`\gamma _{g,9}`$ does not reduce the rank of the unbroken group $`G_gG`$, that is, $`r(G_g)=r(G)`$. This follows from the fact that $`g`$ simply permutes elements of $`\stackrel{~}{\mathrm{\Gamma }}`$ that are pure shifts among each other. In fact, this implies that we can always find $`\gamma _{g,9}`$ such that the unbroken gauge group $`G_g=G`$. In this case the action of $`\gamma _{g,9}`$ is non-trivial only on the discrete gauge quantum numbers<sup>∥¶</sup><sup>∥¶</sup>∥¶Our discussion here straightforwardly generalizes to the most general case. However, to illustrate the point we would like to make here, it suffices to consider $`\gamma _{g,9}`$ of the aforementioned type.. In fact, this action breaks the discrete gauge symmetry $`D`$ either completely or to its smaller subgroup. Thus, the action of $`\gamma _{g,9}`$ on the discrete quantum numbers $`d_\alpha `$ is given by:
$$\gamma _{g,9}:d_\alpha \underset{\alpha ^{}}{}c_{\alpha \alpha ^{}}d_\alpha ^{},$$
(231)
where the matrix $`c_{\alpha \alpha ^{}}`$ is not diagonal as $`\gamma _{g,9}`$ does not commute with the Wilson lines. This then implies that the $`9p`$ sector states invariant under the $`g`$ projection cannot have well defined gauge quantum numbers under the unbroken subgroup $`G`$. This can be seen by noting that the states in (229) (that correspond to the gauge bosons of the unbroken gauge group $`G`$) are invariant under the action of $`g`$ as<sup>∥∥</sup><sup>∥∥</sup>∥∥Here we are using the fact that $`_\alpha c_{\alpha \beta }\overline{c}_{\gamma \alpha }=\delta _{\beta \gamma }`$, which is the statement that $`\gamma _{g,9}\gamma _{g,9}^1=1`$, in particular, when acting on the discrete gauge charges $`d_\alpha `$ and $`\overline{d}_\alpha `$. Also, note that Chan-Paton matrices are unitary, so that the action of $`g`$ on $`\overline{𝐍}`$ of $`U(N)`$ is given by the matrix $`\gamma _{g,9}^1`$ \- see below.
$$\gamma _{g,9}:|D_{\mathrm{inv}}\frac{1}{\sqrt{D}}\underset{\alpha }{}\underset{\beta ,\gamma }{}c_{\alpha \beta }\overline{c}_{\gamma \alpha }|d_\beta \overline{d}_\gamma =\frac{1}{\sqrt{D}}\underset{\alpha }{}|d_\alpha \overline{d}_\alpha =|D_{\mathrm{inv}},$$
(232)
that is, the aforementioned states are actually invariant under the $`\gamma _{g,9}`$ action, so that they provide the correct basis for the gauge bosons. On the other hand, the states in the $`9p`$ sector that are invariant under the $`g`$ projection are in a basis different from that of the $`G`$ gauge bosons. This can be seen by noting that in the basis where $`\gamma _{g,9}`$ acts diagonally on $`𝐍`$ of $`U(N)`$ we have:
$$𝐍=\underset{\mathrm{},k}{}(𝐍_{\mathrm{}},k),$$
(233)
where the vertex operators corresponding to the quantum numbers $`k`$ are given by:
$$|k\underset{\alpha }{}f_{k\alpha }|d_\alpha .$$
(234)
Here $`f_{k\alpha }`$ are the eigenvectors of the matrix $`c_{\alpha \alpha ^{}}`$:
$$\underset{\alpha }{}f_{k\alpha }c_{\alpha \alpha ^{}}=\lambda _kf_{k\alpha ^{}},$$
(235)
where $`\lambda _k`$ are the eigenvalues of $`c_{\alpha \alpha ^{}}`$. It is important to note that for each $`k`$ $`f_{k\alpha }0`$ for at least two different values of $`\alpha `$, which follows from the fact that the twist $`g`$ and the Wilson lines do not commute. Next, consider a massless $`9p`$ sector state containing $`(𝐍_{\mathrm{}},k)`$. Its scattering with its own conjugate state then will contain $`\mathrm{𝐀𝐝𝐣}_{\mathrm{}}`$ together with the following vertex operator corresponding to the discrete gauge quantum numbers:
$$|k|\overline{k}=\underset{\alpha \alpha ^{}}{}f_{k\alpha }\overline{f}_{k\alpha ^{}}|d_\alpha \overline{d}_\alpha ^{}.$$
(236)
Thus, the aforementioned $`9p`$ sector states would scatter into states, in particular, gauge bosons, containing non-diagonal terms with $`|d_\alpha \overline{d}_\alpha ^{}`$ with $`\alpha \alpha ^{}`$ absent in (229). Thus, we see that the $`g`$ invariant states in the $`9p`$ sector indeed do not carry well defined gauge quantum numbers.
The reason why $`g`$ acts so differently in the 99 and $`9p`$ sectors is actually very simple. The Chan-Paton part of the $`9p`$ vertex operators is proportional to
$$V_{9p}\lambda _9\overline{\lambda }_p,$$
(237)
while in the 99 sector we have
$$V_{99}\lambda _9\overline{\lambda }_9,$$
(238)
where $`\lambda _9`$ and $`\overline{\lambda }_9`$ correspond to the fundamental and antifundamental of the 99 gauge group $`U(N)`$, while $`\lambda _p`$ and $`\overline{\lambda }_p`$ correspond to the fundamental and antifundamental of the $`pp`$ gauge group $`U(N^{})`$. Thus, the action of $`g`$ in the $`9p`$ sector is given by
$$g:\lambda _9\overline{\lambda }_p\gamma _{g,9}\lambda _9\overline{\lambda }_p\gamma _{g,p}^1,$$
(239)
while in the 99 sector we have
$$g:\lambda _9\overline{\lambda }_9\gamma _{g,9}\lambda _9\overline{\lambda }_9\gamma _{g,9}^1.$$
(240)
Thus, the action of $`\gamma _{g,9}`$ in the 99 sector is bilinear, while in the $`9p`$ sector it is linear, and this is precisely the reason why the two actions are incompatible in the presence of non-commuting Wilson lines (which affect the 99 quantum numbers only) as explained above.
The above discussion implies that we indeed have a conflict between the facts that gauge bundles of D9-branes wrapped on tori with $`B`$-flux lack vector structure, while the presence of $`9p`$ sectors with D$`p`$-branes transverse to such tori implies the presence of certain vector structure. In fact, the inconsistencies we have encountered in orientifolds of Type IIB on $`T^4/𝐙_M`$ orbifolds with $`B`$-flux simply imply that the corresponding choices of the gauge bundles are not consistent within this framework (albeit consistent choices could be found for the cases without the $`B`$-flux). The obstruction to having consistent gauge bundles, once again, is related to the lack of vector structure.
## IV Four Dimensional Orientifolds with $`B`$-flux
In this section we will discuss four dimensional orientifolds with $`B`$-flux. In particular, we will consider compactifications with $`𝒩=1`$ as well as $`𝒩=2`$ supersymmetry. More concretely, we will discuss orientifolds of Type IIB on $`T^6/\mathrm{\Gamma }`$ orbifolds. If $`\mathrm{\Gamma }`$ is a (non-trivial) subgroup of $`SU(2)`$, then we have $`𝒩=2`$ supersymmetry, and if $`\mathrm{\Gamma }`$ is a subgroup of $`SU(3)`$ but not of $`SU(2)`$, then we have $`𝒩=1`$ supersymmetry. In subsection A we will discuss $`𝒩=2`$ examples. In subsections B,C,D,E we will discuss $`𝒩=1`$ examples with the orbifold groups $`\mathrm{\Gamma }=𝐙_3,𝐙_7,𝐙_3𝐙_3,𝐙_6`$, respectively.
### A The $`𝒩=2`$ Models
In this subsection we will discuss orientifolds of Type IIB on $`T^2\mathrm{K3}`$, where $`\mathrm{K3}=T^4/𝐙_M`$ ($`M=2,3,4,6`$). A priori the $`B`$-flux can be turned on either on $`T^2`$ or K3 or both. However, we will focus on the cases with $`B`$-flux such that we will not encounter the difficulties analogous to those we have found in six dimensional orientifolds with $`B`$-flux. In particular, the latter will always be either transverse to or inside of the world-volumes of all D-branes present in a given model<sup>∥\**</sup><sup>∥\**</sup>∥\**If only one type of D-branes is present, which is the case in the $`𝐙_3`$ models, then we can have $`B`$-flux simultaneously turned on in the directions inside of as well as transverse to the D-brane world-volumes..
Let us discuss this point in a bit more detail. What we have found in the previous section is that if we simultaneously have D-branes wrapping tori (or, more precisely, orbifolds thereof) and D-branes transverse to such tori, then we run into various subtle inconsistencies. The latter would not, for instance, occur if all D-branes where transverse to $`B`$-flux. Similarly, such inconsistencies would also be absent if all D-branes are wrapping such tori. This is precisely the strategy we are going to employ here to construct consistent four dimensional $`𝒩=2`$ supersymmetric orientifolds with $`B`$-flux. Note that in the case of K3 orientifolds with both D9- and D5-branes we cannot have such a setup. But with three compact complex dimensions this now becomes possible to achieve.
To begin with, we can consider the following setup. Consider the $`\mathrm{\Omega }J^{}`$ orientifold<sup>∥††</sup><sup>∥††</sup>∥††The action of $`J^{}`$ was defined in the previous section. In particular, note that $`J^{}=1`$ in the $`𝐙_2`$ models. of Type IIB on $`T^2\mathrm{K3}`$, where $`\mathrm{K3}=T^4/𝐙_M`$ ($`M=2,3,4,6`$). For $`M=2,4,6`$, where we have both D9- and D5-branes, we will assume that the $`B`$-field on K3 is trivial, while we have non-zero $`B`$-flux on $`T^2`$. In the $`M=3`$ case a priori we can have $`B`$-flux on both $`T^2`$ as well as K3.
Let us first consider the models with $`M=2,4,6`$. The $`B`$-flux on $`T^2`$ can be described in terms of the freely acting $`𝐙_2𝐙_2`$ orbifold with discrete torsion. Note that the latter now commutes with the $`𝐙_M`$ orbifold action. In fact, the spectra of these models can be obtained by compactifying the corresponding K3 models of with trivial $`B`$-flux on $`T^2`$ with $`B`$-flux<sup>∥‡‡</sup><sup>∥‡‡</sup>∥‡‡Note that in the open string sector this is equivalent to compactifying a six dimensional gauge theory on a non-commutative $`T^2`$.. The closed string spectra are given by a straightforward dimensional reduction of the corresponding six dimensional spectra given in <sup>\***</sup><sup>\***</sup>\***Note, in particular, that the $`B`$-flux on $`T^2`$ does not affect the number of six dimensional tensor multiplets, which upon the dimensional reduction actually give rise to four dimensional vector multiplets.. As to the open string sector, its massless spectrum is obtained by performing the freely acting $`𝐙_2𝐙_2`$ orbifold projections. Note that now the Wilson lines act in both 99 as well as 55 sectors. The corresponding Chan-Paton matrices $`\gamma _{S_i,9}`$ and $`\gamma _{S_i,5}`$ must be the same (up to equivalent representations). This is necessary for the action of the $`𝐙_2𝐙_2`$ freely acting orbifold on the 59 sector states, which is given by the matrices $`\gamma _{S_i,9}\gamma _{S_i,5}^1`$, to be consistent. Moreover, the twisted Chan-Paton matrices $`\gamma _{R,9}`$ and $`\gamma _{R,5}`$ (which up to equivalent representations must be the same so that the $`𝐙_2𝐙_M`$ orbifold twist $`R`$ acts consistently on the 59 sector states) have eigenvalues $`\pm i`$ (and not $`\pm 1`$). This follows from the corresponding statement for the six dimensional $`𝐙_2`$ model with trivial $`B`$-flux. Note that depending upon whether the Wilson lines on $`T^2`$ are of the $`D_4`$ or $`D_4^{}`$ type we have different models, which, however, are connected as they belong to different points of the Coulomb branch of the same $`𝒩=2`$ gauge theory.
Note that the ranks of both the 99 and 55 gauge groups are equal<sup>\**†</sup><sup>\**†</sup>\**†Here we assume that the Wilson lines in both the 99 and 55 sectors correspond to the points of the respective moduli spaces which can be described by the freely acting $`𝐙_2𝐙_2`$ orbifold. 8. In fact, both gauge groups contain $`𝐙_2`$ discrete gauge subgroups, but there are no massless states carrying non-trivial $`[𝐙_2]_{99}`$ or $`[𝐙_2]_{55}`$ charges. This, in particular, applies to the $`59`$ sector states as well where the multiplicity of states now is $`\xi _{59}=1`$. This is because now the Wilson lines do act in the 59 sector. In particular, the 59 states carrying non-trivial $`[𝐙_2]_{99}`$ or $`[𝐙_2]_{55}`$ charges are now massive - they are at heavy Kaluza-Klein levels corresponding to the compactification on $`T^2`$. Here we note that in the limit of the large volume $`T^2`$ the effect of the $`B`$-flux is ameliorated. Thus, as was pointed out in , in this limit in all three 99, 55 and 59 sectors the Kaluza-Klein states with non-trivial $`[𝐙_2]_{99}`$ and/or $`[𝐙_2]_{55}`$ charges become massless (along with the Kaluza-Klein states with trivial discrete gauge charges) so that the freely acting orbifold action is ameliorated, and the massless six dimensional states arising in this limit are in the representations of the full rank $`16+16`$ 99 plus 55 gauge symmetry. That is, in this limit we recover the six dimensional K3 orientifolds of .
For illustrative purposes let us consider the $`𝐙_2`$ example. The gauge group (at the $`𝐙_2𝐙_2`$ freely acting orbifold points) is $`U(8)_{99}U(8)_{55}`$. (Here we drop the $`[𝐙_2]_{99}[𝐙_2]_{55}`$ discrete gauge subgroup as no massless states carry non-trivial charges with respect to the latter.) If the Wilson lines on $`T^2`$ are of the $`D_4`$ type, then the massless hypermultiplets are given by:
$$2\times (\mathrm{𝟑𝟔};\mathrm{𝟏}),2\times (\mathrm{𝟏};\mathrm{𝟑𝟔}),(\mathrm{𝟖};\mathrm{𝟖}).$$
(241)
On the other hand, if the Wilson lines on $`T^2`$ are of the $`D_4^{}`$ type, then the massless hypermultiplets are given by:
$$2\times (\mathrm{𝟐𝟖};\mathrm{𝟏}),2\times (\mathrm{𝟏};\mathrm{𝟐𝟖}),(\mathrm{𝟖};\mathrm{𝟖}).$$
(242)
As we have already mentioned, these points belong to the same branch of the moduli space corresponding to the Coulomb branch.
In fact, most of the above discussion also applies to the $`M=3`$ cases, except that here we can also have $`B`$-flux on K3, so that the corresponding four dimensional models are obtained by compactifying the six dimensional $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ models with $`b=0,2,4`$ on $`T^2`$ with $`B`$-flux, and the former are recovered in the large $`T^2`$ limit<sup>\**‡</sup><sup>\**‡</sup>\**‡Here we can also consider $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ orientifolds with $`B`$-flux turned on both on $`T^2`$ and K3. In the limit of large volume $`T^2`$ we recover the six dimensional $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ models (with D5-branes only) discussed in subsection B of section III. Here we note that there is a possibility of a non-perturbative inconsistency in the four dimensional $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ models with $`B`$-flux turned on on K3 (regardless of the $`B`$-flux on $`T^2`$). This is in complete parallel with our discussion for the corresponding six dimensional $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ models. Note, however, that such a non-perturbative inconsistency is not expected in the four dimensional $`\mathrm{\Omega }J^{}`$ $`𝐙_3`$ models with $`B`$-flux turned on on $`T^2`$ only. Neither should it occur in the four dimensional $`\mathrm{\Omega }RJ^{}`$ $`𝐙_3`$ models with $`B`$-flux on K3 (regardless of the $`B`$-flux on $`T^2`$)..
Another class of orientifolds we can consider here is the following. Let $`z_1`$ parametrize $`T^2`$, and $`z_2,z_3`$ parametrize $`T^4`$ in $`\mathrm{K3}=T^4/𝐙_M`$. For simplicity let us assume that $`T^4=T^2T^2`$, where $`z_2,z_3`$ parametrize these two 2-tori. (To avoid confusion, from now on we will refer to the 2-torus parametrized by $`z_1`$ as $`\stackrel{~}{T}^2`$.) Then we can consider the $`\mathrm{\Omega }R^{}J^{}`$ orientifold of Type IIB on $`T^2(T^4/𝐙_M)`$, where the action of $`R^{}`$ is given by $`R^{}z_1=z_1`$, $`R^{}z_2=z_2`$, $`R^{}z_3=z_3`$. For $`M=3`$ we then have only D5-branes wrapping the 2-torus in $`T^4`$ (or, more precisely, an orbifold thereof) parametrized by $`z_3`$. That is, the locations of these D5-branes are given by points in the directions $`z_1,z_2`$. These $`𝐙_3`$ models with $`B`$-field turned on in various directions are straightforward to analyze along the lines of our previous discussions. We will therefore focus on the models with $`M=2,4,6`$, where we have two types of D-branes (intersecting at right angles). Thus, we have D5-branes wrapping the $`T^2`$ parametrized by $`z_3`$. We also have D5-branes wrapping the $`T^2`$ parametrized by $`z_2`$. This follows from the fact that we have O5-planes whose world-volumes coincide with the set of points fixed under $`R^{}`$, and we also have O5-planes whose world-volumes coincide with the set of points fixed under $`RR^{}`$, $`R`$ being the generator of the $`𝐙_2`$ subgroup of $`𝐙_M`$. Let $`\stackrel{~}{R}z_1=z_1`$. There are four points on $`\stackrel{~}{T}^2`$ fixed under the action of $`\stackrel{~}{R}`$: $`\xi _0=0`$, $`\xi _a=e_a/2`$, $`a=1,2,3`$, where $`e_i`$, $`i=1,2`$, are the vielbeins on $`\stackrel{~}{T}^2`$, and $`e_3e_1e_2`$. Note that we have total of 12 O5<sup>-</sup>-planes and 4 O5<sup>+</sup>-planes. Similarly, we have total of 12 O5<sup>′-</sup>-planes and 4 O5<sup>′+</sup>-planes. (This follows from the fact that we have half-integer $`B`$-flux on $`\stackrel{~}{T}^2`$ but trivial $`B`$-flux on K3.) For definiteness let us assume that 4 O5-planes corresponding to the $`\xi _0`$ fixed point are of the O5<sup>+</sup> type. Then the other 12 O5-planes corresponding to the fixed points $`\xi _a`$ are of the O5<sup>-</sup> type. What about the O5-planes? It is not difficult to see that with the above assumption the 4 O5-planes corresponding to the fixed point $`\xi _0`$ must be of the O5<sup>′+</sup> type. Similarly, the other 12 O5-planes corresponding to the fixed points $`\xi _a`$ must be of the O5<sup>′-</sup> type. This can be seen as follows. First note that the consistency of the $`R`$ projection in the $`55^{}`$ sector requires that $`\gamma _{R,5}`$ and $`\gamma _{R,5^{}}`$ both have either $`\pm i`$ or $`\pm 1`$ eigenvalues. Second, arguments similar to those in subsection A of section III imply that if $`\gamma _{R,5}`$ has eigenvalues $`\pm 1`$, then the O5-planes corresponding to a given fixed point on $`\stackrel{~}{T}^2`$ are of the type opposite to that of the corresponding O5-planes. That is, if the latter are, say, of the O5<sup>-</sup> type, then the former are of the O5<sup>′+</sup> type. On the other hand, if $`\gamma _{R,5}`$ has eigenvalues $`\pm i`$, then the O5-planes corresponding to a given fixed point on $`\stackrel{~}{T}^2`$ are of the same type as the corresponding O5-planes<sup>\**§</sup><sup>\**§</sup>\**§In fact, one can check the above statements in the following simple way. Note that if we T-dualize, say, on $`T^2`$ parametrized by $`z_2`$ (note that there are no subtleties with this T-duality procedure as the $`B`$-flux on this $`T^2`$ is trivial), then D5-branes turn into D7-branes, while D5-branes turn in to D3-branes. Analogous statements also apply to the corresponding O-planes. Thus, after T-duality transformation we obtain a background with O3- and O7-planes as well as the corresponding D-branes. For this system we can straightforwardly repeat the argument in the beginning of subsection A of section III (which was carried out for the 59 system but is identical for the 37 system), which (after T-dualizing back to the $`55^{}`$ system) leads precisely to the aforementioned conclusions. Equivalently, we can carry out these arguments for the $`55^{}`$ system by noting that in the $`55^{}`$ sector we have 4 Neumann-Dirichlet boundary conditions just as in the 37 sector or 59 sector.. Since we must have 4 O5<sup>+</sup>-planes and 12 O5<sup>-</sup>-planes as well as 4 O5<sup>′+</sup>-planes and 12 O5<sup>′-</sup>-planes (which follows from the tadpole cancellation conditions), it is then clear that $`\gamma _{R,5}`$ must have eigenvalues $`\pm i`$, and O5- and O5-planes corresponding to a given fixed point on $`\stackrel{~}{T}^2`$ must be of the same type<sup>\**¶</sup><sup>\**¶</sup>\**¶Note that this statement for the $`\mathrm{\Omega }R^{}J^{}`$ orientifolds is the analogue of the statement for the corresponding $`\mathrm{\Omega }J^{}`$ orientifolds that the Wilson lines in the 99 and 55 sectors must be of the same type - see above for details..
The massless spectra of the $`\mathrm{\Omega }R^{}J^{}`$ orientifolds are the same as those of the corresponding $`\mathrm{\Omega }J^{}`$ orientifolds. However, the massive spectra differ. Note, for instance, that in the $`\mathrm{\Omega }J^{}`$ orientifold we have 32 D9-branes as well as 32 D5-branes. The rank of the 99 and 55 gauge groups, however, is 8, and we have the $`[𝐙_2]_{99}[𝐙_2]_{55}`$ discrete gauge symmetry under which some massive Kaluza-Klein states are charged non-trivially. Such a discrete gauge symmetry is absent in the corresponding $`\mathrm{\Omega }R^{}J^{}`$ orientifolds - we have only 16 D5-branes and 16 D5-branes. Furthermore, if we take the size of $`\stackrel{~}{T}^2`$ in the $`\mathrm{\Omega }R^{}J^{}`$ orientifolds to infinity, we will not obtain six dimensional theories - Lorentz invariance in these backgrounds is always broken to that of a four dimensional theory. In particular, as was pointed out in , the rank of the gauge group in such models in not enhanced in the large $`\stackrel{~}{T}^2`$ limit - this is, in fact, a direct consequence of the fact that the number of each type of branes is only 16. What happens in the large $`\stackrel{~}{T}^2`$ limit is that the O-planes corresponding to the fixed points $`\xi _a`$ on $`\stackrel{~}{T}^2`$ are removed to infinity and decouple from the remaining O-planes corresponding to the fixed point $`\xi _0`$.
Before we end this subsection, we would like to make a remark on the motion of branes in the $`\mathrm{\Omega }R^{}J^{}`$ models. In particular, note that the motion of branes in the direction of $`\stackrel{~}{T}^2`$ corresponds to different points on the Coulomb branch of the $`𝒩=2`$ gauge theory. Suppose that D5-branes and D5-branes are on top of the respective O-planes corresponding to the same fixed point on $`\stackrel{~}{T}^2`$. Then we have a non-trivial $`55^{}`$ massless matter content. If we move, say, D5-branes off the corresponding O5-plane while leaving D5-branes untouched, then this corresponds to Higgsing the 55 sector gauge group, while the $`5^{}5^{}`$ gauge group is untouched. Note that the $`55^{}`$ states in this process become heavy (due to Higgsing). In the brane language this is simply the statement that $`55^{}`$ strings now cannot have zero length, so that the corresponding states are always heavy. On the other hand, if we move D5- and D5-branes off the corresponding fixed points together, then the $`55^{}`$-sector states remain massless. In the gauge theory language this corresponds to a special subspace of the Coulomb branch where the $`55^{}`$ hypermultiplets remain massless as the mass term coming from the coupling to the 55 Higgs field is precisely cancelled by the mass term coming from the coupling to the $`5^{}5^{}`$ Higgs field.
### B The $`𝒩=1`$ $`𝐙_3`$ Models
In this subsection we would like to discuss four dimensional $`𝒩=1`$ orientifolds of Type IIB on $`T^6/𝐙_3`$, where the generator $`\theta `$ of $`𝐙_3`$ acts on the complex coordinates $`z_I`$, $`I=1,2,3`$, parametrizing $`T^6`$ as follows: $`\theta z_I=\omega z_I`$, where $`\omega \mathrm{exp}(2\pi i/3)`$. A priori we can consider various orientifolds with O9-, O7-, O5- or O3-planes with $`B`$-flux turned on inside of and/or transverse to their world-volumes. Since other cases are straightforward to consider along the lines of our previous discussions, here we will focus on the models with O3-planes<sup>\**∥</sup><sup>\**∥</sup>\**∥Here we note that if we have $`B`$-flux inside of the world-volumes of O$`p`$-planes and the corresponding D$`p`$-branes in this background, then, just as in the corresponding six dimensional $`𝐙_3`$ models, there is a possibility of a non-perturbative inconsistency. If we, however, consider the models with O3-planes, such a non-perturbative inconsistency is not expected to arise.. Thus, we will discuss $`\mathrm{\Omega }R^{}J^{}(1)^{F_L}`$ orientifolds of Type IIB on $`T^6/𝐙_3`$, where $`R^{}z_I=z_I`$, and the action of $`J^{}`$ is analogous<sup>\****</sup><sup>\****</sup>\****Note, however, that, unlike the six dimensional cases, in the four dimensional orientifolds where some of the orbifold group elements twist all three complex coordinates $`z_I`$, the action of $`J^{}`$ (which is trivial in the untwisted sector, while it maps a $`g`$ twisted sector to its inverse $`g^1`$ twisted sector) does not seem to have a well defined geometric interpretation . Here we will ignore potential difficulties with such an interpretation (which could be seen , for instance, by considering a map of these orientifolds to F-theory ), and assume that such an action is well defined in the conformal field theory context. to that in the six dimensional orientifolds discussed in subsections B,C,D of section III. In such an orientifold we have $`n_f=32+32/2^{b/2}`$ O3<sup>-</sup>-planes, and $`n_{f+}=3232/2^{b/2}`$ O3<sup>+</sup>-planes, where $`b`$ is the rank of the $`B`$-flux. Moreover, it is not difficult to show that the twisted tadpole cancellation conditions read:
$$\mathrm{Tr}(\gamma _\theta )=(1)^{b/2}\times 4.$$
(243)
Note that for the untwisted Chan-Paton matrix we have $`\mathrm{Tr}(\gamma _I)=32/2^{b/2}`$.
Let us consider the $`b=2,4,6`$ cases separately<sup>\**††</sup><sup>\**††</sup>\**††Here we should point out that these models were originally discussed in . More precisely, the spectra of the $`b=2,6`$ models given in were erroneous as it was not realized there that the consistent orientifold projection in these cases is of the $`Sp`$ type. This was originally corrected in , and more recently in ..
$``$ For $`b=2`$ we have $`n_f=48`$ O3<sup>-</sup>-planes and $`n_{f+}=16`$ O3<sup>+</sup>-planes. The $`𝐙_3`$ symmetry requires that the O3-plane at the origin of $`T^6`$ be of the O3<sup>+</sup> type. If we place all 16 D3-branes on top of this O3-plane, then the gauge group is (note that $`\mathrm{Tr}(\gamma _\theta )=+4`$ in this case) $`U(4)Sp(8)`$, and the massless open string sector contains chiral supermultiplets in
$$\mathrm{\Phi }_s=3\times (\mathrm{𝟏𝟎},\mathrm{𝟏}),Q_s=3\times (\overline{\mathrm{𝟒}},\mathrm{𝟖}),s=1,2,3.$$
(244)
There is a non-trivial superpotential in this model given by:
$$𝒲=ϵ_{ss^{}s^{\prime \prime }}\mathrm{\Phi }_sQ_s^{}Q_{s^{\prime \prime }}.$$
(245)
$``$ For $`b=4`$ we have $`n_f=40`$ O3<sup>-</sup>-planes and $`n_{f+}=24`$ O3<sup>+</sup>-planes. The $`𝐙_3`$ symmetry requires that the O3-plane at the origin of $`T^6`$ be of the O3<sup>-</sup> type. If we place all 8 D3-branes on top of this O3-plane, then the gauge group is (note that $`\mathrm{Tr}(\gamma _\theta )=4`$ in this case) $`U(4)`$, and the massless open string sector contains chiral supermultiplets in $`\mathrm{\Phi }_s=3\times \mathrm{𝟔}`$. There are no renormalizable couplings in this model.
$``$ For $`b=6`$ we have $`n_f=36`$ O3<sup>-</sup>-planes and $`n_{f+}=28`$ O3<sup>+</sup>-planes. The $`𝐙_3`$ symmetry requires that the O3-plane at the origin of $`T^6`$ be of the O3<sup>+</sup> type. If we place all 4 D3-branes on top of this O3-plane, then the gauge group is (note that $`\mathrm{Tr}(\gamma _\theta )=+4`$ in this case) $`Sp(4)`$, and there are no massless open string sector states in this model.
### C The $`𝒩=1`$ $`𝐙_7`$ Models
In this subsection we discuss four dimensional $`𝒩=1`$ orientifolds of Type IIB on $`T^6/𝐙_7`$, where the generator $`g`$ of $`𝐙_7`$ acts on the complex coordinates $`z_I`$, $`I=1,2,3`$, parametrizing $`T^6`$ as follows: $`gz_1=\alpha `$, $`gz_2=\alpha ^2z_2`$, $`gz_3=\alpha ^4z_3`$, where $`\alpha \mathrm{exp}(2\pi i/7)`$. As in the previous section, let us focus on the cases where we have O3-planes. Thus, consider the $`\mathrm{\Omega }R^{}J^{}(1)^{F_L}`$ orientifold of Type IIB on $`T^6/𝐙_7`$ ($`R^{}z_I=z_I`$). Here we note that the rank of the $`B`$-flux can take only two values in this case: $`b=0,6`$. The reason why is that only for these values of $`b`$ does the $`𝐙_7`$ orbifold act crystallographically on $`T^6`$. Another way of seeing this is as follows. Note that we have $`n_f=32\pm 32/2^{b/2}`$ O3-planes for the rank $`b`$ $`B`$-flux. The O3-plane at the origin is invariant under the $`𝐙_7`$ twist $`g`$. However, O3-planes at other fixed points of $`R^{}`$ must come in groups of 7 such that they are permuted by the $`𝐙_7`$ orbifold within each group. However, for $`b=2,4`$ neither $`n_f1`$ nor $`n_{f+}1`$ are divisible by 7. For $`b=0,6`$ both $`n_f1`$ and $`n_{f+}`$ are divisible by 7, so we conclude that in these cases the O3-plane at the origin is of the O3<sup>-</sup> type. In the $`b=6`$ case we must place all 4 D3-branes at this O3-plane, which is consistent with the twisted tadpole cancellation condition $`\mathrm{Tr}(\gamma _g)=+4`$ . The gauge group in this case is $`SO(4)`$ with no massless open string matter.
### D The $`𝒩=1`$ $`𝐙_3𝐙_3`$ Models
In this subsection we would like to discuss four dimensional $`𝒩=1`$ orientifolds of Type IIB on $`(T^2T^2T^2)/(𝐙_3𝐙_3)`$, where the action of the generators $`\theta `$ and $`\theta ^{}`$ of the two $`𝐙_3`$ subgroups on the complex coordinates $`z_I`$, $`I=1,2,3`$, parametrizing the three 2-tori is as follows ($`\omega \mathrm{exp}(2\pi i/3)`$): $`\theta z_1=\omega z_1`$, $`\theta z_2=\omega ^1z_2`$, $`\theta z_3=z_3`$, $`\theta ^{}z_1=z_1`$, $`\theta ^{}z_2=\omega z_2`$, $`\theta ^{}z_3=\omega ^1z_3`$. Here we will focus on the $`\mathrm{\Omega }R^{}J^{}(1)^{F_L}`$ orientifolds ($`R^{}z_I=z_I`$). Using our previous results it is then no difficult to show that we have the following models<sup>\**‡‡</sup><sup>\**‡‡</sup>\**‡‡Solutions to the tadpole cancellation conditions for these models were found in .:
$``$ $`b=2`$. The gauge group is $`U(4)U(4)`$ with the following massless open string sector chiral multiplets
$$Q=(\overline{\mathrm{𝟒}},\overline{\mathrm{𝟒}}),R=(\mathrm{𝟒},\overline{\mathrm{𝟒}}),\mathrm{\Phi }=(\mathrm{𝟏},\mathrm{𝟏𝟎}),$$
(246)
and the superpotential
$$𝒲=QR\mathrm{\Phi }.$$
(247)
$``$ $`b=4`$. The gauge group is $`U(4)`$, and in the open string sector we have a massless chiral multiplet in $`\mathrm{𝟔}`$ of $`U(4)`$. There are no renormalizable couplings in this case.
$``$ $`b=6`$. The gauge group is $`Sp(4)`$, and there is no massless matter in the open string sector.
### E The $`𝒩=1`$ $`𝐙_6`$ Models
In this subsection we would like to discuss four dimensional $`𝒩=1`$ orientifolds of Type IIB on $`T^6/𝐙_6`$, where the generators $`\theta `$ and $`R`$ of the $`𝐙_3`$ and $`𝐙_2`$ subgroups of the $`𝐙_6`$ orbifold group act on the complex coordinates $`z_I`$, $`I=1,2,3`$, parametrizing the three 2-tori (we assume that $`T^6=T^2T^2T^2`$) as follows ($`\omega \mathrm{exp}(2\pi i/3)`$): $`\theta z_I=\omega z_I`$, $`Rz_1=z_1`$, $`Rz_{2,3}=z_{2,3}`$. Note that the $`\mathrm{\Omega }J^{}`$ $`𝐙_6`$ model with trivial $`B`$-flux was originally constructed in the second reference in . In <sup>††\*</sup><sup>††\*</sup>††\*This model, among other four dimensional $`𝒩=1`$ models with and without $`B`$-flux, was also discussed in . the following model with $`B`$-flux was discussed<sup>†††</sup><sup>†††</sup>†††This model was discussed in in the phenomenological context of TeV-scale brane world. For a partial list of other recent developments in these directions, see .. Consider the $`\mathrm{\Omega }J^{}`$ orientifold of Type IIB on $`(T^2T^2T^2)/𝐙_6`$ with $`b=2`$ $`B`$-flux turned on on the second or third $`T^2`$ only. It should be clear that in this model we are going to have inconsistencies similar to those we have encountered in the six dimensional $`𝐙_6`$ models with $`B`$-flux<sup>††‡</sup><sup>††‡</sup>††‡Note that the O5-plane at the $`R`$ fixed point $`z_2=z_3=0`$ must be of the O5<sup>+</sup> type for the corresponding background to be $`𝐙_3`$ symmetric. This would then require that the twisted Chan-Paton matrices $`\gamma _{R,5}`$ as well as $`\gamma _{R,9}`$ have eigenvalues $`\pm 1`$. In , however, these Chan-Paton matrices were assumed to have eigenvalues $`\pm i`$.. To avoid these difficulties we can assume that $`B`$-flux is turned on inside of both D9- and D5-branes present in this case. That is, consider the $`\mathrm{\Omega }J^{}`$ $`𝐙_6`$ model with $`b=2`$ $`B`$-flux turned on on the first $`T^2`$. The corresponding Wilson lines must be of the same type in both 99 and 55 sectors. In fact, to be compatible with the $`𝐙_3`$ orbifold action, they must be of the $`D_4`$ (and not $`D_4^{}`$) type. The important point here is that the multiplicity of states in the 59 sector (before the $`𝐙_3`$ orbifold projection) is $`\xi _{59}=1`$ (and not 2, which would be the case in the model of had it been consistent).
Instead of describing the massless spectrum of the above model, we will discuss that of a different model<sup>††§</sup><sup>††§</sup>††§The reason for this is that in the former model there is a possibility of a non-perturbative inconsistency arising along the lines of our previous discussions, while in the model we are going to discuss next such an inconsistency is not expected to arise. (these two models actually have identical massless spectra albeit their massive spectra are different). Thus, consider the $`\mathrm{\Omega }R^{}J^{}`$ orientifold of Type IIB on $`(T^2T^2T^2)/𝐙_6`$, where $`R^{}z_{1,2}=z_{1,2}`$, $`Rz_3=z_3`$. Let us turn on $`b=2`$ $`B`$-flux on the first $`T^2`$ parametrized by $`z_1`$. Note that in this model we have two types of D-branes (intersecting at right angles). Thus, we have D5-branes wrapping the $`T^2`$ parametrized by $`z_3`$. We also have D5-branes wrapping the $`T^2`$ parametrized by $`z_2`$. Just as in the previous subsection, the O5- and O5-planes corresponding to the same fixed point on $`T^2`$ parametrized by $`z_1`$ are of the same type. Moreover, to be compatible with the $`𝐙_3`$ orbifold action, the 4 O5-planes corresponding to the origin of this $`T^2`$ must be of the O5<sup>+</sup> type, and, similarly, the 4 O5-planes corresponding to the origin of this $`T^2`$ must be of the O5<sup>′+</sup> type. The other 12 O5-planes are of the O5<sup>-</sup> type, and, similarly, the other 12 O5-planes are of the O5<sup>′-</sup> type. Let us place all 16 D5-branes and 16 D5-branes at an O5<sup>+</sup>-plane and an O5<sup>′+</sup>-plane, respectively. The solution to the twisted tadpole cancellation conditions reads (up to equivalent representations) :
$`\gamma _{\theta ,5}=\gamma _{\theta ,5^{}}=\mathrm{diag}(1,1,\omega ,\omega ^1)I_4,`$ (248)
$`\gamma _{R,5}=\gamma _{R,5^{}}=I_4i\sigma _3I_2.`$ (249)
The gauge group of this model is $`[U(2)U(2)U(4)]_{55}[U(2)U(2)U(4)]_{99}`$, and the massless open string chiral matter reads:
$`\mathrm{\Phi }_{1,2}=2\times (\mathrm{𝟑},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},\stackrel{~}{\mathrm{\Phi }}_{1,2}=2\times (\mathrm{𝟏},\overline{\mathrm{𝟑}},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},`$ (250)
$`P_{1,2}=2\times (\overline{\mathrm{𝟐}},\mathrm{𝟏},\overline{\mathrm{𝟒}};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},\stackrel{~}{P}_{1,2}=2\times (\mathrm{𝟏},\mathrm{𝟐},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},`$ (251)
$`P_3=(\overline{\mathrm{𝟐}},\mathrm{𝟏},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},\stackrel{~}{P}_3=(\mathrm{𝟏},\mathrm{𝟐},\overline{\mathrm{𝟒}};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},R=(\mathrm{𝟐},\overline{\mathrm{𝟐}},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏})_{55},`$ (252)
$`\mathrm{\Phi }_{1,2}^{}=2\times (\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟑},\mathrm{𝟏},\mathrm{𝟏})_{5^{}5^{}},\stackrel{~}{\mathrm{\Phi }}_{1,2}^{}=2\times (\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\overline{\mathrm{𝟑}},\mathrm{𝟏})_{5^{}5^{}},`$ (253)
$`P_{1,2}^{}=2\times (\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\overline{\mathrm{𝟐}},\mathrm{𝟏},\overline{\mathrm{𝟒}})_{5^{}5^{}},\stackrel{~}{P}_{1,2}^{}=2\times (\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟐},\mathrm{𝟒})_{5^{}5^{}},`$ (254)
$`P_3^{}=(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\overline{\mathrm{𝟐}},\mathrm{𝟏},\mathrm{𝟒})_{5^{}5^{}},\stackrel{~}{P}_3^{}=(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟐},\overline{\mathrm{𝟒}})_{5^{}5^{}},R^{}=(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟐},\overline{\mathrm{𝟐}},\mathrm{𝟏})_{5^{}5^{}},`$ (255)
$`S=(\mathrm{𝟐},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟐},\mathrm{𝟏},\mathrm{𝟏})_{55^{}},T=(\mathrm{𝟏},\mathrm{𝟐},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟒})_{55^{}},U=(\mathrm{𝟏},\mathrm{𝟏},\mathrm{𝟒};\mathrm{𝟏},\mathrm{𝟐},\mathrm{𝟏})_{55^{}},`$ (256)
$`\stackrel{~}{S}=(\mathrm{𝟏},\overline{\mathrm{𝟐}},\mathrm{𝟏};\mathrm{𝟏},\overline{\mathrm{𝟐}},\mathrm{𝟏})_{55^{}},\stackrel{~}{T}=(\overline{\mathrm{𝟐}},\mathrm{𝟏},\mathrm{𝟏};\mathrm{𝟏},\mathrm{𝟏},\overline{\mathrm{𝟒}})_{55^{}},\stackrel{~}{U}=(\mathrm{𝟏},\mathrm{𝟏},\overline{\mathrm{𝟒}};\overline{\mathrm{𝟐}},\mathrm{𝟏},\mathrm{𝟏})_{55^{}},`$ (257)
where $`\mathrm{𝟐}`$ and $`\overline{\mathrm{𝟐}}`$ of $`U(2)`$ carry the $`U(1)`$ charges $`+1`$ and $`1`$, respectively, while $`\mathrm{𝟑}`$ and $`\overline{\mathrm{𝟑}}`$ of $`U(2)`$ carry the $`U(1)`$ charges $`+2`$ and $`2`$, respectively. Similarly, $`\mathrm{𝟒}`$ and $`\overline{\mathrm{𝟒}}`$ of $`U(4)`$ carry the $`U(1)`$ charges $`+1`$ and $`1`$, respectively. Note that this spectrum is the same as that in except for the $`55^{}`$ sector multiplicity of states, which in the above model is 1, while in it was assumed to be 2 (in the corresponding 59 sector)<sup>††¶</sup><sup>††¶</sup>††¶Here we should point out that the aforementioned multiplicity being 1 is consistent with the $`U(1)`$ anomaly cancellation in the above model via the generalized Green-Schwarz mechanism, while attempts to implement the latter with the aforementioned multiplicity being 2 (as in ) appear to run into various difficulties ..
The above model is expected to be consistent. The superpotential in this model is given by:
$`𝒲=`$ $`P_1\stackrel{~}{P}_2R+P_2\stackrel{~}{P}_1R+\mathrm{\Phi }_1P_2P_3+\mathrm{\Phi }_2P_1P_3+\stackrel{~}{\mathrm{\Phi }}_1\stackrel{~}{P}_2\stackrel{~}{P}_3+\stackrel{~}{\mathrm{\Phi }}_2\stackrel{~}{P}_1\stackrel{~}{P}_3+`$ (260)
$`P_1^{}\stackrel{~}{P}_2^{}R^{}+P_2^{}\stackrel{~}{P}_1^{}R^{}+\mathrm{\Phi }_1^{}P_2^{}P_3^{}+\mathrm{\Phi }_2^{}P_1^{}P_3^{}+\stackrel{~}{\mathrm{\Phi }}_1^{}\stackrel{~}{P}_2^{}\stackrel{~}{P}_3^{}+\stackrel{~}{\mathrm{\Phi }}_2^{}\stackrel{~}{P}_1^{}\stackrel{~}{P}_3^{}+`$
$`S\stackrel{~}{U}P_3+U\stackrel{~}{S}\stackrel{~}{P}_3+T\stackrel{~}{T}R+S\stackrel{~}{T}P_3^{}+T\stackrel{~}{S}\stackrel{~}{P}_3^{}+U\stackrel{~}{U}R^{}.`$
Using this superpotential it is not difficult to see that if we Higgs the gauge group along the lines of <sup>††∥</sup><sup>††∥</sup>††∥In particular, assume that the $`S`$ and $`\stackrel{~}{S}`$ fields acquire non-zero VEVs, which break the gauge group down to $`U(2)_{\mathrm{diag}}U(2)_{\mathrm{diag}}U(4)_{55}U(4)_{5^{}5^{}}`$. Then the observation of is that one can treat, say, the $`U(2)_{\mathrm{diag}}U(2)_{\mathrm{diag}}U(4)_{55}`$ part of this gauge group as the Pati-Salam gauge symmetry. Here we should point out that there are certain subtleties related to the $`U(1)`$ factors in the aforementioned Higgsing, which we will not discuss in this paper as the key point here is that even if such Higgsing is possible, the number of chiral Pati-Salam generations still cannot be 3., then the number of remaining chiral generations for the Pati-Salam gauge group is 2 (and not 3 as it was originally intended in ), which is due to the fact that the multiplicity of the $`55^{}`$ states in the above model is only 1 (and not 2 as was assumed in for the 59 sector states).
Before we end this subsection, let us note that in orientifolds of the above $`𝐙_6`$ orbifold compactification we cannot turn on $`b=4,6`$ $`B`$-flux without running into the aforementioned inconsistencies - for $`b=4,6`$ we cannot avoid having two sets of D-branes such that $`B`$-flux is inside of the world-volumes of one set of D-branes while the other set is transverse to it<sup>††\**</sup><sup>††\**</sup>††\**Note that, for essentially the same reasons, the $`𝒩=1`$ $`𝐙_2𝐙_2`$ as well as $`𝐙_2𝐙_2𝐙_3`$ models with $`B`$-flux discussed in also suffer from various subtle inconsistencies for all three values of $`b=2,4,6`$. Note, however, that the $`𝐙_2𝐙_2𝐙_3`$ model with trivial $`B`$-flux originally constructed in the first reference in is consistent. This model when interpreted in the phenomenological context has three chiral generations, and its phenomenological implications were studied in ..
## V Comments
In this section we would like to comment on various issues related to discussions in the previous sections. To begin with, let us note that non-trivial multiplicity of states, which is due to non-trivial $`B`$-flux, in the 59 sectors of some of the models discussed in the previous sections appears to be in conflict with $`𝐙_M`$ orbifold projections (with $`M=4,6`$) acting on the corresponding coordinates (that is, those transverse to D5-branes). This observation might be relevant for other compactifications with such non-trivial multiplicity of states in $`pp^{}`$ sectors. In particular, such sectors arise in some of the models recently discussed in , where the orientifold action involves complex conjugation of the compact coordinates. The analogue of the quantized $`B`$-flux in such backgrounds is given by the components of the complex structure corresponding to the $`B`$-flux in the Kähler structure under the interchange of the complex and Kähler structures. In fact, the origin of non-trivial multiplicity of states in such orientifolds is also analogous to that in orientifolds with non-zero $`B`$-flux, so a priori there might be a possibility of subtle inconsistencies, similar to those we have found in the latter backgrounds, also arising in the former ones. It would be interesting to understand this issue in a bit more detail, but this would be outside of the scope of this paper.
The second comment concerns possible generalizations to non-supersymmetric backgrounds with $`B`$-flux. Recently compact non-supersymmetric orientifolds have been constructed in, for instance, by introducing both D-branes as well as anti-D-branes in orientifolds of Type IIB on orbifolds that preserve some number of supersymmetries. However, regardless of the $`B`$-flux such models have one peculiar feature that some of the NS-NS tadpoles must be non-vanishing. The reason for this is that in the aforementioned construction one introduces only O\- and/or O<sup>++</sup>-planes (see section II for notations) which are supersymmetric in the sense that the NS-NS and R-R tadpoles for each of these O-planes are identical. However, if we now introduce both D-branes and anti-D-branes, then it is clear that we cannot cancel both NS-NS and R-R tadpoles simultaneously. Thus, a D-brane has R-R charge $`+1`$, while an anti-D-brane has R-R charge $`1`$. Both of these objects, however, give rise to identical NS-NS tadpoles. Thus, if we require R-R tadpole cancellation, then we have some uncanceled NS-NS tadpoles. In the standard argument was employed that such tadpoles might not be dangerous as they could possibly be dealt with via the Fischler-Susskind mechanism. However, a priori it is unclear what the corresponding consistent backgrounds would be if any<sup>††††</sup><sup>††††</sup>††††This is analogous to the situation in the non-supersymmetric $`Sp(32)`$ theory discussed in section II. In fact, some of the aforementioned backgrounds could be thought of as compactifications of this theory..
Actually, we can understand this point in a bit more detail. If some of the twisted NS-NS tadpoles are non-zero, then we would have to employ the Fischler-Susskind mechanism for twisted (as well as untwisted) scalar fields that couple to D-branes and O-planes. Shifting the VEVs of the twisted scalars, however, implies that we blow up the orbifold, and the consistent background is no longer an exactly solvable conformal field theory. Orientifolds of such backgrounds are difficult to study, so it is unclear what the resulting theory would look like if there at all exists the corresponding consistent background. One things, however, is quite clear - a priori there is no reason to believe that, if such a non-orbifold background indeed exists, the open string spectrum of the theory would be the same.
There is, however, a way to avoid the aforementioned difficulty by considering backgrounds where all twisted NS-NS tadpoles cancel. Nonetheless, we have some uncanceled untwisted NS-NS tadpoles, which imply that VEVs of some of the untwisted closed string scalars must be shifted. This, in turn, gives us a hint of what the corresponding consistent backgrounds could look like. Thus, imagine that we have D$`p`$-branes as well as D$`\overline{p}`$-branes transverse to some compact coordinates. To avoid appearance of open string tachyons, we must place D$`p`$-branes and D$`\overline{p}`$-branes far enough apart from each other. In fact, to avoid a possibility of brane-anti-brane annihilation, we can assume that branes and anti-branes are stuck at, say, the corresponding orbifold fixed points (for definiteness we will assume that at the orbifold fixed point located at the origin we have branes and not anti-branes). Then tachyons are absent if the separation between the fixed points (related to the compactification radii) is large enough. However, supersymmetry is broken, and the cosmological constant is non-zero. In fact, it depends upon the compactification radii. For large enough values of the latter the vacuum energy monotonically decreases with radii. In fact, in the decompactification limit the branes and anti-branes decouple<sup>††‡‡</sup><sup>††‡‡</sup>††‡‡This is the case if the number of decompactified dimensions is larger than 2. from each other, and the resulting theory with branes located at the origin is now supersymmetric. In fact, all tadpoles in this theory (that is, both R-R and NS-NS tadpoles) cancel. Thus, here we see that there exists a consistent background for such theories which can be reached via the Fischler-Susskind mechanism - it is a (partially) decompactified background with branes only, which is supersymmetric. The latter has lower vacuum energy compared with the original compactified theory with runaway scalar potential (as a function of compactification radii). In this respect such compactifications are similar to the standard Scherk-Schwarz type of compactifications (or generalizations thereof) which are unstable to decompactification and eventually end up in a supersymmetric vacuum.
Note that we can construct non-supersymmetric theories with both O/O<sup>++</sup> (that is, supersymmetric) as well as O<sup>-+</sup>/O<sup>+-</sup> (that is, non-supersymmetric) orientifold planes, where all R-R as well as NS-NS tadpoles cancel. Non-compact versions of such theories (which contain tachyons) were originally constructed in <sup>‡‡\*</sup><sup>‡‡\*</sup>‡‡\*There such models were discussed in the context of large $`N`$ gauge theories, so that the presence of closed string tachyons did not pose a problem .. One can easily generalize this construction to compact cases where one can avoid tachyons by considering (partially) freely acting orbifolds. However, as usual, the vacuum energy in such models has a runaway behavior with the stable supersymmetric vacuum reached in a (partial) decompactification limit.
Finally, we note that some of the discussions of non-perturbative K3 orientifolds with $`B`$-flux in the third reference in appear to be affected by the results of this paper, in particular, we expect various aforementioned subtleties arising in these models as well, so that they should also be revisited. This, however, is outside of the scope of this paper, and a more detailed discussion of such and related compactifications will be given elsewhere<sup>‡‡†</sup><sup>‡‡†</sup>‡‡†In certain (limited) cases non-perturbative orientifolds of can be described in terms of perturbative orientifolds of non-geometric backgrounds. The latter approach makes it possible to better understand the former, including the issues related to non-zero $`B`$-flux.. Also, some other $`𝒩=1`$ models (such as the $`𝐙_6^{}`$ models discussed in the first reference in ) should also be considered in this context.
###### Acknowledgements.
I would like to thank Ofer Aharony, Alex Buchel, Amihay Hanany, Martin Roček, Gary Shiu, Tom Taylor and Henry Tye for valuable discussions. This work was supported in part by the National Science foundation. Parts of this work were completed while I was visiting ITP at UCSB as part of the 1999 ITP program “Supersymmetric Gauge Dynamics and String Theory”. I would like to thank the organizers and participants of the program for creating a stimulating atmosphere, and ITP for their kind hospitality. Parts of this manuscript were typed up during my stays at Rutgers University and Harvard University. I would also like to thank Albert and Ribena Yu for financial support. |
warning/0001/cond-mat0001253.html | ar5iv | text | # References
International Journal of Theoretical and Applied Finance
c World Scientific Publishing Company and Imperial College Press
LEARNING SHORT–OPTION VALUATION IN THE PRESENCE OF RARE EVENTS
M. RABERTO<sup>1</sup>, G. CUNIBERTI<sup>2</sup>, E. SCALAS<sup>3</sup>, M. RIANI<sup>1</sup>, F. MAINARDI<sup>4</sup>, and G. SERVIZI<sup>4</sup>
<sup>1</sup>INFM and Dipartimento di Fisica, Università di Genova, via Dodecaneso 33, I-16146 Genova
<sup>2</sup>Max-Planck-Institut PKS, Nöthnitzer Str. 38, D-01187, Dresden
<sup>3</sup>Lab33 S.r.l., corso Perrone 24, I-16152 Genova
<sup>4</sup>INFN and Dipartimento di Fisica, Università di Bologna, via Irnerio 46, I-40126 Bologna
Web: www.econophysics.org
(Leave 1 inch blank space for publisher.)
Keywords: Econophysics; learning; option pricing; statistical finance.
We extend the neural–network approach for the valuation of financial derivatives developed by Hutchinson et al.<sup>?</sup> to the case of fat–tailed distributions of the underlying asset returns. We use a two–layer perceptron with three inputs, four hidden neurons, and one output. The input parameters of the network are: the simulated price of the underlying asset $`F`$ divided by the strike price $`E`$, the time–to–maturity $`T`$, and the ratio $`|FE|/T`$. The latter takes into account the volatility smile, whereas the price $`F`$ is generated by the method of Gorenflo et al.<sup>?</sup> based on fractional calculus. The output parameter is the call price $`C`$ over $`E`$. The learning–set option price $`C`$ is computed by means of a formula given by Bouchaud and Potters<sup>?</sup>. Option prices obtained by means of this learning scheme are compared with liffe option prices on German treasury bond (bund) futures.
The figure on the left shows the network performance. In the inset, the difference between the market price and the estimated price is plotted vs. the market price. Further details can be found at www.econophysics.org under the link “research”. |
warning/0001/cond-mat0001326.html | ar5iv | text | # Improved Theory of Single-Particle Properties of Fermi Systems Using Sea-Bosons
## I Introduction
Both the Hartree-Fock and the Random Phase approximations(RPA) are widely used in the physics literature to study Fermi systems. The Bogoliubov theory is the analog of the random-phase approximation for Bose systems. This latter fact has been demonstrated in detail in our article. The analysis presented here could be repeated for Bose systems as well. But let us focus on the more important Fermi system. Here we try and address the question of validity of the Hartree-Fock approximation and the RPA. This is important since we showed in our earlier work that the simplest and most natural application of bosonization was in the regime where the RPA was exact. It is therefore quite pertinent to ask whether this approximation is adequate. It is well-known that there are many problems with the RPA. In particular, the pair-correlation function has unphysical behaviour at short distances(it becomes negative) . Therefore it is necessary to find better approximations. Many attempts have been made in the literature in this regard. The Hubbard approximation\[Hubbard\], the Singwi-Sjolander approach\[Singwi\] and many others discussed in the text by Mahan are among the more prominent. Many of them have found to resolve this problem of short distance behavior of the pair correlation function. However it seems that none of them are able to address the question of single-particle properties. This issue has acquired an urgency in the recent past given that a segment of the high $`T_c`$ community is convinced that the non-superconducting state of these materials is non-Fermi liquid. Therefore being able to find a theory that describes the non-Fermi liquid regime accurately is desirable. In this attempt, we shall adopt a two-track approach. On the one hand we reinterpret the RPA so that it becomes more widely applicable. We argue that while the mean-field approximation carried out on the density operator is the Hartree-Fock approximation, the RPA manifests itself as mean-field theory applied not to the density operator but to the number operator. The density operator measures how the electrons are distributed in real space whereas the number operator measures how the electrons are distributed in momentum space. Just as the Hartree-Fock approximation is valid when fluctuations in the density of electrons at each point in real space is small, the RPA is valid when fluctuations in the momentum distribution of electrons is small compared with the average momentum distribution which measures the probability of an electron to possess a given momentum. We also introduce another mathematical transformation within the RPA-scheme and find that this maps a purely repulsive interaction to a purely attractive one but one that makes the particle exchange a momentum different from the usual one involved in two-body collisions. This unusual transformation allows us to define and compute yet another dielectric function of this system. A physical interpretation of this curious change in sign is given. A suitable combination of the purely repulsive(both actually and apparently) and purely attractive(only apparently, the electrons still repel one another) is shown to be better than either one spearately. In the end, we point out how a motivated researcher may be able to compute the phase diagram of the homogeneous electron gas by appropriately combining these three ingredients, although we ourselves restrict our attention to simple analytically solvable cases.
(i) Sea-bosons, indispensable for computing single-particle properties.
(ii) Generalised RPA and beyond (with fluctuations in the momentum distribution). Many new dielectric functions may be found here.
(iii) Repulsion-attraction duality : the enigmatic transformation that allows us to view the purely repulsive interaction between electrons as being attractive and repulsive at the same time when viewed in the sea-boson language.
In what follows, we try and combine the bosonization approach of our earlier articles with this notion of generalised RPA and repulsion-attraction duality and try and compute the single-particle properties. Just in order to whet the reader’s appetite for the calculation of single-particle properties using sea-bosons, we pause here to explore another option, one that suggests itself quite naturally when one views RPA as a mean-field idea applied to the number operator rather than the density operator.
## II Single-Particle Properties Without Using Bosonization
Let us try and write down some representative examples when RPA and Hartree-Fock approximations are used. Let us take for example, the jellium model .
$$H=\underset{𝐤}{}ϵ_𝐤c_𝐤^{}c_𝐤+\underset{𝐪0}{}\frac{v_𝐪}{2V}(\rho _𝐪\rho _𝐪N)$$
(1)
If we apply the mean-field idea on the density, that is, replace $`\rho _𝐪`$ by $`\rho _𝐪`$ then we get a hamiltonian that does not involve any coulomb interaction at all(apart from an additive constant).
$`H={\displaystyle \underset{𝐤}{}}ϵ_𝐤c_𝐤^{}c_𝐤+{\displaystyle \underset{𝐪0}{}}{\displaystyle \frac{v_𝐪}{2V}}(\rho _𝐪\rho _𝐪N)`$
$$=\underset{𝐤}{}ϵ_𝐤c_𝐤^{}c_𝐤N\underset{𝐪0}{}\frac{v_𝐪}{2V}$$
(2)
Therefore this approximation is bad in the extreme. However the random-phase approximation is still valid for this system. It is the mean-field idea applied not to the density but to the number operator. For this we have to rewrite the full hamiltonian given below,
$$H=\underset{𝐤}{}ϵ_𝐤c_𝐤^{}c_𝐤+\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤,𝐤^{^{}}}{}c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}c_{𝐤𝐪/2}$$
(3)
This has to be replaced by,
$$H=H_0\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤𝐤^{^{}}}{}c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤𝐪/2}$$
(4)
where
$$H_0=\underset{𝐤}{}ϵ_𝐤c_𝐤^{}c_𝐤\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤}{}n_{𝐤+𝐪/2}n_{𝐤𝐪/2}$$
(5)
where $`n_𝐤=c_𝐤^{}c_𝐤`$. Let us write,
$$n_𝐤=n_𝐤+\delta n_𝐤$$
(6)
We plug the above decomposition into $`H_0`$ and find that if we neglect terms quadratic in the fluctuations, we get what we are after, namely the generalised RPA called for simplicity as just RPA.
$$H_{RPA}=E_0^{^{}}+\underset{𝐤}{}\stackrel{~}{ϵ}_𝐤c_𝐤^{}c_𝐤$$
(7)
$$\stackrel{~}{ϵ}_𝐤=ϵ_𝐤\underset{𝐪}{}\frac{v_𝐪}{V}n_{𝐤𝐪}$$
(8)
The average occupation is
$$n_𝐤=\frac{1}{exp(\beta (\stackrel{~}{ϵ}_𝐤\mu ))+1}$$
(9)
The chemical potential $`\mu `$ has to be fixed by making sure that,
$$\underset{𝐤}{}n_𝐤=N$$
(10)
At zero temperature $`\mu =ϵ_F`$, this quantity is equal to the usual Fermi energy when $`v_𝐪=0`$.
$$n_𝐤=\theta (ϵ_F\stackrel{~}{ϵ}_𝐤)=\theta (ϵ_Fϵ_𝐤+\underset{𝐪0}{}\frac{v_𝐪}{V}n_{𝐤𝐪})$$
(11)
and $`\theta `$ is the Heaviside step function. We can now demonstrate that the generalised RPA dielectric function(the Lindhard dielectric function being a weak coupling limit of this) may be recovered using the following procedure. If one considers an extremely weak external perturbation applied to the system and follows the discussion in Mahan one arrives at the following formula for the dielectric function,
$$ϵ_{gRPA}(𝐪,\omega )=1+\frac{v_𝐪}{V}\underset{𝐤}{}\frac{n_{𝐤+𝐪/\mathrm{𝟐}}n_{𝐤𝐪/\mathrm{𝟐}}}{\omega \stackrel{~}{ϵ}_{𝐤+𝐪/\mathrm{𝟐}}+\stackrel{~}{ϵ}_{𝐤𝐪/\mathrm{𝟐}}}$$
(12)
The only point to bear in mind is that we have to use the full interacting momentum distribution rather than just its noninteracting value. When one includes fluctuations in the momentum distribution however, the answer given later, is very different from the usual RPA or even the generalised RPA. This feature of having the full momentum distribution in the numerator may be found in our earlier work . Thus we can see that there is a whole new set of approximations that go beyond the RPA. While none of these revelations may come as a surprise to the reader, it should serve as a reminder that even our most cherished approximations may not be controlled in any sense of the term. It is more likely that they were the first to appear in the literature and probably the easiest ones to use thus explaining their popularity. The dielectric function written down above has the attractive feature of reducing to the familiar Lindhard dielectric function for extremely weak coupling and at the same time giving us something very different for stronger coupling. Furthermore, if we ask the Luttinger liquid community to write down a dielectric function of the Luttinger liquid, they are in all probability, going to write down the traditional RPA dielectric function\[Neto\]. But we submit that this is a serious mistake. The more nonideal a system is, the more its dielectric function differs from the traditional RPA, even at high density. Parenthetically, we note that having granted the fact that the system is nonideal(namely, no Fermi surface or close to being devoid of one) it makes no difference whether the system is at high or low density, the point is, it is the generalised RPA that is still valid for this system. The only drawback of the above approach is that if we compute the one-particle Green functions we find that the imaginary part vanishes identically. This is unfortunate and we have to do better in order capture lifetime effects. The RPA hamiltonian neglects fluctuations in the momentum distribution of the electrons. In order to recover a finite lifetime of single particle excitations, we find that it is important to study the generalised $`H_0`$ rather than the more simple $`H_{RPA}`$. The fluctuations in the momentum distribution may be related to the mean by the following observation. Define,
$$N(𝐤,𝐤^{^{}})=n_𝐤n_𝐤^{^{}}n_𝐤n_𝐤^{^{}}$$
(13)
. The fluctuation in the number operator is $`N(𝐤,𝐤)=n_𝐤^2n_𝐤^2`$ . Since $`n_𝐤^2=n_𝐤`$ for fermions, we have
$$N(𝐤,𝐤)=n_𝐤(1n_𝐤)$$
(14)
Therefore, we may conclude that any nonideal momentum distribution fluctuates (there are however, pathological exceptions, see footnote). In fact, a very nonideal momentum distribution such as one for which $`n_𝐤=0.5`$ for most momenta has the largest fluctuation. When dealing with nonideal systems, we are obliged to consider fluctuations in the momentum distribution. In order to study lifetime effects therefore, we have to include the full $`H_0`$. This may be written more transparently as (apart from additive constants) $`H_0=H_{RPA}+H_{fl}`$,
$$H_{fl}=\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤}{}\delta n_{𝐤+𝐪/2}\delta n_{𝐤𝐪/2}$$
(15)
The full Fermi propagator may be evaluated by treating the fluctuation part as a perturbation and using the functional methods of Schwinger illustrated brilliantly by Kadanoff and Baym. Before we plunge into the details it is important to keep in mind that the mean also changes when we consider fluctuations. That is,
$$n_𝐤=n_𝐤_{RPA}+n_𝐤_{fl}$$
(16)
here $`n_𝐤_{RPA}`$ is given by Eq.( 9) or Eq.( 11) the rest is nonzero only when the fluctuation in the momentum distribution is large (which is, unfortunately, almost always the case when $`n_𝐤_{RPA}`$ is nonideal) Therefore, now, $`\delta n_𝐤`$ refers to fluctuation around the full average rather than the RPA average. The final answers are given below, and it is hoped that the reader can rederive them using the references quoted in the bibliography (mainly , ). We shall adhere to the notation of Kadanoff and Baym . In their notation the final answers for the single-particle Green functions are as follows(we assume in the following that $`F(𝐩)0`$, however one may investigate the limit $`F(𝐩)0`$, here $`z_n=(2n+1)\pi /\beta `$),
$$G_n(𝐤)=\frac{1}{iz_n\stackrel{~}{ϵ}_𝐤+\mu \mathrm{\Sigma }_n(𝐤)}$$
(17)
and
$$\mathrm{\Sigma }_n(𝐤)=G_n(𝐤)F(𝐤)$$
(18)
$$F(𝐤)=\underset{𝐪,𝐪^{^{}}0}{}\frac{v_𝐪v_𝐪^{^{}}}{V^2}N(𝐤𝐪,𝐤𝐪^{^{}})$$
(19)
From Eq.( 18) we may obtain the real and imaginary parts of the retarded self-energy, and from there the spectral function and the collision rates.
$$\mathrm{\Gamma }(𝐩,\omega )=\sqrt{\kappa (𝐩,\omega )}$$
(20)
Similarly,
$$A(𝐩,\omega )=\frac{\sqrt{\kappa (𝐩,\omega )}}{F(𝐩)}$$
(21)
if $`\kappa (𝐩,\omega )=(\omega \stackrel{~}{ϵ}_𝐩+\mu )^24F(𝐩)<0`$ and both are zero otherwise(when $`F(𝐩)0`$). It can be seen that the spectral function is peaked around $`\stackrel{~}{ϵ}_𝐩\mu `$ with a width of the order of $`2\sqrt{F(𝐩)}`$, and the collision rate is vanishingly small for those values of $`𝐩`$ for which $`F(𝐩)`$ is close to zero. It may be shown that the momentum distribution including possible fluctuations is given by,
$$n_𝐩=(\frac{2}{\pi })_{\pi /2}^{\pi /2}𝑑\theta \text{ }cos^2\theta \text{ }\frac{1}{e^{\beta (\stackrel{~}{ϵ}_𝐩\mu )}e^{2\beta \sqrt{F(𝐩)}sin\theta }+1}$$
(22)
and $`n_𝐩=0`$ for $`F(𝐩)<0`$. Again, it may be seen quite easily that Eq.( 22) is identical to Eq.( 9) when fluctuations in the momentum distribution are ignored. The only sticking point now is the computation of the fluctuation in the momentum distribution. This may be done in a similar manner by employing the functional methods of Kadanoff and Baym. However, we shall defer the computation of this quantity until the next section. Incidentally, the formulas we suggested in our preprints for this quantity in retrospect, are probably rather poor. Instead we shall adopt a more rigorous approach in the next section. The dielectric function is also modified as a result of these fluctuations and the final formula for the dielectric function including possible fluctuations in the momentum distribution is,
$$ϵ_{eff}(𝐪,\omega )=ϵ_{gRPA}(𝐪,\omega )(\frac{v_𝐪}{V})^2\frac{P_2(𝐪,\omega )}{ϵ_{gRPA}(𝐪,\omega )}$$
(23)
Here,
$$P_2(𝐪,\omega )=\underset{𝐤,𝐤^{^{}}}{}\frac{N(𝐤+𝐪/2,𝐤^{^{}}+𝐪/2)N(𝐤𝐪/2,𝐤^{^{}}+𝐪/2)N(𝐤+𝐪/2,𝐤^{^{}}𝐪/2)+N(𝐤𝐪/2,𝐤^{^{}}𝐪/2)}{(\omega \stackrel{~}{ϵ}_{𝐤𝐪/2}+\stackrel{~}{ϵ}_{𝐤+𝐪/2})(\omega \stackrel{~}{ϵ}_{𝐤^{^{}}𝐪/2}+\stackrel{~}{ϵ}_{𝐤^{^{}}+𝐪/2})}$$
(24)
A full derivation of this is relegated to the appendix in order to avoid interupting the flow of ideas. Assuming that $`\omega `$ has a small imaginary part, we recover both the real and imaginary parts of the full dielectric function.
## III Repulsion-Attraction Duality
In this section, we point out an interesting and even a seemingly paradoxical duality between attraction and repulsion. In a homogeneous electron gas, electron repel each other. However, we may show that by virtue of the electron being fermions, an exchange operation on the interaction part of the full hamiltonian leads to a change in sign of the interaction and also a change in the momentum exchanged by the interacting electrons. To see this, let us write down the interaction part of the full hamiltonian again.
$$H_{int}=\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤𝐤^{^{}}}{}c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}c_{𝐤𝐪/2}$$
(25)
We may rearrange the various fermion operators in this interaction so that precisely the same quantity may be rewritten as,
$$H_{int}=\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{2V}\underset{𝐪0}{}c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}c_{𝐤𝐪/2}$$
(26)
Two important changes have occured. First there is the change in sign. This is the reason for the term repulsion-attraction duality. The second is the momentum that is exchanged is different. The momentum carried away by the virtual photon is no longer $`𝐪`$ but is $`𝐤𝐤^{^{}}`$. The change in sign is peculiar to fermions, whereas the other feature is present for bosons too. Let us now use this new form and recompute the dielectric function. Actually, we could have done the same shuffle for the $`𝐤=𝐤^{^{}}`$ part of the interaction. But we shall relegate a more careful examination of these issues to future publications. The main purpose of the present article is to lay before the reader a scheme that is robust, rich and sufficiently general so that she may apply the ideas to other more practical problems. In order to compute the dielectric function suggested by this shuffled interaction let us proceed as follows. The full hamiltonian may be written as ($`n_0(𝐤)=n_𝐤=c_𝐤^{}c_𝐤`$),
$`H={\displaystyle \underset{𝐤}{}}\stackrel{~}{ϵ}_𝐤n_0(𝐤){\displaystyle \underset{𝐪0}{}}{\displaystyle \frac{v(𝐪)}{2V}}{\displaystyle \underset{𝐤}{}}\delta n_0(𝐤+𝐪/2)\text{ }\delta n_0(𝐤𝐪/2)`$
$$\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{2V}\underset{𝐪0}{}n_𝐪(𝐤)n_𝐪(𝐤^{^{}})$$
(27)
here $`n_𝐪(𝐤)=c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2}`$ Let us now apply an external field,
$$H_{ext}(t)=\underset{𝐤,𝐪}{}(U_{ext}(𝐪,t)+U_{ext}^{}(𝐪,t))n_𝐪(𝐤)$$
(28)
In order to proceed, we appeal to the random-phase approximation. In a suitably generalised sense, it may be defined to mean the following approximation :
$`[n_𝐪(𝐤),n_𝐪^{^{}}(𝐤^{^{}})]=c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}\delta _{𝐤𝐪/2,𝐤^{^{}}+𝐪^{^{}}/2}c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤𝐪/2}\delta _{𝐤^{^{}}𝐪^{^{}}/2,𝐤+𝐪/2}`$
$$\delta _{𝐤,𝐤^{^{}}}\delta _{𝐪,𝐪^{^{}}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(29)
$$[n_𝐪(𝐤),n_0(𝐩)]=n_𝐪(𝐤)(\delta _{𝐩,𝐤𝐪/2}\delta _{𝐩,𝐤+𝐪/2})$$
(30)
Here we have to retain $`n_0(𝐤)`$ as an operator in order to include effects due to fluctuations in the momentum distribution. Let us now write down the equation of motion of this operator.
$`i{\displaystyle \frac{}{t}}n_𝐪(𝐤)=(\stackrel{~}{ϵ}_{𝐤𝐪/2}\stackrel{~}{ϵ}_{𝐤+𝐪/2})n_𝐪(𝐤){\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v(𝐪^{^{}})}{V}}(\delta n_0(𝐤𝐪/2𝐪^{^{}})n_𝐪(𝐤)`$
$`\delta n_0(𝐤+𝐪/2𝐪^{^{}})n_𝐪(𝐤)){\displaystyle }_{𝐤𝐤^{^{}}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(n_0(𝐤+𝐪/2)n_𝐪(𝐤^{^{}})n_0(𝐤𝐪/2)n_𝐪(𝐤^{^{}}))`$
$$+(U_{ext}(𝐪,t)+U_{ext}^{}(𝐪,t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(31)
Write $`n_0(𝐤)=n_0(𝐤)+\delta \text{ }n_0(𝐤)`$
$`i{\displaystyle \frac{}{t}}n_𝐪(𝐤)=(\stackrel{~}{ϵ}_{𝐤𝐪/2}\stackrel{~}{ϵ}_{𝐤+𝐪/2})n_𝐪(𝐤)`$
$`{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v(𝐪^{^{}})}{V}}(\delta n_0(𝐤𝐪/2𝐪^{^{}})n_𝐪(𝐤)\delta n_0(𝐤+𝐪/2𝐪^{^{}})n_𝐪(𝐤))`$
$`{\displaystyle \underset{𝐤𝐤^{^{}}}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))n_𝐪(𝐤^{^{}})`$
$`{\displaystyle \underset{𝐤𝐤^{^{}}}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(\delta n_0(𝐤+𝐪/2)n_𝐪(𝐤^{^{}})\delta n_0(𝐤𝐪/2)n_𝐪(𝐤^{^{}}))`$
$$+(U_{ext}(𝐪,t)+U_{ext}^{}(𝐪,t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(32)
$`i{\displaystyle \frac{}{t}}\delta n_0(𝐩)n_𝐪(𝐤)=(\stackrel{~}{ϵ}_{𝐤𝐪/2}\stackrel{~}{ϵ}_{𝐤+𝐪/2})\delta n_0(𝐩)n_𝐪(𝐤)`$
$`{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v(𝐪^{^{}})}{V}}(\delta n_0(𝐩)\delta n_0(𝐤𝐪/2𝐪^{^{}})\delta n_0(𝐩)\delta n_0(𝐤+𝐪/2𝐪^{^{}}))n_𝐪(𝐤)`$
$`{\displaystyle \underset{𝐤𝐤^{^{}}}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))\delta n_0(𝐩)n_𝐪(𝐤^{^{}})`$
$`{\displaystyle \underset{𝐤𝐤^{^{}}}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(\delta n_0(𝐩)\delta n_0(𝐤+𝐪/2)\delta n_0(𝐩)\delta n_0(𝐤𝐪/2))n_𝐪(𝐤^{^{}})`$
$$+(U_{ext}(𝐪,t)+U_{ext}^{}(𝐪,t))(\delta n_0(𝐩)\delta n_0(𝐤+𝐪/2)\delta n_0(𝐩)\delta n_0(𝐤𝐪/2))$$
(33)
Assuming that,
$$U_{ext}(𝐪,t)=U_{ext}(𝐪,0)e^{i\omega \text{ }t}$$
(34)
$`(\omega \stackrel{~}{ϵ}_{𝐤𝐪/2}+\stackrel{~}{ϵ}_{𝐤+𝐪/2})\text{ }\delta \text{ }n_0(𝐩)\text{ }n_𝐪(𝐤)={\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(N(𝐩,𝐤𝐪/2𝐪^{^{}})N(𝐩,𝐤+𝐪/2𝐪^{^{}}))n_𝐪(𝐤)`$
$`(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)){\displaystyle \underset{𝐤^{^{}}𝐤}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}\delta \text{ }n_0(𝐩)n_𝐪(𝐤^{^{}})`$
$`(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2)){\displaystyle \underset{𝐤^{^{}}𝐤}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}n_𝐪(𝐤^{^{}})`$
$$+U_{ext}(𝐪,0)(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2))$$
(35)
$`(\omega \stackrel{~}{ϵ}_{𝐤𝐪/2}+\stackrel{~}{ϵ}_{𝐤+𝐪/2})n_𝐪(𝐤)={\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(\delta n_0(𝐤𝐪/2𝐪^{^{}})n_𝐪(𝐤)\delta n_0(𝐤+𝐪/2𝐪^{^{}})n_𝐪(𝐤))`$
$`(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)){\displaystyle \underset{𝐤^{^{}}𝐤}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}n_𝐪(𝐤^{^{}})`$
$`{\displaystyle \underset{𝐤^{^{}}𝐤}{}}{\displaystyle \frac{v_{𝐤𝐤^{^{}}}}{V}}(\delta n_0(𝐤+𝐪/2)n_𝐪(𝐤^{^{}})\delta n_0(𝐤𝐪/2)n_𝐪(𝐤^{^{}}))`$
$$+U_{ext}(𝐪,0)(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(36)
In order to simplify this further, let us assume that we are in the weakly nonideal regime. That is, it is legitimate to treat the momentum distribution as possessing a sharp Fermi surface and no other striking features. Let us now define,
$$\stackrel{~}{n}_𝐪(\stackrel{}{r})=\frac{1}{V}\underset{𝐤}{}e^{i𝐤.\stackrel{}{r}}n_𝐪(𝐤)$$
(37)
$`(\omega +i{\displaystyle \frac{𝐪.}{m}})\text{ }\delta \text{ }n_0(𝐩)\text{ }\stackrel{~}{n}_𝐪(\stackrel{}{r})={\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤,𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})\text{ }e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(N(𝐩,𝐤𝐪/2𝐪^{^{}})N(𝐩,𝐤+𝐪/2𝐪^{^{}}))\text{ }\stackrel{~}{n}_𝐪(\stackrel{}{r}^{^{}})}`$
$`{\displaystyle \frac{1}{V}}{\displaystyle d^3r^{^{}}\text{ }\underset{𝐤}{}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))v(r^{^{}})\delta \text{ }n_0(𝐩)n_𝐪(\stackrel{}{r}^{^{}})}`$
$`{\displaystyle \frac{1}{V}}{\displaystyle d^3r^{^{}}\text{ }\underset{𝐤}{}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2))v(r^{^{}})n_𝐪(\stackrel{}{r}^{^{}})}`$
$$+U_{ext}(𝐪,0)\frac{1}{V}\underset{𝐤}{}e^{i𝐤.\stackrel{}{r}}(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2))$$
(38)
$`(\omega +i{\displaystyle \frac{𝐪.}{m}})\stackrel{~}{n}_𝐪(\stackrel{}{r})={\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤,𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(\delta n_0(𝐤𝐪/2𝐪^{^{}})n_𝐪(𝐤)\delta n_0(𝐤+𝐪/2𝐪^{^{}})n_𝐪(\stackrel{}{r}^{^{}}))`$
$`{\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤}{}}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))n_𝐪(\stackrel{}{r}^{^{}})}`$
$`{\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤}{}}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})(\delta n_0(𝐤+𝐪/2)n_𝐪(\stackrel{}{r}^{^{}})\delta n_0(𝐤𝐪/2)n_𝐪(\stackrel{}{r}^{^{}}))}`$
$$+U_{ext}(𝐪,0)\frac{1}{V}\underset{𝐤}{}e^{i𝐤.\stackrel{}{r}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(39)
Let us now compute the following quantity,
$$f_𝐪(\stackrel{}{R})=\frac{1}{V}\underset{𝐤}{}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(40)
where $`\stackrel{}{R}=\stackrel{}{r}^{^{}}\stackrel{}{r}`$. If $`k_f`$ is sufficiently large ( $`k_f>>q,k_f>>1/R`$) we may write,
$$f_𝐪(\stackrel{}{R})C_0𝐪._R\delta ^3(R)$$
(41)
$$C_0=\frac{(4\pi )^2}{(2\pi )^3}(\frac{i}{q})_0^{\mathrm{}}𝑑R\text{ }\frac{2^3}{q^2R^2}(sin(k_fR)(k_fR)cos(k_fR))(sin(\frac{qR}{2})(\frac{qR}{2})cos(\frac{qR}{2}))$$
(42)
$`(\omega +i{\displaystyle \frac{𝐪.}{m}})\delta n_0(𝐩)\text{ }\stackrel{~}{n}_𝐪(\stackrel{}{r})={\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤,𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})\text{ }e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(N(𝐩,𝐤𝐪/2𝐪^{^{}})N(𝐩,𝐤+𝐪/2𝐪^{^{}}))\text{ }\stackrel{~}{n}_𝐪(\stackrel{}{r}^{^{}})}`$
$`+C_0𝐪._\stackrel{}{r}v(r)\delta n_0(𝐩)\stackrel{~}{n}_𝐪(\stackrel{}{r})`$
$`{\displaystyle \frac{1}{V}}{\displaystyle d^3r^{^{}}\text{ }\underset{𝐤}{}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2))v(r^{^{}})n_𝐪(\stackrel{}{r}^{^{}})}`$
$$+U_{ext}(𝐪,0)\frac{1}{V}\underset{𝐤}{}e^{i𝐤.\stackrel{}{r}}(N(𝐩,𝐤+𝐪/2)N(𝐩,𝐤𝐪/2))$$
(43)
$`(\omega +i{\displaystyle \frac{𝐪.}{m}})\stackrel{~}{n}_𝐪(\stackrel{}{r})={\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤,𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}(\delta n_0(𝐤𝐪/2𝐪^{^{}})n_𝐪(\stackrel{}{r}^{^{}})\delta n_0(𝐤+𝐪/2𝐪^{^{}})n_𝐪(\stackrel{}{r}^{^{}}))}`$
$`+C_0𝐪._\stackrel{}{r}v(r)\stackrel{~}{n}_𝐪(\stackrel{}{r})`$
$`{\displaystyle \frac{1}{V}}{\displaystyle \underset{𝐤}{}}e^{i𝐤.(\stackrel{}{r}^{^{}}\stackrel{}{r})}{\displaystyle d^3r^{^{}}\text{ }v(r^{^{}})(\delta n_0(𝐤+𝐪/2)n_𝐪(\stackrel{}{r}^{^{}})\delta n_0(𝐤𝐪/2)n_𝐪(\stackrel{}{r}^{^{}}))}`$
$$+U_{ext}(𝐪,0)\frac{1}{V}\underset{𝐤}{}e^{i𝐤.\stackrel{}{r}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(44)
At this stage the author is unable to complete this calculation, but hopefully the reader appreciates the spirit of the discussion. The author apologises in advance for this omission.
## IV Improved Theory of Single-Particle Properties Using Bosonization
In this section, we use the sea-boson method suitably generalised to accomodate fluctuations in the momentum distribution to compute single-particle properties. In the appendix we show how to derive the dielectric function using this method. The dielectric function may be evaluated by more conventional means as well . However, the single-particle properties evaluated using the sea-boson method is more accurate as one is able to treat the problem more systematically and the physical interpretation of the formulas is also more transparent. The really systematic approach would be to write down the sea-boson correspondence suitably generalised to accomodate fluctuations in the momentum distribution and evaluate the various propagators like we did in our earlier work. However, we shall adopt a simpler but hopefully the correct approach here. The idea is to simply borrow from our earlier work except that the zeros of the dielectric function are no longer the zeros of the RPA dielectric function but that of $`ϵ_{eff}`$ and we need to interpret the derivate of the dielectric funcion with respect to frequency that appears in these formulas as being the derivative of the full dielectric function. Therefore, let us write down the various formulas. The full hamiltonian in diagonalised form has the following appearence,
$$H_{full}=\underset{𝐪,i}{}\omega _i(𝐪)d_i^{}(𝐪)d_i(𝐪)$$
(45)
$$\overline{n}_𝐤=n^\beta (𝐤)\text{ }F_1(𝐤)+(1n^\beta (𝐤))\text{ }F_2(𝐤)$$
(46)
where,
$$F_1(𝐤)=\frac{1}{1+\frac{S_B(𝐤)}{1+S_A(𝐤)}}$$
(47)
$$F_2(𝐤)=\frac{1}{1+\frac{1+S_B(𝐤)}{S_A(𝐤)}}$$
(48)
$$S_A(𝐤)=\underset{𝐪,i}{}\frac{\overline{n}_{𝐤𝐪}}{(\omega _i(𝐪)+𝐤.𝐪/mϵ_𝐪)^2}g_i^2(𝐪)$$
(49)
$$S_B(𝐤)=\underset{𝐪,i}{}\frac{1\overline{n}_{𝐤+𝐪}}{(\omega _i(𝐪)+𝐤.𝐪/m+ϵ_𝐪)^2}g_i^2(𝐪)$$
(50)
$$g_i(𝐪)=[\underset{𝐤}{}\frac{\overline{n}_{𝐤𝐪/2}\overline{n}_{𝐤+𝐪/2}}{(\omega _i(𝐪)\frac{𝐤.𝐪}{m})^2}]^{\frac{1}{2}}$$
(51)
It may be observed that this quantity is just the frequency derivative of the polarization. In other words,
$$g_i^2(𝐪)=\frac{V}{v_𝐪}(\frac{}{\omega })_{\omega =\omega _i(𝐪)}ϵ_{gRPA}^P(𝐪,\omega )$$
(52)
By analogy we may generalise this so that when fluctuations are introduced,
$$g_i^2(𝐪)=\frac{V}{v_𝐪}(\frac{}{\omega })_{\omega =\omega _i(𝐪)}ϵ_{eff}^P(𝐪,\omega )$$
(53)
We shall employ this latter definition in our analysis. $`\omega _i(𝐪)`$ is the zero of the overall dielectric function,
$$ϵ_{eff}^P(𝐪,\omega _i)=0$$
(54)
It is worthwhile simplifying the effective dielectric function.
$$ϵ_{eff}(𝐪,\omega )=ϵ_{gRPA}(𝐪,\omega )+\frac{(ϵ_{gRPA}(𝐪,\omega )1)(ϵ_{gRPA}(𝐪,\omega )ϵ_\beta (𝐪,\omega ))}{ϵ_{gRPA}(𝐪,\omega )}$$
(55)
The principal part of this is given by,
$$ϵ_{eff}^P(𝐪,\omega )=ϵ_{gRPA}^P(𝐪,\omega )+\frac{(ϵ_{gRPA}^P(𝐪,\omega )1)(ϵ_{gRPA}^P(𝐪,\omega )ϵ_\beta ^P(𝐪,\omega ))}{ϵ_{gRPA}^P(𝐪,\omega )}$$
(56)
where,
$$ϵ_{gRPA}(𝐪,\omega )=1+\frac{v_𝐪}{V}\underset{𝐤}{}\frac{\overline{n}_{𝐤+𝐪/2}\overline{n}_{𝐤𝐪/2}}{\omega \frac{𝐤.𝐪}{m}}$$
(57)
$$ϵ_\beta (𝐪,\omega )=1+\frac{v_𝐪}{V}\underset{𝐤}{}\frac{n_{𝐤+𝐪/2}^\beta n_{𝐤𝐪/2}^\beta }{\omega \frac{𝐤.𝐪}{m}}$$
(58)
and the principal parts of these functions are just the real parts.
$$ϵ_{gRPA}^P(𝐪,\omega )=1+\frac{v_𝐪}{V}\underset{𝐤}{}\text{ }P\text{ }\frac{\overline{n}_{𝐤+𝐪/2}\overline{n}_{𝐤𝐪/2}}{\omega \frac{𝐤.𝐪}{m}}$$
(59)
### A Just How Many Zeros Does the Dielectric Function Really Have ?
In this section, we address what is perhaps the most vexing problem in this approach. What is the size of the set to which $`i`$ in Eq.( 45) belongs ? If one counts only the collective mode which is what a naive approach would lead us to do, then we are ignoring a large (infinite !) number of particle-hole modes. And yet, it seems that the particle-hole modes don’t come about naturally and must be forced into the formalism. It is really important to address this issue since ignoring it would mean that the excitation spectrum of the homogeneous electron gas possesses a gap when in fact it should not. In other words, if one counts only the collective mode(plasmon) one arrives at the unavoidable conclusion that the excitation spectrum has a finite gap(equal to the plasmon energy). How then should one introduce the particle-hole mode ? In this section, we show how to introduce both the particle-hole mode and collective mode in a unified manner. In our previous work we suggested that the zeros of the RPA dielectric function should be interpreted as the maxima of the dynamical structure factor. This is quite an unusual and drastic departure from the notion of a root of a function. Being physicists we accept it as it has a physical interpretation. Let us examine the diagonalised form of the full hamiltonian:
$$H_{full}=\underset{𝐪,i}{}\omega _i(𝐪)d_i^{}(𝐪)d_i(𝐪)$$
(60)
For small $`𝐪`$, we expect this sum to involve just the collective mode. In other words,
$$H_{full}=\underset{𝐪=small}{}\omega _c(𝐪)d_c^{}(𝐪)d_c(𝐪)$$
(61)
This may be achieved by employing the following device,
$$H_{full}=\underset{𝐪}{}_0^{\mathrm{}}\text{ }𝑑\omega \text{ }\omega \text{ }W(𝐪,\omega )\text{ }d_\omega ^{}(𝐪)d_\omega (𝐪)$$
(62)
Now these operators, the dressed sea-bosons in the new language obey a different kind of commutation rule:
$$[d_\omega (𝐪),d_\omega ^{^{}}(𝐪^{^{}})]=0$$
(63)
$$[d_\omega (𝐪),d_\omega ^{^{}}^{}(𝐪^{^{}})]=\delta _{\omega ,\omega ^{^{}}}\delta _{𝐪,𝐪^{^{}}}$$
(64)
where $`\delta _{\omega ,\omega ^{^{}}}`$ is a Kronecker delta function. Also the object $`W(𝐪,\omega )`$ is a weight suitably chosen so that the the sum reduces to just the one over the collective mode in the small $`𝐪`$ limit. It is also clear that $`W(𝐪,\omega )`$ has to have dimensions of inverse energy. Let us therefore postulate a form,
$$W(𝐪,\omega )=c_0\text{ }Im(\frac{1}{ϵ(𝐪,\omega )})$$
(65)
where $`c_0`$ is a suitable constant. We know from textbooks that(in 3D),
$$Limit_{𝐪0}W(𝐪,\omega )=c_0\text{ }Limit_{𝐪0}\text{ }Im(\frac{1}{ϵ(𝐪,\omega )})=\frac{c_0\pi \omega _p}{2}\delta (\omega \omega _p)$$
(66)
In order that we reproduce the right collective mode we must choose,
$$c_0=\frac{2}{\pi \omega _p}$$
(67)
The choices for $`W(𝐪,\omega )`$ in other dimensions are worked out in the appendix. Now we have in our hands a convenient way of labeling excited states of our system. The eigenstates of the system have energy eigenvalues given by,
$$\mathrm{\Omega }_\omega (𝐪)=\omega \text{ }W(𝐪,\omega )\text{ }\mathrm{\Delta }\omega $$
(68)
The spacing $`\mathrm{\Delta }\omega `$ is the smallest possible spacing between the energies, which is arbitrarily small. Thus we can see that, in most cases it does not cost a finite energy to excite the system. This corresponds to the particle-hole mode. On the other hand, if the weight $`W(𝐪,\omega )`$ is singular as it happens in the vicinity of $`\omega =\omega _p`$and $`𝐪=0`$, we have a finite gap.
$$\mathrm{\Omega }_{\omega \omega _p}(𝐪0)=\omega _p\text{ }\delta (\omega \omega _p)\text{ }\mathrm{\Delta }\omega $$
(69)
Now if $`Lim_{\omega \omega _p}\delta (\omega \omega _p)\text{ }\mathrm{\Delta }\omega =1`$ (this defines $`\mathrm{\Delta }\omega `$ if you like), then we see the emergence of the collective mode. In gapped systems we expect only the collective mode and no particle hole mode. This is possible only if for any $`\omega 0`$,
$$W(𝐪,\omega )=\mathrm{}$$
(70)
For the homogeneous Fermi system, this situation corresponds to Wigner crystallisation. In a Wigner crystal, in order to create excited states, you need to create phonons which require a nonzero-amount of energy each time.
Thus, all energies are allowed but each comes with a ”weight” corresponding to how strong the structure factor is at that energy. It may be seen that in 3-dimensions this would mean that for small $`𝐪`$ the sum over $`i`$ is just the collective mode, but for larger $`𝐪`$, we start summing the particle hole modes as well. The sums in the objects $`S_A`$ and $`S_B`$ have to be reinterpreted as well. It is rather unfortunate that simple and straightforward interpretations are not possible, and one is forced to take recourse to devious means such as the one we shall now describe. Let us postulate,
$$S_A(𝐤)=\underset{𝐪}{}_0^{\mathrm{}}\text{ }𝑑\omega \text{ }\stackrel{~}{W}(𝐪,\omega )\text{ }\frac{\overline{n}_{𝐤+𝐪}}{(\omega 𝐤.𝐪/mϵ_𝐪)^2}g_\omega ^2(𝐪)$$
(71)
$$S_B(𝐤)=\underset{𝐪}{}_0^{\mathrm{}}\text{ }𝑑\omega \text{ }\stackrel{~}{W}(𝐪,\omega )\text{ }\frac{1\overline{n}_{𝐤𝐪}}{(\omega 𝐤.𝐪/m+ϵ_𝐪)^2}g_\omega ^2(𝐪)$$
(72)
$$g_\omega ^2(𝐪)=\frac{v_𝐪}{V}\frac{1}{\frac{}{\omega }ϵ_{eff}^P(𝐪,\omega )}$$
(73)
and,
$$\stackrel{~}{W}(𝐪,\omega )\text{ }=W(𝐪,\omega )/_0^{\mathrm{}}\text{ }𝑑\omega \text{ }W(𝐪,\omega )$$
(74)
This seemingly adhoc ansatz becomes more plausible when one realises that in the small $`𝐪`$ limit it has a familiar form corresponding to the collective mode.
$$S_A(𝐤)\underset{𝐪=small}{}\frac{\overline{n}_{𝐤+𝐪}}{(\omega _c(𝐪)𝐤.𝐪/mϵ_𝐪)^2}g_c^2(𝐪)$$
(75)
$$S_B(𝐤)\underset{𝐪=small}{}\frac{1\overline{n}_{𝐤𝐪}}{(\omega _c(𝐪)𝐤.𝐪/m+ϵ_𝐪)^2}g_c^2(𝐪)$$
(76)
Let us now move on to the full propagator.
### B The Full Propagator
Let us borrow from our earlier work and write down the final formulas,
$$\psi ^{}(𝐱,t)\psi (𝐱^{^{}},t^{^{}})=|_0|^2|𝒵_0|^4e^{_{𝐤,𝐪,i}U_{𝐤,𝐪}^i(𝐱)U_{𝐤,𝐪}^i(𝐱^{^{}})e^{i\text{ }\stackrel{~}{\omega }_i(𝐪)(t^{^{}}t)}}e^{_{𝐤,𝐪}g_{𝐤,𝐪}^{}(𝐱)g_{𝐤,𝐪}(𝐱^{^{}})e^{i\text{ }\omega _𝐤(𝐪)(t^{^{}}t)}}\psi ^{}(𝐱,t)\psi (𝐱^{^{}},t^{^{}})_{free}$$
(77)
$$\psi (𝐱^{^{}},t^{^{}})\psi ^{}(𝐱,t)=|_0|^2|𝒵_0|^4e^{_{𝐤,𝐪,i}U_{𝐤,𝐪}^i(𝐱^{^{}})U_{𝐤,𝐪}^i(𝐱)e^{i\text{ }\stackrel{~}{\omega }_i(𝐪)(tt^{^{}})}}e^{_{𝐤,𝐪}f_{𝐤,𝐪}^{}(𝐱)f_{𝐤,𝐪}(𝐱^{^{}})e^{i\text{ }\omega _𝐤(𝐪)(tt^{^{}})}}\psi (𝐱^{^{}},t^{^{}})\psi ^{}(𝐱,t)_{free}$$
(78)
In the above formula, the index $`i`$ runs over both the collective mode as well as the particle-hole modes($`i=c,𝐤_i`$). As we have pointed out earlier, this must be reinterpreted to mean a weighted sum with the dynamical structure factor suitably normalised appearing as the weight.
$$g_{𝐤,𝐪}(𝐱)=e^{i\text{ }𝐪.𝐱}(\frac{1}{2\text{ }N\text{ }ϵ_𝐪})\mathrm{\Lambda }_𝐤^0(𝐪)\omega _𝐤(𝐪)+i\text{ }U_𝐪(𝐱)\mathrm{\Lambda }_𝐤^0(𝐪)=f_{𝐤,𝐪}^{}(𝐱)$$
(79)
Set $`U_𝐪(𝐱)=e^{i\text{ }𝐪.𝐱}U_0(𝐪)`$
$$U_0(𝐪)=\frac{1}{N}(\frac{\theta (k_f|𝐪|)w_1(𝐪)}{w_2(𝐪)})^{\frac{1}{2}}$$
(80)
$$w_1(𝐪)=(\frac{1}{4\text{ }N\text{ }ϵ_𝐪^2})\underset{𝐤}{}(\frac{𝐤.𝐪}{m})^2(\mathrm{\Lambda }_𝐤^0(𝐪))^2$$
(81)
$$w_2(𝐪)=(\frac{1}{N})\underset{𝐤}{}(\mathrm{\Lambda }_𝐤^0(𝐪))^2$$
(82)
Here $`\mathrm{\Lambda }_𝐤^0(𝐪)`$ is the $`\mathrm{\Lambda }`$ with a noninteracting momentum distribution. The interacting coefficients are given as,
$$U_{𝐤,𝐪}^i(𝐱)=f_{𝐤,𝐪}^{}(𝐱)[a_𝐤(𝐪),b_i^{}(𝐪)]+f_{𝐤,𝐪}(𝐱)[a_𝐤(𝐪),b_i(𝐪)]$$
(83)
$`_0=exp({\displaystyle \underset{𝐤,𝐪,i}{}}f_{𝐤,𝐪}^{}(𝐱)f_{𝐤,𝐪}(𝐱)[b_i(𝐪),a_𝐤^{}(𝐪)][a_𝐤(𝐪),b_i^{}(𝐪)])`$
$`\times exp({\displaystyle \frac{1}{2}}{\displaystyle \underset{𝐤,𝐪,i}{}}f_{𝐤,𝐪}^{}(𝐱)f_{𝐤,𝐪}^{}(𝐱)[a_𝐤(𝐪),b_i(𝐪)][a_𝐤(𝐪),b_i^{}(𝐪)])`$
$$\times exp(\frac{1}{2}\underset{𝐤,𝐪,i}{}f_{𝐤,𝐪}(𝐱)f_{𝐤,𝐪}(𝐱)[a_𝐤(𝐪),b_i(𝐪)][a_𝐤(𝐪),b_i^{}(𝐪)])$$
(84)
$$𝒵_0=e^{i\text{ }_{𝐤,𝐪0}U_0(𝐪)(\frac{1}{2\text{ }Nϵ_𝐪})(\mathrm{\Lambda }_𝐤^0(𝐪))^2\omega _𝐤(𝐪)}e^{\frac{1}{2}\text{ }_{𝐤,𝐪0}(\frac{1}{2\text{ }Nϵ_𝐪})^2(\mathrm{\Lambda }_𝐤^0(𝐪))^2(\omega _𝐤(𝐪))^2}e^{\frac{1}{2}\text{ }_{𝐤,𝐪0}(U_0(𝐪))^2(\mathrm{\Lambda }_𝐤^0(𝐪))^2}$$
(85)
The commutators are given as before,
$$[b_i(𝐪),a_𝐤^{}(𝐪)]=(\frac{\mathrm{\Lambda }_𝐤(𝐪)}{\stackrel{~}{\omega }_i(𝐪)\frac{𝐤.𝐪}{m}})g_i(𝐪)=[a_𝐤(𝐪),b_i^{}(𝐪)]$$
(86)
$$[b_i(𝐪),a_𝐤(𝐪)]=(\frac{\mathrm{\Lambda }_𝐤(𝐪)}{\stackrel{~}{\omega }_i(𝐪)\frac{𝐤.𝐪}{m}})g_i(𝐪)$$
(87)
$$[a_𝐤(𝐪),b_i(𝐪)]=(\frac{\mathrm{\Lambda }_𝐤(𝐪)}{\stackrel{~}{\omega }_i(𝐪)+\frac{𝐤.𝐪}{m}})g_i(𝐪)$$
(88)
$$g_i^2(𝐪)=\frac{v_𝐪}{V}\frac{1}{\frac{}{\omega }|_{\omega =\omega _i}ϵ_{eff}^P(𝐪,\omega )}$$
(89)
### C Exchange Effects within RPA
It may be possible to employ a myriad of other similar approaches each differing from the other in how the notion of RPA is implemented. To give an example, let us rewrite the full hamiltonian differently.
$`H={\displaystyle \underset{𝐤}{}}ϵ_𝐤c_𝐤^{}c_𝐤{\displaystyle \underset{𝐪0}{}}{\displaystyle \frac{v(𝐪)}{2V}}{\displaystyle \underset{𝐤}{}}n_{𝐤+𝐪/2}n_{𝐤𝐪/2}`$
$$+\underset{𝐪0}{}\frac{v(𝐪)}{2V}\underset{𝐤𝐤^{^{}}}{}n_𝐪(𝐤)n_𝐪(𝐤^{^{}})$$
(90)
Here the exchange term has been singled out for special treatment. We could rewrite the last term differently,
$`H={\displaystyle \underset{𝐤}{}}ϵ_𝐤c_𝐤^{}c_𝐤{\displaystyle \underset{𝐪0}{}}{\displaystyle \frac{v(𝐪)}{2V}}{\displaystyle \underset{𝐤}{}}n_{𝐤+𝐪/2}n_{𝐤𝐪/2}`$
$$\underset{𝐪0}{}\frac{v(𝐪)}{2V}\underset{𝐤𝐤^{^{}}}{}n_{𝐤𝐤^{^{}}}(𝐤/2+𝐤^{^{}}/2+𝐪/2)n_{𝐤^{^{}}𝐤}(𝐤/2+𝐤^{^{}}/2𝐪/2)$$
(91)
This may be interpreted as an exchange term. The negative sign suggests precisely this. These two hamiltonians lead to different answers when the RPA is carried out on them. In fact, even the exchange term involving just the number operators can be treated in two different ways. One way is to retain it as it is, the other is to rewrite it as,
$$H_{ex}=\underset{𝐤}{}\stackrel{~}{ϵ}_𝐤c_𝐤^{}c_𝐤\underset{𝐪0}{}\frac{v(𝐪)}{2V}\underset{𝐤}{}\delta \text{ }n_{𝐤+𝐪/2}\text{ }\delta \text{ }n_{𝐤𝐪/2}$$
(92)
The effective dispersion includes the exchange energy,
$$\stackrel{~}{ϵ}_𝐤=ϵ_𝐤\underset{𝐪0}{}\frac{v(𝐪)}{V}n_{𝐤𝐪}$$
(93)
and $`\delta \text{ }n_𝐤=n_𝐤\overline{n}_𝐤`$ Again when the RPA is carried out on them they yield different answers. One has to examine this issue more closely. In particular the following questions spring to mind. Should we use one or the other or a combination of the two ? The answer may be given by appealing to our bosonization scheme. For more details, the reader is refered to our earlier works ,. Just to make this article self-contained, it is desirable to fix some notation.
$$n_0(𝐤)=c_𝐤^{}c_𝐤$$
(94)
the $`a_𝐤(𝐪)`$ is a sea-boson corresponding to the fermions $`c_𝐤`$. The coefficients $`\mathrm{\Lambda }_𝐤(𝐪)=\sqrt{\overline{n}_{𝐤+𝐪/2}(1\overline{n}_{𝐤𝐪/2})}`$ are c-numbers for now.
$$c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2}=\mathrm{\Lambda }_𝐤(𝐪)a_𝐤(𝐪)+a_𝐤^{}(𝐪)\mathrm{\Lambda }_𝐤(𝐪)$$
(95)
Let us now try and figure out how to decompose products of four fermions, (let us further assume that none of the four indices are equal to any other index)($`𝐪0`$ and $`𝐤𝐤^{^{}}`$)
$$c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}=[\mathrm{\Lambda }_𝐤(𝐪)a_𝐤(𝐪)+a_𝐤^{}(𝐪)\mathrm{\Lambda }_𝐤(𝐪)][\mathrm{\Lambda }_𝐤^{^{}}(𝐪)a_𝐤^{^{}}(𝐪)+a_𝐤^{^{}}^{}(𝐪)\mathrm{\Lambda }_𝐤^{^{}}(𝐪)]$$
(96)
the above equality is within RPA. It may be noted that a slightly different regrouping yields a totally different formula,
$`c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}=c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤𝐪/2}`$
$`=[\mathrm{\Lambda }_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}+𝐪/2}(𝐤𝐤^{^{}})a_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}+\frac{𝐪}{2}}(𝐤^{^{}}𝐤)+a_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}+\frac{𝐪}{2}}^{}(𝐤𝐤^{^{}})\mathrm{\Lambda }_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}+\frac{𝐪}{2}}(𝐤^{^{}}𝐤)]`$
$$\times [\mathrm{\Lambda }_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}\frac{𝐪}{2}}(𝐤^{^{}}𝐤)a_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}\frac{𝐪}{2}}(𝐤𝐤^{^{}})+a_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}\frac{𝐪}{2}}^{}(𝐤^{^{}}𝐤)\mathrm{\Lambda }_{\frac{𝐤}{2}+\frac{𝐤^{^{}}}{2}\frac{𝐪}{2}}(𝐤𝐤^{^{}})]$$
(97)
Let us now examine the commutator obtained using Fermi algebra,
$`[c_{𝐤𝐪/2}^{}c_{𝐤+𝐪/2},c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2}c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}]=[n_0(𝐤𝐪/2)n_0(𝐤+𝐪/2)]c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}`$
$$+[\delta _{𝐤^{^{}}𝐪/2,𝐤+𝐪/2}(1n_0(𝐤𝐪/2))c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}+𝐪/2}+\delta _{𝐤𝐪/2,𝐤^{^{}}+𝐪/2}n_0(𝐤+𝐪/2)c_{𝐤^{^{}}𝐪/2}^{}c_{𝐤𝐪/2}]$$
(98)
It is clear that if one uses the ansatz in Eq.( 96) then one obtains only the first term in Eq.( 98). In order to obtain the second term it is necessary to go beyond this and try and include some portion of Eq.( 97). Let us first try to use the ansatz in Eq.( 97) while completely ignoring the ansatz of Eq.( 96) In this case the interaction term takes a different form,
$$H_I=\underset{𝐪0}{}\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{2V}[\mathrm{\Lambda }_𝐤(𝐪)a_𝐤(𝐪)+a_𝐤^{}(𝐪)\mathrm{\Lambda }_𝐤(𝐪)][\mathrm{\Lambda }_𝐤^{^{}}(𝐪)a_𝐤^{^{}}(𝐪)+a_𝐤^{^{}}^{}(𝐪)\mathrm{\Lambda }_𝐤^{^{}}(𝐪)]$$
(99)
Here two important changes have occured. First, the interaction $`v_𝐪`$ is replaced by $`v_{𝐤𝐤^{^{}}}`$ and there is a change in sign. The physical meaning of these mysterious changes will be defered until later when we have investigated the details. Let us now write down the full hamiltonian,
$$H=\underset{𝐤,𝐪}{}\stackrel{~}{\omega }_𝐤(𝐪)a_𝐤^{}(𝐪)a_𝐤(𝐪)\underset{𝐪0}{}\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{2V}[\mathrm{\Lambda }_𝐤(𝐪)a_𝐤(𝐪)+a_𝐤^{}(𝐪)\mathrm{\Lambda }_𝐤(𝐪)][\mathrm{\Lambda }_𝐤^{^{}}(𝐪)a_𝐤^{^{}}(𝐪)+a_𝐤^{^{}}^{}(𝐪)\mathrm{\Lambda }_𝐤^{^{}}(𝐪)]$$
(100)
Here,
$$\stackrel{~}{\omega }_𝐤(𝐪)=\stackrel{~}{ϵ}_{𝐤+𝐪/2}\stackrel{~}{ϵ}_{𝐤𝐪/2}$$
(101)
$$\stackrel{~}{ϵ}_𝐤=ϵ_𝐤\underset{𝐪0}{}\frac{v_𝐪}{V}\overline{n}_{𝐤𝐪}$$
(102)
Let us diagonalise this via a Bogoliubov transformation. To this end one assumes that the diagonalised form has the following apperance,
$$H=\underset{i,𝐪}{}\omega _i(𝐪)d_i^{}(𝐪)d_i(𝐪)$$
(103)
$`\omega _i(𝐪)d_i(𝐪)={\displaystyle \underset{𝐤,𝐪}{}}\stackrel{~}{\omega }_𝐤(𝐪)[d_i(𝐪),a_𝐤^{}(𝐪)]a_𝐤(𝐪)+{\displaystyle \underset{𝐤,𝐪}{}}\stackrel{~}{\omega }_𝐤(𝐪)[d_i(𝐪),a_𝐤(𝐪)]a_𝐤^{}(𝐪)`$
$$\underset{𝐪0}{}\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{V}(\mathrm{\Lambda }_𝐤^{^{}}(𝐪)[d_i(𝐪),a_𝐤^{^{}}(𝐪)]+[d_i(𝐪),a_𝐤^{^{}}^{}(𝐪)]\mathrm{\Lambda }_𝐤^{^{}}(𝐪))(\mathrm{\Lambda }_𝐤(𝐪)a_𝐤(𝐪)+a_𝐤^{}(𝐪)\mathrm{\Lambda }_𝐤(𝐪))$$
(104)
$`\omega _i(𝐪)[d_i(𝐪),a_𝐤^{}(𝐪)]=\stackrel{~}{\omega }_𝐤(𝐪)[d_i(𝐪),a_𝐤^{}(𝐪)]`$
$$\mathrm{\Lambda }_𝐤(𝐪)\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{V}(\mathrm{\Lambda }_𝐤^{^{}}(𝐪)[d_i(𝐪),a_𝐤^{^{}}(𝐪)]+[d_i(𝐪),a_𝐤^{^{}}^{}(𝐪)]\mathrm{\Lambda }_𝐤^{^{}}(𝐪))$$
(105)
$`\omega _i(𝐪)[d_i(𝐪),a_𝐤(𝐪)]=\stackrel{~}{\omega }_𝐤(𝐪)[d_i(𝐪),a_𝐤(𝐪)]`$
$$+\mathrm{\Lambda }_𝐤(𝐪)\underset{𝐤𝐤^{^{}}}{}\frac{v_{𝐤𝐤^{^{}}}}{V}(\mathrm{\Lambda }_𝐤^{^{}}(𝐪)[d_i(𝐪),a_𝐤^{^{}}(𝐪)]+[d_i(𝐪),a_𝐤^{^{}}^{}(𝐪)]\mathrm{\Lambda }_𝐤^{^{}}(𝐪))$$
(106)
$$[d_i(𝐪),a_𝐤^{}(𝐪)]=\frac{\mathrm{\Lambda }_𝐤(𝐪)}{\omega _i(𝐪)\stackrel{~}{\omega }_𝐤(𝐪)}\text{ }𝑑\stackrel{}{r}\text{ }e^{i𝐤.\stackrel{}{r}}v(\stackrel{}{r})\{R_1(𝐪,\stackrel{}{r})+R_2(𝐪,\stackrel{}{r})\}$$
(107)
$$[d_i(𝐪),a_𝐤(𝐪)]=\frac{\mathrm{\Lambda }_𝐤(𝐪)}{\omega _i(𝐪)+\stackrel{~}{\omega }_𝐤(𝐪)}\text{ }𝑑\stackrel{}{r}\text{ }e^{i𝐤.\stackrel{}{r}}v(\stackrel{}{r})\{R_1(𝐪,\stackrel{}{r})+R_2(𝐪,\stackrel{}{r})\}$$
(108)
Here,
$$R_1(𝐪,\stackrel{}{r})=\frac{1}{V}\underset{𝐤^{^{}}}{}e^{i𝐤^{^{}}.\stackrel{}{r}}\mathrm{\Lambda }_𝐤^{^{}}(𝐪)[d_i(𝐪),a_𝐤^{^{}}(𝐪)]$$
(109)
$$R_2(𝐪,\stackrel{}{r})=\frac{1}{V}\underset{𝐤^{^{}}}{}e^{i𝐤^{^{}}.\stackrel{}{r}}\mathrm{\Lambda }_𝐤^{^{}}(𝐪)[d_i(𝐪),a_𝐤^{^{}}^{}(𝐪)]$$
(110)
$$(\omega _i(𝐪)\stackrel{~}{\omega }_{i_\stackrel{}{r}^{^{}}}(𝐪))R_2(𝐪,\stackrel{}{r}^{^{}})=\text{ }𝑑\stackrel{}{r}\text{ }F_1(\stackrel{}{r}\stackrel{}{r}^{^{}};𝐪)v(\stackrel{}{r})\{R_1(𝐪,\stackrel{}{r})+R_2(𝐪,\stackrel{}{r})\}$$
(111)
$$(\omega _i(𝐪)\stackrel{~}{\omega }_{i_\stackrel{}{r}^{^{}}}(𝐪))R_1(𝐪,\stackrel{}{r}^{^{}})=\text{ }𝑑\stackrel{}{r}\text{ }F_2(\stackrel{}{r}\stackrel{}{r}^{^{}};𝐪)v(\stackrel{}{r})\{R_1(𝐪,\stackrel{}{r})+R_2(𝐪,\stackrel{}{r})\}$$
(112)
$$F_1(\stackrel{}{r}\stackrel{}{r}^{^{}};𝐪)=\frac{1}{V}\underset{𝐤}{}\mathrm{\Lambda }_𝐤(𝐪)e^{i𝐤.(\stackrel{}{r}\stackrel{}{r}^{^{}})}$$
(113)
$$F_2(\stackrel{}{r}\stackrel{}{r}^{^{}};𝐪)=\frac{1}{V}\underset{𝐤}{}\mathrm{\Lambda }_𝐤(𝐪)e^{i𝐤.(\stackrel{}{r}\stackrel{}{r}^{^{}})}$$
(114)
Now define,
$$R_1(𝐪,\stackrel{}{r})+R_2(𝐪,\stackrel{}{r})=R(𝐪,\stackrel{}{r})$$
(115)
$$(\omega _i(𝐪)\stackrel{~}{\omega }_{i_\stackrel{}{r}^{^{}}}(𝐪))R(𝐪,\stackrel{}{r}^{^{}})=\text{ }d^3r\text{ }f_0(𝐪;\stackrel{}{r}\stackrel{}{r}^{^{}})v(\stackrel{}{r})R(𝐪,\stackrel{}{r})$$
(116)
$$f_0(𝐪;\stackrel{}{r}\stackrel{}{r}^{^{}})=\text{ }\frac{d^3k}{(2\pi )^3}\text{ }n_F(𝐤)e^{i𝐤.(\stackrel{}{r}\stackrel{}{r}^{^{}})}(2i)sin(\frac{1}{2}𝐪.(\stackrel{}{r}\stackrel{}{r}^{^{}}))$$
(117)
$`f_0(𝐪;\stackrel{}{r}\stackrel{}{r}^{^{}})={\displaystyle _0^{k_F}}\text{ }{\displaystyle \frac{4\pi k\text{ }dk}{(2\pi )^3}}\text{ }{\displaystyle \frac{sin(k|\stackrel{}{r}\stackrel{}{r}^{^{}}|)}{|\stackrel{}{r}\stackrel{}{r}^{^{}}|}}(2i)sin({\displaystyle \frac{1}{2}}𝐪.(\stackrel{}{r}\stackrel{}{r}^{^{}}))`$
$$=\frac{4\pi }{(2\pi )^3}\frac{(2i)sin(\frac{1}{2}𝐪.(\stackrel{}{r}\stackrel{}{r}^{^{}}))}{|\stackrel{}{r}\stackrel{}{r}^{^{}}|}\{k_F\text{ }\frac{cos(k_F|\stackrel{}{r}\stackrel{}{r}^{^{}}|)}{|\stackrel{}{r}\stackrel{}{r}^{^{}}|}+\frac{sin(k_F|\stackrel{}{r}\stackrel{}{r}^{^{}}|)}{|\stackrel{}{r}\stackrel{}{r}^{^{}}|^2}\}=i\text{ }C_0(𝐪)𝐪._\stackrel{}{r}\delta ^3(\stackrel{}{r}\stackrel{}{r}^{^{}})$$
(118)
$$(\omega _i(𝐪)i\frac{𝐪._\stackrel{}{r}^{^{}}}{m})R(𝐪,\stackrel{}{r}^{^{}})=i\text{ }C_0(𝐪)(𝐪._\stackrel{}{r}^{^{}})v(\stackrel{}{r}^{^{}})R(𝐪,\stackrel{}{r}^{^{}})$$
(119)
$$[\omega _i(𝐪)i\frac{𝐪._\stackrel{}{r}^{^{}}}{m}+i\text{ }C_0(𝐪)(𝐪._\stackrel{}{r}^{^{}}v(\stackrel{}{r}^{^{}}))]R(𝐪,\stackrel{}{r}^{^{}})=i\text{ }C_0(𝐪)v(\stackrel{}{r}^{^{}})(𝐪._\stackrel{}{r}^{^{}})R(𝐪,\stackrel{}{r}^{^{}})$$
(120)
This eigenvalue problem admits many eigenvalues. Actually, if not for the constraint of the square integrability of $`R(𝐪,\stackrel{}{r})`$, for each $`𝐪`$ any positive real number would be an eigenvalue.
$$[\omega _i(𝐪)i\frac{𝐪._\stackrel{}{r}^{^{}}}{m}+i\text{ }C_0(𝐪)(𝐪._\stackrel{}{r}^{^{}}v(\stackrel{}{r}^{^{}}))]R(𝐪,\stackrel{}{r}^{^{}})=i\text{ }C_0(𝐪)v(\stackrel{}{r}^{^{}})(𝐪._\stackrel{}{r}^{^{}})R(𝐪,\stackrel{}{r}^{^{}})$$
(121)
Let us now solve this equation,
$$R(𝐪,\stackrel{}{r})=R(𝐪,\stackrel{}{r}_0)exp(\frac{1}{i\text{ }q\text{ }cos\theta }_{r_0}^r𝑑r^{^{}}\frac{\omega _i(𝐪)+i\text{ }C_0(𝐪)\text{ }q\text{ }cos\theta \text{ }\frac{v(r^{^{}})}{r^{^{}}}}{\frac{1}{m}C_0(𝐪)v(r^{^{}})})$$
(122)
For some suitable $`\stackrel{}{r}_0`$.
This work was supported by the Dept. of Physics at University of Illinois at Urbana-Champaign. The authors may be contacted at the e-mail address setlur@mrlxpa.mrl.uiuc.edu.
## V Appendix
In this appendix, we show how to derive the full dielectric function using the generalised RPA. Along the way we point out some pitfalls and possible generalisations. Let us write the generalised RPA hamiltonian and try and compute the dielectric function.
$$H_0=\underset{𝐤}{}ϵ_𝐤n_0(𝐤)\underset{𝐪0}{}\frac{v_𝐪}{2V}\underset{𝐤}{}n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)$$
(123)
$$H_{ext}(t)=\underset{𝐪}{}(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))\underset{𝐤}{}n_𝐪(𝐤)$$
(124)
here $`n_𝐪(𝐤)=c_{𝐤+𝐪/2}^{}c_{bfk𝐪/2}`$ and $`n_0(𝐤)=c_𝐤^{}c_𝐤`$ Let us now write down the quation of motion for $`n_𝐪(𝐤)`$.
$`i{\displaystyle \frac{}{t}}n_𝐪(𝐤)={\displaystyle \underset{𝐤^{^{}}}{}}ϵ_𝐤^{^{}}[n_𝐪(𝐤),n_0(𝐤^{^{}})]`$
$`{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}{\displaystyle \underset{𝐤^{^{}}}{}}[n_𝐪(𝐤),n_0(𝐤^{^{}})]n_0(𝐤^{^{}}𝐪^{^{}})`$
$$+\underset{𝐤^{^{}},𝐪^{^{}}}{}(U_{ext}(𝐪^{^{}}t)+U_{ext}^{}(𝐪^{^{}}t))[n_𝐪(𝐤),n_𝐪^{^{}}(𝐤^{^{}})]$$
(125)
Now,
$`[n_𝐪(𝐤),n_𝐪^{^{}}(𝐤^{^{}})]=[c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2},c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}]`$
$$=c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}\delta _{𝐤𝐪/2,𝐤^{^{}}+𝐪^{^{}}/2}c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤𝐪/2}\delta _{𝐤+𝐪/2,𝐤^{^{}}𝐪^{^{}}/2}$$
(126)
One may approximate this as,
$`[n_𝐪(𝐤),n_𝐪^{^{}}(𝐤^{^{}})]=[c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2},c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}]`$
$$=[n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)]\delta _{𝐤,𝐤^{^{}}}\delta _{𝐪,𝐪^{^{}}}$$
(127)
The next approximation would be to replace the number operator by its c-number expectation value. But we shall desist from that for the moment.
$$[n_𝐪(𝐤),n_0(𝐤^{^{}})]=n_𝐪(𝐤)(\delta _{𝐤^{^{}},𝐤𝐪/2}\delta _{𝐤^{^{}},𝐤+𝐪/2})$$
(128)
$`i{\displaystyle \frac{}{t}}n_𝐪(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})n_𝐪(𝐤){\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(n_0(𝐤𝐪/2𝐪^{^{}})n_0(𝐤+𝐪/2𝐪^{^{}}))n_𝐪(𝐤)`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(129)
$$n_𝐪(𝐤)=U_{ext}(𝐪t)C_𝐪(𝐤)+U_{ext}^{}(𝐪t)D_𝐪(𝐤)$$
(130)
Let us now ignore fluctuations in the momentum distribution. That is, we are allowed to replace
$$n_0(𝐤^{^{}})n_𝐪(𝐤)=n_0(𝐤^{^{}})n_𝐪(𝐤)$$
(131)
$`\omega \text{ }U_{ext}(𝐪t)C_𝐪(𝐤)\omega \text{ }U_{ext}^{}(𝐪t)D_𝐪(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})U_{ext}(𝐪t)C_𝐪(𝐤)+(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})U_{ext}^{}(𝐪t)D_𝐪(𝐤)`$
$`{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(n_0(𝐤𝐪/2𝐪^{^{}})n_0(𝐤+𝐪/2𝐪^{^{}}))U_{ext}(𝐪t)C_𝐪(𝐤)`$
$`{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(n_0(𝐤𝐪/2𝐪^{^{}})n_0(𝐤+𝐪/2𝐪^{^{}}))U_{ext}^{}(𝐪t)D_𝐪(𝐤)`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(132)
$`\omega \text{ }C_𝐪(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})C_𝐪(𝐤){\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{V}}(n_0(𝐤𝐪/2𝐪^{^{}})n_0(𝐤+𝐪/2𝐪^{^{}}))C_𝐪(𝐤)`$
$$+(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(133)
Since,
$$U_{eff}(𝐪t)=U_{ext}(𝐪t)+\frac{v_𝐪}{V}\rho _𝐪^{^{}}U_{ext}(𝐪t)$$
(134)
$$\rho _𝐪^{^{}}=\underset{𝐤}{}C_𝐪(𝐤)$$
(135)
$$\underset{𝐤}{}C_𝐪(𝐤)=\underset{𝐤}{}\frac{n_0(𝐤𝐪/2)n_0(𝐤+𝐪/2)}{\omega \stackrel{~}{ϵ}_{𝐤+𝐪/2}+\stackrel{~}{ϵ}_{𝐤𝐪/2}}$$
(136)
If we use the definition of the dielectric function we get a wrong answer.
$$ϵ_{WRONG}(𝐪,\omega )=\frac{U_{ext}(𝐪t)}{U_{eff}(𝐪t)}=\frac{1}{1+\frac{v_𝐪}{V}_𝐤\frac{n_0(𝐤𝐪/2)n_0(𝐤+𝐪/2)}{\omega \stackrel{~}{ϵ}_{𝐤+𝐪/2}+\stackrel{~}{ϵ}_{𝐤𝐪/2}}}$$
(137)
The reason is because we have been very cavalier in our treatment of correlations. One has to consider the full Hamiltonian when dealing with the dielectric function rather than just $`H_0`$. Let us now try and do this.
$`i{\displaystyle \frac{}{t}}n_𝐪^t(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})n_𝐪^t(𝐤)+{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{2V}}[n_𝐪(𝐤),\rho _𝐪^{^{}}]\rho _𝐪^{^{}}^t+{\displaystyle \underset{𝐪^{^{}}0}{}}{\displaystyle \frac{v_𝐪^{^{}}}{2V}}\rho _𝐪^{^{}}^t[n_𝐪(𝐤),\rho _𝐪^{^{}}]`$
$$+\underset{𝐪^{^{}}0}{}(U_{ext}(𝐪^{^{}}t)+U_{ext}^{}(𝐪^{^{}}t))[n_𝐪(𝐤),\rho _𝐪^{^{}}]$$
(138)
$`[n_𝐪(𝐤),\rho _𝐪^{^{}}]={\displaystyle \underset{𝐤^{^{}}}{}}[c_{𝐤+𝐪/2}^{}c_{𝐤𝐪/2},c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}]`$
$`={\displaystyle \underset{𝐤^{^{}}}{}}c_{𝐤+𝐪/2}^{}c_{𝐤^{^{}}𝐪^{^{}}/2}\delta _{𝐤𝐪/2,𝐤^{^{}}+𝐪^{^{}}/2}{\displaystyle \underset{𝐤^{^{}}}{}}c_{𝐤^{^{}}+𝐪^{^{}}/2}^{}c_{𝐤𝐪/2}\delta _{𝐤+𝐪/2,𝐤^{^{}}𝐪^{^{}}/2}`$
$$\delta _{𝐪,𝐪^{^{}}}[n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)]$$
(139)
$`i{\displaystyle \frac{}{t}}n_𝐪^t(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})n_𝐪^t(𝐤)+{\displaystyle \frac{v_𝐪}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))\rho _𝐪^t`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(140)
Let us make a first pass at the computation of the dielectric function. Here, we make use of mean-field theory, that is, replace
$$n_0(𝐤^{^{}})\rho _𝐪=n_0(𝐤^{^{}})\rho _𝐪$$
(141)
$`\omega \text{ }n_𝐪^t(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})n_𝐪^t(𝐤)+{\displaystyle \frac{v_𝐪}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))\rho _𝐪^t`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(142)
$`n_𝐪^t(𝐤)={\displaystyle \frac{v_𝐪}{V}}{\displaystyle \frac{n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}}\rho _𝐪^t`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))\frac{n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}$$
(143)
This means,
$$\rho _𝐪=U_{ext}(𝐪t)\frac{P_0(𝐪,\omega )}{ϵ(𝐪,\omega )}$$
(144)
$$P_0(𝐪,\omega )=\underset{𝐤}{}\frac{n_0(𝐤𝐪/2)n_0(𝐤+𝐪/2)}{\omega ϵ_{𝐤+𝐪/2}+ϵ_{𝐤𝐪/2}}$$
(145)
$$ϵ(𝐪,\omega )=1\frac{v_𝐪}{V}P_0(𝐪,\omega )$$
(146)
From this and the fact that
$$ϵ_{gRPA}(𝐪,\omega )=\frac{U_{ext}(𝐪t)}{U_{eff}(𝐪t)}=ϵ(𝐪,\omega )$$
(147)
Next we would like to include fluctuations. Let us do this differently this time via the use of the BBGKY heirarchy.
$`i{\displaystyle \frac{}{t}}n_𝐪^t(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})n_𝐪^t(𝐤)+{\displaystyle \frac{v_𝐪}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))\rho _𝐪^t`$
$`+{\displaystyle \frac{v_𝐪}{V}}(F_{2A}(𝐤+𝐪/2,𝐪)F_{2A}(𝐤𝐪/2,𝐪))`$
$$+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(148)
Here,
$$F_{2A}(𝐤^{^{}},𝐪;t)=n_0(𝐤^{^{}})\rho _𝐪^tn_0(𝐤^{^{}})\rho _𝐪^t$$
(149)
$$F_2(𝐤^{^{}};𝐤,𝐪;t)=n_0(𝐤^{^{}})n_𝐪^t(𝐤)n_0(𝐤^{^{}})n_𝐪^t(𝐤)$$
(150)
$$F_{2A}(𝐤^{^{}},𝐪;t)=\underset{𝐤}{}F_2(𝐤^{^{}};𝐤,𝐪;t)$$
(151)
$`i{\displaystyle \frac{}{t}}F_2(𝐤^{^{}};𝐤,𝐪;t)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})F_2(𝐤^{^{}};𝐤,𝐪;t)+{\displaystyle \frac{v_𝐪}{V}}(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))\rho _𝐪^t`$
$$+\frac{v_𝐪}{V}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))F_{2A}(𝐤^{^{}},𝐪;t)+(U_{ext}(𝐪t)+U_{ext}^{}(𝐪t))(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))$$
(152)
Let us write,
$$F_2(𝐤^{^{}};𝐤,𝐪;t)=U_{ext}(𝐪t)F_{2,a}(𝐤^{^{}};𝐤,𝐪)+U_{ext}^{}(𝐪t)F_{2,b}(𝐤^{^{}};𝐤,𝐪)$$
(153)
$$\rho _𝐪^t=U_{ext}(𝐪t)\rho _𝐪^{^{}}+U_{ext}^{}(𝐪t)\rho _𝐪^{^{\prime \prime }}$$
(154)
Also define,
$$N(𝐤,𝐤^{^{}})=n_0(𝐤)n_0(𝐤^{^{}})n_0(𝐤)n_0(𝐤^{^{}})$$
(155)
$`\omega \text{ }F_{2,a}(𝐤^{^{}};𝐤,𝐪)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})F_{2,a}(𝐤^{^{}};𝐤,𝐪)+{\displaystyle \frac{v_𝐪}{V}}(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))\rho _𝐪^{^{}}`$
$$+\frac{v_𝐪}{V}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))F_{2A}^a(𝐤^{^{}},𝐪)+(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))$$
(156)
$`\omega \text{ }F_{2,b}(𝐤^{^{}};𝐤,𝐪)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})F_{2,b}(𝐤^{^{}};𝐤,𝐪)+{\displaystyle \frac{v_𝐪}{V}}(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))\rho _𝐪^{^{\prime \prime }}`$
$$+\frac{v_𝐪}{V}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))F_{2A}^b(𝐤^{^{}},𝐪)+(N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2))$$
(157)
$`ϵ(𝐪,\omega )\text{ }F_{2A}^a(𝐤^{^{}},𝐪)={\displaystyle \frac{v_𝐪}{V}}{\displaystyle \underset{𝐤}{}}{\displaystyle \frac{N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}}\rho _𝐪^{^{}}`$
$$+\underset{𝐤}{}\frac{N(𝐤^{^{}},𝐤+𝐪/2)N(𝐤^{^{}},𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}$$
(158)
$`\omega \text{ }C_𝐪(𝐤)=(ϵ_{𝐤𝐪/2}ϵ_{𝐤+𝐪/2})C_𝐪(𝐤)+{\displaystyle \frac{v_𝐪}{V}}(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))\rho _𝐪^{^{}}+{\displaystyle \frac{v_𝐪}{V}}(F_{2A}^a(𝐤+𝐪/2)F_{2A}^a(𝐤𝐪/2))`$
$$+(n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2))$$
(159)
$`C_𝐪(𝐤)={\displaystyle \frac{v_𝐪}{V}}{\displaystyle \frac{n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}}\rho _𝐪^{^{}}`$
$`+{\displaystyle \frac{n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}}`$
$$+\frac{v_𝐪}{V}\frac{F_{2A}^a(𝐤+𝐪/2,𝐪)F_{2A}^a(𝐤𝐪/2,𝐪)}{\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2}}$$
(160)
After all this, it may be shown that the overall dielectric function including possible fluctutations in the momentum distribution is given by,
$$ϵ_{eff}(𝐪,\omega )=ϵ_{gRPA}(𝐪,\omega )(\frac{v_𝐪}{V})^2\frac{P_2(𝐪,\omega )}{ϵ_{gRPA}(𝐪,\omega )}$$
(161)
Here,
$$P_2(𝐪,\omega )=\underset{𝐤,𝐤^{^{}}}{}\frac{N(𝐤+𝐪/2,𝐤^{^{}}+𝐪/2)N(𝐤𝐪/2,𝐤^{^{}}+𝐪/2)N(𝐤+𝐪/2,𝐤^{^{}}𝐪/2)+N(𝐤𝐪/2,𝐤^{^{}}𝐪/2)}{(\omega ϵ_{𝐤𝐪/2}+ϵ_{𝐤+𝐪/2})(\omega ϵ_{𝐤^{^{}}𝐪/2}+ϵ_{𝐤^{^{}}+𝐪/2})}$$
(162)
$$ϵ_{gRPA}(𝐪,\omega )=1+\frac{v_𝐪}{V}\underset{𝐤}{}\frac{n_0(𝐤+𝐪/2)n_0(𝐤𝐪/2)}{\omega ϵ_{𝐤+𝐪/2}+ϵ_{𝐤𝐪/2}}$$
(163) |
warning/0001/astro-ph0001240.html | ar5iv | text | # Probing The Gravity Induced Bias with Weak Lensing: Test of Analytical results Against Simulations
## 1 Introduction
Weak gravitational lensing (Bartelmann & Schneider 1999) is the distortion of images of the background galaxies through the tidal gravitational field of large scale background mass distribution. Study of such distortions provides us an unique way to probe statistical properties of intervening mass distribution.
Although there is no generally applicable definition of weak lensing the common aspect of all studies in weak gravitational lensing is that measurements of its effects are statistical in nature. Pioneering works in this direction were done by Blandford et al. (1991), Miralda-Escude (1991) and Kaiser (1992) based on the earlier work by Gunn (1967). Current progress in weak lensing can broadly be divided into two distinct categories. Where as Villumsen (1996), Stebbins (1996), Bernardeau et al. (1997) and Kaiser (1987) have focussed on linear and quasi-linear regime by assuming a large smoothing angle, several authors have developed a numerical technique to simulate weak lensing catalogs. Numerical simulations of weak lensing typically employ N-body simulations, through which ray tracing experiments are conducted (Schneider & Weiss 1988; Jarosszn’ski et al. 1990; Lee & Paczyn’ski 1990; Jarosszn’ski 1991; Babul & Lee 1991; Bartelmann & Schneider 1991, Blandford et al. 1991). Building on the earlier work of Wambsganns et al. (1995, 1997, 1998) most detailed numerical study of lensing was done by Wambsganns, Cen & Ostriker (1998). Other recent studies using ray tracing experiments have been conducted by Premadi, Martel & Matzner (1998), van Waerbeke, Bernardeau & Mellier (1998), Bartelmann et al (1998) and Couchman, Barber & Thomas (1998). While peturbative analysis can provide valuable information at large smoothing angle such analysis can not be used to study lensing at small angular scales as the whole perturbative series starts to diverge.
A complete analysis of the weak lensing statistics at small angular scales will require a similar analysis for the underlying dark matter distribution which we do not have at present. However there are several non-linear ansatze which predict a tree hierarchy for matter correlation functions and are thought to be successful to some degree in modeling results from numerical simulations. Most of these ansatze assumes a tree hierarchy for higher order correlation functions and they disagree with each other by the way they assign weights to trees of same order but of different topologies (Balian & Schaeffer 1989, Bernardeau & Schaeffer 1992; Szapudi & Szalay 1993). Evolution of two-point correlation functions in all such approximations are left arbitrary. However recent studies by several authors (Hamilton et al 1991; Nityananda & Padmanabhan 1994; Padmanabhan et al. 1996; Jain, Mo & White 1995; Peacock & Dodds 1996) have provided us a very accurate fitting formula for the evolution of the two-point correlation function which can be used in combination with these hierarchical ansatze to predict clustering properties of dark matter distribution in the universe.
Most recent studies in weak lensing have mainly focussed on lower order cumulants (van Waerbeke, Bernardeau & Mellier 1998, Hui 1999, Munshi & Jain 1999a), cumulant correlators (Munshi & Jain 1999b) and errors associated with their measurement from observational data using different filter functions (Reblinsky et al. 1999). However it is well known that higher order moments and more sensitive to the tail of the distribution function which they represent and are sensitive to measurement errors due to finite size of catalogs (Szapudi & Colombi 1996). On the other hand numerical simulations involving ray tracing techniques have already shown that while probability distribution function (PDF) associated with density field is not sensitive to cosmological parameters its weak lensing counterpart can lift such a degeneracy and can help us in estimation of cosmological parameters from observational data (Jain, Seljak & White 1999). Munshi & Jain (1999a,b) extended such studies to show that the hierarchical ansatz can actually be used to make concrete analytical predictions for lower order statistical properties of the convergence field. Valageas (1999a) has used hierarchical ansatz for computing error involved in estimation of $`\mathrm{\Omega }_0`$ and $`\mathrm{\Lambda }`$ from SNeIa observations. A similar fitting function was recently proposed by Wang (1999). Our formalism is similar to that of Valageas (1999a) although the results obtained by us include the effect of smoothing and analysis presented here correspond to the joint probability distribution function of two smoothed patches and hence the bias associated with them. Our formalism presented here is an extension of the study of the PDF using similar techniques (Munshi & Jain 1999b, Valageas 1999b), it can very easily be extended to multi-point PDF and hence to compute the $`S_N`$ parameters associated with “hot spots” of the convergence maps. Analytical results and detailed comparison against numerical simulations will be presented elsewhere (Munshi & Coles 1999a,b).
The paper is organized as follows. In section $`2`$ we describe briefly ray tracing simulations. Section $`3`$ presents most of the analytical results necessary for computation of the bias of the smoothed convergence field $`\kappa _s`$. Results of data analysis are presented in section $`4`$. Section $`5`$ is left for discussion of our result in general cosmological framework.
## 2 Generation of Convergence Maps from N-body Simulations
Convergence maps are generated by solving the geodesic equations for the propagation of light rays through N-body simulations of dark matter clustering. The simulations used for our study are adaptive $`P^3M`$ simulations with $`256^3`$ particles and were carried out using codes kindly made available by the VIRGO consortium. These simulations can resolve structures larger than $`30h^1kpc`$ at $`z=0`$ accurately. These simulations were carried out using 128 or 256 processors on CRAY T3D machines at Edinburgh Parallel Computer Center and at the Garching Computer Center of the Max-Planck Society. These simulations were intended primarily for studies of the formation and clustering of galaxies (Kauffmann et al 1999a, 1999b; Diaferio et al 1999) but were made available by these authors and by the Virgo Consortium for this and earlier studies of gravitational lensing (Jain, Seljak & White 1999, Reblinsky et al. 1999, Munshi & Jain 1999a,b).
Ray tracing simulations were carried out by Jain et al. (1999) using a multiple lens-plane calculation which implements the discrete version of recursion relations for mapping the photon position and the Jacobian matrix (Schneider & Weiss 1988; Schneider, Ehler & Falco 1992). In a typical experiment $`4\times 10^6`$ rays are used to trace the underlying mass distribution. The dark matter distribution between the source and the observer is projected onto $`20`$ \- $`30`$ planes. The particle positions on each plane are interpolated onto a $`2048^2`$ grid. On each plane the shear matrix is computed on this grid from the projected density by using Fourier space relations between the two. The photons are propagated starting from a rectangular grid on the first lens plane. The regular grid of photon position gets distorted along the line of sight. To ensure that all photons reach the observer, the ray tracing experiments are generally done backward in time from the observer to the source plane at red-shift $`z=z_s`$. The resolution of the convergence maps is limited by both the resolution scale associated with numerical simulations and also due to the finite resolution of the grid to propagate photons. The outcome of these simulations are shear and convergence maps on a two dimensional grid. Depending on the background cosmology the two dimensional box represents a few degree scale patch on the sky. For more details on the generation of $`\kappa `$-maps, see Jain et al (1999).
## 3 Analytical Predictions
In this section we will provide necessary theoretical background for the bias function in the context of hierarchical clustering.
We will be using the following line element for the background geometry of the universe:
$$d\tau ^2=c^2dt^2+a^2(t)(d\chi ^2+r^2(\chi )d^2\mathrm{\Omega })$$
(1)
Where we have denoted angular diameter distance by $`r(\chi )`$ and scale factor of the universe by $`a(t)`$. $`r(\chi )=K^{1/2}\mathrm{sin}(K^{1/2}\chi )`$ for positive curvature, $`r(\chi )=(K)^{1/2}\mathrm{sinh}((K)^{1/2}\chi )`$ for negative curvature and $`\chi `$ for the flat universe. For a present value of of $`H_0`$ and $`\mathrm{\Omega }_0`$ we have $`K=(\mathrm{\Omega }_01)H_0^2`$. The various parameters charecterising different cosmological models are listed in Table - 1.
### 3.1 The Formalism
The statistics of weak lensing convergence $`\kappa `$ is very much similar to that of the projected density field. In what follows we will be considering a small patch of the sky where we can use the plane parallel approximation or small angle approximation to replace spherical harmonics by Fourier modes. The 3D density contrast $`\delta `$ along the line of sight when projected into 2D sky with a weight function $`\omega (\chi )`$ will provide us the projected density contrast or the weak-lensing convergence at a direction $`\gamma `$.
$$\kappa (\gamma _1)=_0^{\chi _s}𝑑\chi _1\omega (\chi _1)\delta (r(\chi )\gamma _1)$$
(2)
In all our discussion we will be placing the sources at a fixed red-shift (an approximation not too difficult to modify for more realistic description), the weight function can be expressed as $`\omega (\chi )=3/4ac^2H_0^2\mathrm{\Omega }_mr(\chi )r(\chi _s\chi )/r(\chi _s)`$. Where $`\chi _s`$ is the comoving radial distance to the source at a redshift $`z_s`$. Fourier decomposition of $`\delta `$ can be written as:
$$\kappa (\gamma _1)=_0^{\chi _s}𝑑\chi _1\omega (\chi _1)\frac{d^3𝐤}{(2\pi )^3}\mathrm{exp}(i\chi _1k_{}+ir\theta k_{})\delta _k$$
(3)
Where we have used $`k_{}`$ and $`k_{}`$ to denote components of wave vector $`𝐤`$ parallel and perpendicular to line of sight direction $`\gamma `$. In small angle approximation however one assumes that $`k_{}`$ much larger compared to $`k_{}`$. We will denote the angle between the line of sight direction $`\gamma `$ and the wave vector $`𝐤`$ by $`\theta `$. Using the definitions we have introduced above we can compute the smoothed projected two-point correlation function (Peebles 1980, Kaiser 1992, Kaiser 1998):
$$\kappa (\gamma _1)\kappa (\gamma _2)_c=_0^{\chi _s}𝑑\chi _1\frac{\omega ^2(\chi _1)}{r^2(\chi _1)}\frac{d^2𝐥}{(2\pi )^2}\mathrm{exp}(\theta l)\mathrm{P}\left(\frac{l}{r(\chi )}\right)W_2^2(l\theta _0).$$
(4)
Where we have introduced a new notation $`𝐥=r(\chi )𝐤_{}`$ which denotes a scaled wave vector projected on the surface of the sky. The average of two-point correlation function $`\kappa _s^2`$ smoothed over an angle $`\theta _0`$ with a top-hat smoothing window $`W_2(l\theta _0)`$ is useful to quantify the fluctuations in $`\kappa _s`$ which is often used to reconstruct the matter power spectrum $`P(𝐤)`$ (Jain, Selzak & White 1998).
$$\kappa _s^2_c=_0^{\chi _s}𝑑\chi _1\frac{\omega ^2(\chi _1)}{r^2(\chi _1)}\frac{d^2𝐥}{(2\pi )^2}\mathrm{P}\left(\frac{l}{r(\chi )}\right)W_2^2(l\theta _0)$$
(5)
Similar analysis for the higher order cumulant correlators (Szapudi & Szalay 1997, Munshi & Coles 1999) of the smoothed convergence field relating $`\kappa _s^m(\gamma _1)\kappa _s^n(\gamma _2)_c`$ with multi-spectra of underlying dark matter distribution $`B_p`$ (Munshi & Coles 1999a):
$$\kappa _s^2(\gamma _1)\kappa _s(\gamma _2)_c=_0^{\chi _s}\frac{\omega ^3(\chi )}{r^4(\chi )}𝑑\chi \frac{d^2𝐥_1}{(2\pi )^2}\frac{d^2𝐥_2}{(2\pi )^2}W_2(l_1\theta _0)W_2(l_2\theta _0)W_2(l_3\theta _0)\mathrm{exp}(il_2\theta _{12})B_3(\frac{𝐥_1}{r(\chi )},\frac{𝐥_2}{r(\chi )},\frac{𝐥_3}{r(\chi )});$$
(6)
$`\kappa _s^3(\gamma _1)\kappa _s(\gamma _2)_c={\displaystyle _0^{\chi _s}}{\displaystyle \frac{\omega ^3(\chi )}{r^4(\chi )}}𝑑\chi {\displaystyle \frac{d^2𝐥_1}{(2\pi )^2}\frac{d^2𝐥_2}{(2\pi )^2}\frac{d^2𝐥_3}{(2\pi )^2}W_2(l_1\theta _0)W_2(l_2\theta _0)W_2(l_3\theta _0)W_2(l_4\theta _0)\mathrm{exp}(il_3\theta _{12})}`$
$`B_4({\displaystyle \frac{𝐥_1}{r(\chi )}},{\displaystyle \frac{𝐥_2}{r(\chi )}},{\displaystyle \frac{𝐥_3}{r(\chi )}},{\displaystyle \frac{𝐥_4}{r(\chi )}});`$ (7)
$`\kappa _s^2(\gamma _1)\kappa _s^2(\gamma _2)_c={\displaystyle _0^{\chi _s}}{\displaystyle \frac{\omega ^3(\chi )}{r^4(\chi )}}𝑑\chi {\displaystyle \frac{d^2𝐥_1}{(2\pi )^2}\frac{d^2𝐥_2}{(2\pi )^2}\frac{d^2𝐥_3}{(2\pi )^2}W(l_1\theta _0)W_2(l_2\theta _0)W_2(l_3\theta _0)W_2(l_4\theta _0)\mathrm{exp}(i(l_1+l_2)\theta _{12})}`$
$`B_4({\displaystyle \frac{𝐥_1}{r(\chi )}},{\displaystyle \frac{𝐥_2}{r(\chi )}},{\displaystyle \frac{𝐥_3}{r(\chi )}},{\displaystyle \frac{𝐥_4}{r(\chi )}}).`$ (8)
In general we can express the cumulant correlators of arbitrary order $`\kappa _s^m(\gamma _1)\kappa _s^n(\gamma _2)_c`$ in terms of multi-spectra $`B_{m+n}`$ as:
$`\kappa _s^m(\gamma _1)\kappa _s^n(\gamma _2)_c={\displaystyle _0^{\chi _s}}{\displaystyle \frac{\omega ^{n+m}(\chi )}{r^{2(n+m1)}(\chi )}}𝑑\chi {\displaystyle \frac{d^2𝐥_1}{(2\pi )^2}\mathrm{}\frac{d^2𝐥_{n+m1}}{(2\pi )^2}W_2(l_1\theta _0)\mathrm{}W_2(l_{n+m}\theta _0)\mathrm{exp}[i(l_1+\mathrm{}+l_m)\theta _{12}]}`$
$`B_{m+n}({\displaystyle \frac{𝐥_1}{r(\chi )}},\mathrm{},{\displaystyle \frac{𝐥_{m+n}}{r(\chi )}}).`$ (9)
We will use and extend these results in this paper to show that it is possible to compute the whole bias function $`b(>\kappa _s)`$, i.e. the bias associated with those spots in convergence map which $`\kappa _s`$ is above certain threshold (which acts as a generating function for these cumulant correlators) from the statistics of underlying over-dense dark objects.Details of analytical results presented here can be found in Munshi & Coles (1999b).
### 3.2 Hierarchical Ansatze
In deriving the above expressions we have not used any specific form for the matter correlation hierarchy, however the length scales involved in small angles are in the highly non-linear regime. Assuming a tree model for the matter correlation hierarchy in the highly non-linear regime one can write the most general case as (Groth & Peebles 1977; Fry & Peebles 1978; Davis & Peebles 1983; Bernardeau & Schaeffer 1992; Szapudi & Szalay 1993):
$$\xi _N(𝐫_\mathrm{𝟏},\mathrm{}𝐫_𝐍)=\underset{\alpha ,\mathrm{N}\mathrm{trees}}{}Q_{N,\alpha }\underset{\mathrm{labellings}}{}\underset{\mathrm{edges}(\mathrm{i},\mathrm{j})}{\overset{(N1)}{}}\xi (𝐫_𝐢,𝐫_𝐣)$$
(10)
It is interesting to note that an exactly similar hierarchy develops in the quasi-linear regime in the limit of vanishing variance (Bernardeau 1992), however the hierarchal amplitudes $`Q_{N,\alpha }`$ become shape dependent in such a case. In the highly nonlinear regime there are some indications that these functions become independent of shape parameters as has been proved by studies of lowest order parameter $`Q_3=Q`$ using high resolution numerical simulations (Sccociamarro et al. 1998). In the Fourier space such an ansatz will mean that the whole hierarchy of multi-spectra $`B_p`$ can be written in terms of sum of products of power-spectra, e.g. in low orders we can write:
$`B_2(𝐤_1,𝐤_2,𝐤_3)_{{\scriptscriptstyle k_i}=0}=Q(P(𝐤_\mathrm{𝟏})P(𝐤_\mathrm{𝟐})+P(𝐤_\mathrm{𝟐})P(𝐤_\mathrm{𝟑})+P(𝐤_\mathrm{𝟑})P(𝐤_\mathrm{𝟏})),`$ (11)
$`B_3(𝐤_1,𝐤_2,𝐤_3,𝐤_4)_{{\scriptscriptstyle k_i}=0}=R_aP(𝐤_\mathrm{𝟏})P(𝐤_\mathrm{𝟏}+𝐤_\mathrm{𝟐})P(𝐤_\mathrm{𝟏}+𝐤_\mathrm{𝟐}+𝐤_\mathrm{𝟑})+\mathrm{cyc}.\mathrm{perm}.+R_bP(𝐤_\mathrm{𝟏})P(𝐤_\mathrm{𝟐})p(𝐤_\mathrm{𝟑})+\mathrm{cyc}.\mathrm{perm}.`$ (12)
Different hierarchal models differ the way they predict the amplitudes of different tree topologies. Bernardeau & Schaeffer (1992) considered the case where amplitudes in general are factorizable, at each order one has a new “star” amplitude and higher order “snake” and “hybrid” amplitudes are constructed from lower order “star” amplitudes (see Munshi, Melott & Coles 1999a,b,c for a detailed description). In models proposed by Szapudi & Szalay (1993) it is assumed that all hierarchal amplitudes of a given order are actually degenerate.
We do not use any of these specific models for clustering and only assume a hierarchal nature for the higher order correlation functions. Galaxy surveys have been used to study these ansatze. Our main motivation here is to show that weak-lensing surveys can also provide valuable information in this direction, in addition to constraining matter power-spectra and the background geometry of the universe. The most general form for the lower order cumulant correlators in the large separation limit can be expressed as:
$`\kappa _s^2(\gamma _1)\kappa _s(\gamma _2)_c`$ $`=`$ $`2Q_3𝒞_3[\kappa _{\theta _0}\kappa _{\theta _{12}}]=C_{21}^\eta 𝒞_3[\kappa _{\theta _0}\kappa _{\theta _{12}}]C_{21}^\kappa \kappa _s^2_c\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c,`$ (13)
$`\kappa _s^3(\gamma _1)\kappa _s(\gamma _2)_c`$ $`=`$ $`(3R_a+6R_b)𝒞_4[\kappa _{\theta _0}^2\kappa _{\theta _{12}}]=C_{31}^\eta 𝒞_4[\kappa _{\theta _0}^2\kappa _{\theta _{12}}]C_{31}^\kappa \kappa _s^2_c^2\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c,`$ (14)
$`\kappa _s^2(\gamma _1)\kappa _s^2(\gamma _2)_c`$ $`=`$ $`4R_b𝒞_4[\kappa _{\theta _0}^2\kappa _{\theta _{12}}]=C_{22}^\eta 𝒞_4[\kappa _{\theta _0}\kappa _{\theta _{12}}]C_{22}^\kappa \kappa _s^2_c^2\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c,`$ (15)
$`\kappa _s^4(\gamma _1)\kappa _s(\gamma _2)_c`$ $`=`$ $`(24S_a+36S_b+4S_c)𝒞_5[\kappa _{\theta _0}^3\kappa _{\theta _{12}}]=C_{41}^\eta 𝒞_5[\kappa _{\theta _0}^3\kappa _{\theta _{12}}]C_{41}^\kappa \kappa _s^2_c^3\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c,`$ (16)
$`\kappa _s^3(\gamma _1)\kappa _s^2(\gamma _2)_c`$ $`=`$ $`(12S_a+6S_b)𝒞_5[\kappa _{\theta _0}^3\kappa _{\theta _{12}}]=C_{32}^\eta 𝒞_5[\kappa _{\theta _0}^3\kappa _{\theta _{12}}]C_{32}^\kappa \kappa _s^2_c^3\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c.`$ (17)
Where $`C_{mn}^\kappa `$ denotes the cumulant correlators of the convergence field and $`C_{mn}^\eta `$ denotes the cumulant correlators for the underlying mass distribution. Extending above results to arbitrary order we can write:
$$\kappa _s^p(\gamma _1)\kappa _s^q(\gamma _2)_c=C_{pq}^\eta 𝒞_{p+q}[\kappa _{\theta _0}^{(p+q2)}\kappa _{\theta _{12}}]=C_{pq}^\kappa \kappa _s^2_c^{(p+q2)}\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c.$$
(18)
where $`C_{pq}^\eta `$ denotes the cumulant correlators for the underlying mass distribution,
$$𝒞_t[\kappa _{\theta _0}^m\kappa _{\theta _{12}}]=_0^{\chi _s}\frac{\omega ^t(\chi )}{r^{2(t1)}(\chi )}\kappa _{\theta _0}^m\kappa _{\theta _{12}}𝑑\chi ,$$
(19)
similarly the following notations were used to simplify the above expressions:
$`\kappa _{\theta _0}`$ $``$ $`{\displaystyle \frac{d^2𝐥}{(2\pi )^2}P\left(\frac{l}{r(\chi )}\right)W_2^2(l\theta _0)},`$ (20)
$`\kappa _{\theta _{12}}`$ $``$ $`{\displaystyle \frac{d^2𝐥}{(2\pi )^2}P\left(\frac{l}{r(\chi )}\right)W_2^2(l\theta _0)\mathrm{exp}(l\theta _{12})}.`$ (21)
.
The hierarchical expression for the lowest order cumulant i.e. $`S_3`$ was derived by Hui 1998. He also showed that his result agrees well with numerical ray tracing experiments by Jain, Seljak and White (1998). More recent studies have shown that higher order cumulants and even the two-point statistics such as cumulant correlators can also be reliably modeled in a similar way (Munshi & Coles 1999, Munshi & Jain 1999a). We extend such results in this paper to compute the complete bias function $`b_\kappa (\kappa _s)`$ (which is related to the low order two-point statistics such as the cumulant correlators defined above) associated with high $`\kappa _s`$ spots in the convergence map,
$`p_\kappa (\kappa _1,\kappa _2)d\kappa _1d\kappa _2=p_\kappa (\kappa _1)p_\kappa (\kappa _2)(1+b_\kappa (\kappa _1)\xi _{12}^\kappa b_\kappa (\kappa _2))d\kappa _1d\kappa _2,`$ (22)
and its relation to the bias associated with collapsed objects in underlying density field $`1+\delta `$.
### 3.3 The bias associated with collapsed objects
The success of analytical results in predicting the lower order cumulants, cumulant correlators and the one-point smoothed PDF(Munshi & Coles 199a,b; Munshi & Jain 1999a,b; Valageas 1999a,b), motivates a general analysis of the bias associated with the high $`\kappa _s`$ spots in convergence maps and their relation with bias associated with the high peaks in the underlying mass distribution. For this purpose we found that formalism developed by Balian & Schaeffer (1989) and Bernardeau & Schaeffer (1992) (later extended Bernardeau (1992, 1994)) to be most suitable. These results are based on a very general tree hierarchy for higher order correlation function and the assumption that amplitudes associated with different tree-topologies are constant once in the highly non-linear regime. Later these results were generalized by Bernardeau (1992, 1994) for the case of quasi-linear regime too, where perturbative dynamics can be used to make more concrete predictions. Errors associated with top-hat smoothing and its use with hierarchical ansatz has been studied in detail by Szapudi et al.(1992) and Boschan et al. (1994) and will be assumed small even in the case of two-point cumulant correlators throughout this paper. In this section we review the basic results from scaling models in the highly non-linear regime and quasi-linear regime before extending such models to the statistics of smoothed convergence field $`\kappa (\theta _0)`$ where we will show that although weak lensing statistics probes the highly non-linear regime, projection effects make the variance smaller than unity justifying the use of quasi-linear results even though the generating function remains the one for highly non-linear regime. We will be using the small angle approximation in our derivation. Our results can in principle be generalized for the case of projected galaxy catalogs too, a detailed analysis will be presented elsewhere.
#### 3.3.1 The Generating Function
In scaling analysis of the probability distribution function (PDF) the void probability distribution function (VPF) plays most fundamental role, which can related to the generating function of the cumulants or $`S_N`$ parameters, $`\varphi (y)`$ (White 1979, Balian & Schaeffer 1989)
$$P_v(0)=\mathrm{exp}(\overline{N}\sigma (N_c))=\mathrm{exp}\left(\frac{\varphi (N_c)}{\overline{\xi }_2}\right).$$
(23)
Where $`P_v(0)`$ is the probability of having no “particles” in a cell of of volume $`v`$, $`\overline{N}`$ is the average occupancy of these “particles” and $`N_c=\overline{N}\overline{\xi }_2`$. The VPF is meaningful only for discrete distribution of particles and can not be defined for smooth density fields such as $`\delta `$ or $`\kappa (\theta _0)`$. However the scaling functions defined above $`\sigma (y)=\frac{\varphi (y)}{y}`$ are very much useful even for continuous distributions where they can be used as a generating function of one-point cumulants or $`S_p`$ parameters,
$$\varphi (y)=\underset{p=1}{\overset{\mathrm{}}{}}\frac{S_p}{p!}y^p.$$
(24)
The function $`\varphi (y)`$ satisfies the constraint $`S_1=S_2=1`$ necessary for proper normalization of PDF. The other generating function which plays a very important role in such analysis is the generating function for vertex amplitudes $`\nu _n`$, associated with nodes appearing in tree representation of higher order correlation hierarchy ($`Q_3=\nu _2`$, $`R_a=\nu _2^2`$ and $`R_b=\nu _3`$).
$$𝒢(\tau )=1\tau +\frac{\nu _2}{2!}\tau ^2\frac{\nu _3}{3!}\tau ^3+\mathrm{}$$
(25)
A more specific model for $`𝒢(\tau )`$ can be used, which is useful to make more specific predictions (Bernardeau & Schaeffer 1979):
$$𝒢(\tau )=\left(1+\frac{\tau }{\kappa _a}\right)^{\kappa _a}.$$
(26)
We will relate $`\kappa _a`$ with other parameters of scaling models. While the definition of VPF do not use any specific form of hierarchical ansatz it is to realize that writing the tree amplitudes in terms of the weights associated with nodes is only possible when one assumes a factorizable model for tree hierarchy (Bernardeau & Schaeffer 1992) and other possibilities which do not violate the tree models are indeed possible too (Bernardeau & Schaeffer 1999). The generating functions for tree nodes can be related to the VPF by solving a pair of implicit equations (Balian & Schaeffer 1989),
$`\varphi (y)=y𝒢(\tau ){\displaystyle \frac{1}{2}}y\tau {\displaystyle \frac{d}{d\tau }}G(\tau ),`$ (27)
$`\tau =y{\displaystyle \frac{d}{d\tau }}𝒢(\tau ).`$ (28)
However a more detailed analysis is needed to include the effect of correlation between two or more correlated volume element which will provide information about bias, cumulants and cumulant correlators of these collapsed object (as opposed to the cumulants and cumulant correlators of the whole convergence map, e.g. Munshi & Jain (1999a)). However we will only quote results useful for measurement of bias from ray-tracing simulations as detailed derivations of related results including related error analysis can be found elsewhere (Bernardeau & Schaeffer 1992, 1999; Munshi et al. 1999a,b,c; Coles et al. 1999).
Notice that $`\tau (y)`$ (also denoted by $`\beta (y)`$ in the literature) plays the role of generating function for factorized cumulant correlators $`C_{p1}`$ ($`C_{pq}=C_{p1}C_{q1}`$):
$$\tau (y)=\underset{p=1}{\overset{\mathrm{}}{}}\frac{C_{p1}}{p!}y^p$$
(29)
#### 3.3.2 The Highly Non-linear Regime
The PDF $`p(\delta )`$ and bias $`b(\delta )`$ can be related to their generating functions VPF $`\varphi (y)`$ and $`\tau (y)`$ respectively by following equations (Balian & Schaeffer 1989, Bernardeau & Schaeffer 1992, Bernardeau & Schaeffer 1999),
$`p(\delta )={\displaystyle _i\mathrm{}^i\mathrm{}}{\displaystyle \frac{dy}{2\pi i}}\mathrm{exp}\left[{\displaystyle \frac{(1+\delta )y\varphi (y)}{\overline{\xi }_2}}\right],`$ (30)
$`b(\delta )p(\delta )={\displaystyle _i\mathrm{}^i\mathrm{}}{\displaystyle \frac{dy}{2\pi i}}\tau (y)\mathrm{exp}\left[{\displaystyle \frac{(1+\delta )y\varphi (y)}{\overline{\xi }_2}}\right].`$ (31)
It is clear that the function $`\varphi (y)`$ completely determines the behavior of the PDF $`p(\delta )`$ for all values of $`\delta `$. However different asymptotic expressions of $`\varphi (y)`$ govern the behavior of $`p(\delta )`$ for different intervals of $`\delta `$. For large $`y`$ we can express $`\varphi (y)`$ as:
$$\varphi (y)=ay^{1\omega }.$$
(32)
Where we have introduced a new parameter $`\omega `$ for the description of VPF. This parameter plays a very important role in scaling analysis. No theoretical analysis has been done so far to link $`\omega `$ with initial power spectral index $`n`$. Numerical simulations are generally used to fix $`\omega `$ for a specific initial condition. Such studies have confirmed that the increase in power on smaller scales increases the value of $`\omega `$. Typically initial power spectrum with spectral index $`n=2`$ (which should model CDM like spectra we considered in our simulations at small length scales) produces a value of $`.3`$ which we will be using in our analysis of PDF of the convergence field $`\kappa _s`$ (Colombi et. al. (1992, 1994, 1995). The VPF $`\varphi (y)`$ and its two-point analog $`\tau (y)`$ both exhibit singularity for small but negative value of $`y_s`$,
$`\varphi (y)=\varphi _sa_s\mathrm{\Gamma }(\omega _s)(yy_s)^{\omega _s},`$ (33)
$`\tau (y)=\tau _sb_s(yy_s)^{\omega _s1}.`$
For the factorizable model of the hierarchical clustering the parameter $`\omega _s`$ takes the value $`3/2`$ and $`a_s`$ and $`b_s`$ can be expressed in terms of the nature of the generating function $`𝒢(\tau )`$ and its derivatives near the singularity $`\tau _s`$ (Bernardeau & Schaeffer 1992):
$`a_s={\displaystyle \frac{1}{\mathrm{\Gamma }(1/2)}}𝒢^{}(\tau _s)𝒢^{\prime \prime }(\tau _s)\left[{\displaystyle \frac{2𝒢^{}(\tau _s)𝒢^{\prime \prime }(\tau _s)}{𝒢^{\prime \prime \prime }(\tau _s)}}\right]^{3/2},`$ (34)
$`b_s=\left[{\displaystyle \frac{2𝒢^{}(\tau _s)𝒢^{\prime \prime }(\tau _s)}{𝒢^{\prime \prime \prime }(\tau _s)}}\right]^{1/2}.`$ (35)
As mentioned before the parameter $`k_a`$ which we have introduced in the definition of $`𝒢(\tau )`$ can be related to the parameters $`a`$ and $`\omega `$ appearing in the asymptotic expressions of $`\varphi (y)`$ (Balian & Schaeffer 1989, Bernardeau & Schaeffer 1992),
$`\omega =k_a/(k_a+2),`$ (36)
$`a={\displaystyle \frac{k_a+2}{2}}k_a^{k_a/k_a+2}.`$ (37)
Similarly the parameter $`y_s`$ which describe the behavior of the function $`\varphi (y)`$ near its singularity can be related to the behavior of $`𝒢(\tau )`$ near $`\tau _s`$ which is the solution of the equation (Balian & Schaeffer 1989, Bernardeau & Schaeffer 1992),
$$\tau _s=\frac{𝒢^{}(\tau _s)}{𝒢^{\prime \prime }(\tau _s)},$$
(38)
finally we can relate $`k_a`$ to $`y_s`$ by following expression (see eq. (36)):
$$y_s=\frac{\tau _s}{𝒢^{}(\tau _s)},$$
(39)
or
$$\frac{1}{y_s}=x_{}=\frac{1}{k_a}\frac{(k_a+2)^{k_a+2}}{(k_a+1)^{k_a+1}}.$$
(40)
The newly introduced variable $`x_{}`$ will be useful to define large $`\delta `$ tail of the PDF $`p(\delta )`$ and the bias $`b(\delta )`$. Different asymptotes in $`\varphi (y)`$ are linked with behavior of $`p(\delta )`$ for various regimes of $`\delta `$. For very large values of variance i.e. $`\xi _2`$ it is possible to define a scaling function $`p(\delta )=\frac{1}{\xi _2^2}h(x)`$ which will encode the scaling behavior of PDF, where plays the role of the scaling variable and is defined as $`\frac{1+\delta }{\xi _2}`$. We list different ranges of $`\delta `$ and specify the behavior of $`p(\delta )`$ and $`b(\delta )`$ in these regimes (Balian & Schaeffer 1989).
$$\overline{\xi }^{\frac{\omega }{(1\omega )}}<<1+\delta <<\overline{\xi };p(\delta )=\frac{a}{\overline{\xi }_2^2}\frac{1\omega }{\mathrm{\Gamma }(\omega )}\left(\frac{1+\delta }{\xi _2}\right)^{\omega 2};b(\delta )=\left(\frac{\omega }{2a}\right)^{1/2}\frac{\mathrm{\Gamma }(\omega )}{\mathrm{\Gamma }[\frac{1}{2}(1+\omega )]}\left(\frac{1+\delta }{\overline{\xi }_2}\right)^{(1\omega )/2}$$
(41)
$$1+\delta >>\overline{\xi }_2;p(\delta )=\frac{a_s}{\overline{\xi }_2^2}\left(\frac{1+\delta }{\overline{\xi }_2}\right)\mathrm{exp}\left(\frac{1+\delta }{x_{}\overline{\xi }_2}\right);b(\delta )=\frac{1}{𝒢^{}(\tau _s)}\frac{(1+\delta )}{\overline{\xi }_2}$$
(42)
The integral constraints satisfied by scaling function are $`S_1=_0^{\mathrm{}}xh(x)𝑑x=1`$ and $`S_2=_0^{\mathrm{}}x^2h(x)𝑑x=1`$. These take care of normalization of the function $`p(\delta )`$. Similarly the normalization constraint over $`b(\delta )`$ can be expressed as $`C_{11}=_0^{\mathrm{}}xb(x)h(x)𝑑x=1`$, which translates into $`_1^{\mathrm{}}𝑑\delta b(\delta )p(\delta )=0`$ and $`_1^{\mathrm{}}𝑑\delta \delta b(\delta )p(\delta )=1`$. Several numerical studies have been conducted to study the behavior of $`h(x)`$ and $`b(x)`$ for different initial conditions (e.g. Colombi et al. 1992,1994,1995; Munshi et al. 1999, Valageas et al. 1999). For very small values of $`\delta `$ the behavior of $`p(\delta )`$ is determined by the asymptotic behavior of $`\varphi (y)`$ for large values of $`y`$, and it is possible to define another scaling function $`g(z)`$ which is completely determined by $`\omega `$, the scaling parameter can be expressed as $`z=(1+\delta )a^{1/(1\omega )}\overline{\xi }_2^{\omega /(1\omega )}`$. However numerically it is much easier to determine $`\omega `$ from the study of $`\sigma (y)`$ compared to the study of $`g(z)`$ (e.g. Bouchet & Hernquist 1992).
$$1+\delta <<\overline{\xi }_2;p(\delta )=a^{\frac{1}{1\omega }}\overline{\xi }_2^{\frac{\omega }{1\omega }}\sqrt{\frac{(1\omega )^{1/\omega }}{2\pi \omega z^{(1+\omega )/\omega }}}\mathrm{exp}\left[\omega \left(\frac{z}{1\omega }\right)^{\frac{1\omega }{\omega }}\right];b(\delta )=\left(\frac{2\omega }{\overline{\xi }_2}\right)^{1/2}\left(\frac{1\omega }{z}\right)^{(1\omega )/2\omega }$$
(43)
To summarize, we can say that the entire behavior of the PDF $`P(\delta )`$ is encoded in two different scaling functions, h(x) and g(z) and one can also study the scaling properties of $`b(\delta )`$ in terms of the scaling variables $`x`$ and $`z`$ in a very similar way. These scaling functions are relevant for small and large $`\delta `$ behavior of the function $`p(\delta )`$ and $`b(\delta )`$. Typically the PDF $`p(\delta )`$ shows a cutoff at both large and small values of $`\delta `$ and it exhibits a power-law in the middle. The power law behavior is observed when both $`g(z)`$ and $`h(x)`$ overlap and is typical of highly non-linear regime. With the decrease in $`\overline{\xi }_2`$ the range of $`\delta `$ for which $`p(\delta )`$ shows such a power law behavior decreases finally to vanish for the case of very small variance i.e. in the quasi-linear regime. Similarly the bias is very small and a slowly varying function for moderately over dense objects but increases rapidly for over dense objects which are in qualitative agreement with PS formalism.
#### 3.3.3 The Quasi-linear Regime
In the quasi-linear regime, a similar formalism can be used to study the PDF. However the generating function now can be explicitly evaluated by using tree-level perturbative dynamics (Bernardeau 1992; Bernardeau 1994). It is also possible to take smoothing corrections into account in which case one can have explicit expression of $`\omega `$ in terms of the initial power spectral index $`n`$. In general the parameters $`k_a`$ or $`\omega `$ charecterising VPF or CPDF are different from there highly non-linear values.
For the purpose of weak lensing calculations it is important to notice that although the generating function for the matter correlation hierarchy for very small angular smoothing is the one from the highly non-linear regime, the analytical results that are useful are the ones from the quasi-linear regime as the variance of projected density field remains very small even for small smoothing scales.
The PDF and bias now can be expressed in terms of $`G_\delta (\tau )`$ (Bernardeau 1992, Bernardeau 1994):
$`p(\delta )d\delta ={\displaystyle \frac{1}{G_\delta ^{}(\tau )}}\left[{\displaystyle \frac{1\tau G_\delta ^{\prime \prime }(\tau )/G_\delta ^{}(\tau )}{2\pi \overline{\xi }_2}}\right]^{1/2}\mathrm{exp}\left({\displaystyle \frac{\tau ^2}{2\overline{\xi }_2}}\right)d\tau ;b(\delta )=\left({\displaystyle \frac{k_a}{\overline{\xi }_2}}\right)\left[(1+G_\delta (\tau ))^{1/k_a}1\right],`$ (44)
$`G_\delta (\tau )=G(\tau )1=\delta .`$ (45)
The above expression is valid for $`\delta <\delta _c`$ where the $`\delta _c`$ is the value of $`\delta `$ which cancels the numerator of the pre-factor of the exponential function appearing in the above expression. For $`\delta >\delta _c`$ the PDF develops an exponential tail which is related to the presence of singularity in $`\varphi (y)`$ in a very similar way as in the case of its highly non-linear counterpart (Bernardeau 1992, Bernardeau 1994).
$$p(\delta )d\delta =\frac{3a_s\sqrt{\overline{\xi }_2}}{4\sqrt{\pi }}\delta ^{5/2}\mathrm{exp}\left[|y_s|\frac{\delta }{\overline{\xi }_2}+\frac{|\varphi _s|}{\overline{\xi }_2}\right]d\delta ;b(\delta )=\frac{1}{𝒢^{}(\tau _s)}\frac{(1+\delta )}{\overline{\xi }_2}$$
(46)
In recent studies it was shown that the underlying smoothed density PDF and its convergence counterpart $`p_\kappa (\kappa _s)`$ are exactly same (Munshi & Jain 1999b) under small angle approximation. In this paper we will extend such result to show that even the bias associated with the convergence map can be related to the bias associated with over-dense objects in a very similar manner. Such results can also be obtained from the extended PS formalism and we plan to present details of such analytical results elsewhere.
It may be noted that similar analytical expressions for the PDF and bias can also be derived for the case of approximate dynamics sometime used to simulate gravitational clustering in the weakly non-linear regime or to reconstruct the projected density maps from convergence maps with large smoothing angles (e.g. Lagrangian perturbation theory which is an extension of Zeldovich approximation (Munshi et al. 1994)).
### 3.4 The Bias of the Convergence Field
To compute the bias associated with the peaks in the convergence field we have to fist develop an analytic expression for the generating field $`\beta (y_1,y_2)`$ for the convergence field $`\kappa _s`$. For that we will use the usual definition for the two-point cumulant correlator $`C_{pq}`$ for the convergence field (for a complete treatment of statistical properties of $`\kappa _s`$ see Munshi & Coles, 1999b).
$$C_{pq}^\kappa =\frac{\kappa _s(\gamma _1)^p\kappa _s(\gamma _2)^q}{k_s^2^{p+q2}\kappa _s(\gamma _1)\kappa _s(\gamma _2)}=C_{p1}^\kappa C_{q1}^\kappa $$
(47)
We will show that like its density field counterpart the two-point generating function for the convergence field $`\kappa _s`$ can also be expressed (under certain simplifying assumptions) as a product of two one-point generating functions $`\beta (y)`$ which can then be directly related to the bias associated with “hot-spots”in the convergence field.
$$\beta _\kappa (y_1,y_2)=\underset{p,q}{\overset{\mathrm{}}{}}\frac{C_{pq}^\kappa }{p!q!}y_1^py_2^q=\underset{p}{\overset{\mathrm{}}{}}\frac{C_{p1}^\kappa }{p!}y_1^p\underset{q}{\overset{\mathrm{}}{}}\frac{C_{q1}^\kappa }{q!}y_2^q=\beta _\kappa (y_1)\beta _\kappa (y_2)\tau _\kappa (y_1)\tau _\kappa (y_2)$$
(48)
It is clear that the factorization of generating function actually depend on the factorization property of the cumulant correlators i.e. $`C_{pq}^\eta =C_{p1}^\eta C_{q1}^\eta `$. Note that such a factorization is possible when the correlation of two patches in the directions $`gamma_1`$ and $`\gamma _2`$ $`\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c`$ is smaller compared to the variance $`\kappa _s^2`$ for the smoothed patches
$$\beta _\kappa (y_1,y_2)=\underset{p,q}{\overset{\mathrm{}}{}}\frac{1}{p!q!}\frac{y_1^py_2^q}{\kappa _s^2^{p+q2}}\frac{\kappa _s(\gamma _1)^p\kappa _s(\gamma _2)^q}{\kappa _s(\gamma _1)\kappa _s(\gamma _2)}.$$
(49)
We will now use the integral expression for cumulant correlators (Munshi & Coles 1999a) to express the generating function which in turn uses the hierarchical ansatz and the far field approximation as explained above
$`\beta _\kappa (y_1,y_2)`$ $`=`$ $`{\displaystyle \underset{p,q}{\overset{\mathrm{}}{}}}{\displaystyle \frac{C_{pq}^\eta }{p!q!}}{\displaystyle \frac{1}{\kappa _s^2^{p+q2}}}{\displaystyle \frac{1}{\kappa _s(\gamma _1)\kappa _s(\gamma _2)}}`$ (50)
$`\times {\displaystyle _0^{\chi _s}}d\chi {\displaystyle \frac{\omega ^{p+q}}{r^{2(p+q1)}}}[{\displaystyle }{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}}P({\displaystyle \frac{𝐥}{𝐫(\chi )}})W^2(l\theta _0)\mathrm{exp}[il\theta _{12}]\left[{\displaystyle }{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}}P({\displaystyle \frac{𝐥}{𝐫(\chi )}})W^2(l\theta _0)\right]^{p+q2}y_1^py_2^q.`$
It is possible to further simplify the above expression by separating the summation over dummy variables $`p`$ and $`q`$, which will be useful to establish the factorization property of two-point generating function for bias $`\beta (y_1,y_2)`$.
$`\beta _\kappa (y_1,y_2)`$ $`=`$ $`{\displaystyle _0^{\chi _s}}𝑑\chi {\displaystyle \frac{\left(\frac{1}{r^2(\chi )}\right)\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\mathrm{exp}[il\theta _{12}]\right]}{\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c}}{\displaystyle \frac{\kappa _s^2^2}{\left(\frac{1}{r^2(\chi )}\right)^2\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\right]^2}}`$ (51)
$`\times {\displaystyle \underset{pq}{\overset{\mathrm{}}{}}}{\displaystyle \frac{C_{pq}^\eta }{p!q!}}\left({\displaystyle \frac{y_1}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle }{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}}\mathrm{P}({\displaystyle \frac{𝐥}{𝐫(\chi )}})W^2(l\theta _0)\right]\right)^p\left({\displaystyle \frac{y_2}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle }{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}}\mathrm{P}({\displaystyle \frac{𝐥}{𝐫(\chi )}})W^2(l\theta _0)\right]\right)^q`$
We can now decompose the double sum over the two indices into two separate sums over individual indices. Finally using the definition of the one-point generating function for the cumulant correlators we can write:
$`\beta _\kappa (y_1,y_2)`$ $`=`$ $`{\displaystyle _0^{\chi _s}}𝑑\chi {\displaystyle \frac{\left(\frac{1}{r^2(\chi )}\right)\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\mathrm{exp}[il\theta _{12}]\right]}{\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c}}{\displaystyle \frac{\kappa _s^2^2}{\left(\frac{1}{r^2(\chi )}\right)^2\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\right]^2}}`$ (52)
$`\times \beta _\eta \left({\displaystyle \frac{y_1}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)}\right]\right)\beta _\eta \left({\displaystyle \frac{y_2}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)}\right]\right).`$
The above expression is quite general and depends only on the small angle approximation and the large separation approximation and is valid for any given specific model for the generating function $`G(\tau )`$. However it is easy to notice that the projection effects as encoded in the line of sight integration do not allow us to write down the two-point generating function $`\beta _\kappa (y_1,y_2)`$ simply as a product of two one-point generating functions $`\beta _\eta (y)`$ as was the case for the density field $`1+\delta `$.
As in the case of the derivation of the probability distribution function for the smoothed convergence field $`\kappa _s`$ it will be much easier if we define a reduced smoothed convergence field $`\eta _s`$. The statistical properties of $`\eta _s`$ are very similar to that of the underlying 3D density field (under certain simplifying approximation) and are roughly independent of the background geometry and dynamics of the universe,
$$\eta _s=\frac{\kappa _s\kappa _{min}}{\kappa _{min}}=1+\frac{\kappa _s}{|\kappa _{min}|}.$$
(53)
Where the minimum vale of $`\kappa _s`$ i.e. $`k_{min}`$ is defined as:
$$k_{min}=_0^{\chi _s}𝑑\chi \omega (\chi ).$$
(54)
It is easy to notice that the minimum value of the convergence field will occur in those line of sight which are completely devoid of any matter i.e. $`\delta =1`$ all along the line of sight. We will also find out later that the cosmological dependence of the statistics of $`\kappa _s`$ field is encoded in $`k_{min}`$ and this choice of the new variable $`\eta _s`$ makes its related statistics almost independent of the background cosmology. Repeating the above analysis again for the $`\eta _s`$ field, we can express the cumulant correlator generating function for the reduced convergence field $`\eta _s`$ as:
$`\beta _\eta (y_1,y_2)`$ $`=`$ $`{\displaystyle _0^{\chi _s}}\omega ^2(\chi )𝑑\chi {\displaystyle \frac{\left(\frac{\omega ^2(\chi )}{r^2(\chi )}\right)\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\mathrm{exp}[il\theta _{12}]\right]}{\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c}}{\displaystyle \frac{\kappa _s^2^2}{\left(\frac{\omega (\chi )}{r^2(\chi )}\right)^2\left[\frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)\right]^2\left[_0^{\chi _s}\omega (\chi )𝑑\chi \right]^2}}`$ (55)
$`\times \beta _\eta \left({\displaystyle \frac{y_1}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)}\right]{\displaystyle _0^{\chi _s}}\omega (\chi )𝑑\chi \right)`$
$`\times \beta _\eta \left({\displaystyle \frac{y_2}{\kappa _s^2}}{\displaystyle \frac{\omega (\chi )}{r^2(\chi )}}\left[{\displaystyle \frac{d^2𝐥}{(2\pi )^2)}\mathrm{P}(\frac{𝐥}{𝐫(\chi )})W^2(l\theta _0)}\right]{\displaystyle _0^{\chi _s}}\omega (\chi )𝑑\chi \right).`$
While the above expression is indeed very accurate and relates the generating function of the density field with that of the convergence field, it is difficult to handel for any practical purpose. Also it is important to notice that the scaling functions such as $`h(x)`$ for the density probability distribution function and $`b(x)`$ for the bias associated with over-dense objects are typically estimated from numerical simulations specially in the highly non-linear regime. Such estimations are plagued with several uncertainties such as finite size of the simulation box. It was noted in earlier studies that such uncertainties lead to only a rather approximate estimation of $`h(x)`$. The estimation of the scaling function associated with the bias i.e. $`b(x)`$ is even more complicated due to the fact that the two-point quantities such as the cumulant correlators and the bias are more affected by finite size of the catalogs. So it is not fruitful to actually integrate the exact integral expression we have derived above and we will replace all line of sight integrals with its approximate values. recent study by Munshi & Jain (1999) have used an exactly similar approximation to simplify the one-point probability distribution function for $`\kappa _s`$ and found good agreement with ray tracing simulations. We will show that our approximation reproduces the numerical results quite accurately for a wide range of smoothing angle,
$`|\kappa _{min}|{\displaystyle \frac{1}{2}}\chi _s\omega (\chi _c),`$ (56)
$`\kappa _s^2{\displaystyle \frac{1}{2}}\chi _s\omega ^2(\chi _c)\left[{\displaystyle \frac{d^2k}{(2\pi )^2}}\mathrm{P}(\mathrm{k})W^2(kr(\chi _c)\theta _0)\right],`$ (57)
$`\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c{\displaystyle \frac{1}{2}}\chi _s\omega ^2(\chi _c)\left[{\displaystyle \frac{d^2k}{(2\pi )^2}}\mathrm{P}(\mathrm{k})W^2(kr(\chi _c)\theta _0)\mathrm{exp}[ikr(\chi _c)\theta _{12}]\right].`$ (58)
Use of these approximations gives us the leading order contributions to these integrals and we can check that to this order we recover the factorization property of the generating function i.e. $`\beta _\eta (y_1,y_2)=\beta _\eta (y_1)\beta _\eta (y_2)`$,
$$\beta _\eta (y_1,y_2)=\beta _\eta (y_1)\beta _\eta (y_2)=\beta _{1+\delta }(y_1)\beta _{1+\delta }(y_2)\tau (y_1)\tau (y_2).$$
(59)
So it is clear that at this level of approximation, due to the factorization property of the cumulant correlators, the bias function $`b_\eta (x)`$ associated with the peaks in the convergence field $`\kappa _s`$, beyond certain threshold, obeys a similar factorization property too, which is exactly same as its density field counterpart. Earlier studies have established such a correspondence between convergence field and density field in the case of one-point probability distribution function $`p(\delta )`$ (Munshi & Jain 1999b),
$$b_\eta (x_1)h_\eta (x_1)b_\eta (x_2)h_\eta (x_2)=b_{1+\delta }(x_1)h_{1+\delta }(x_1)b_{1+\delta }(x_2)h_{1+\delta }(x_2).$$
(60)
Where we have used the following relation between $`\beta _\eta (y)`$ and $`b_\eta (x)`$,
$$b_\eta (x)h_\eta (x)=\frac{1}{2\pi i}_i\mathrm{}^i\mathrm{}𝑑y\tau (y)\mathrm{exp}(xy).$$
(61)
For all practical purpose we found that the differential bias as defined above is more difficult to measure from numerical simulations as compared to its integral counterpart where we concentrate on the bias associated with peaks above certain threshold,
$$b_\eta (>x)h_\eta (>x)=\frac{1}{2\pi i}_i\mathrm{}^i\mathrm{}dy\frac{\tau (y)}{y}\mathrm{exp}(xy).$$
(62)
It is important to notice that although the bias $`b(x)`$ associated with the convergence field and the underlying density field is exactly same, the variance associated with the density field is very high but the projection effects in the convergence field brings down the variance in the convergence field to less than unity which indicates that we have to use the integral definition of bias to recover it from its generating function (see eq.(30) and eq.(31)).
Now writing down the full two point probability distribution function for two correlated spots in terms of the convergence field $`\kappa `$ and its reduced version $`\eta `$:
$`p_\kappa (\kappa _1,\kappa _2)d\kappa _1d\kappa _2=p_\kappa (\kappa _1)p_\kappa (\kappa _2)(1+b_\kappa (\kappa _1)\xi _{12}^\kappa b_\kappa (\kappa _2))d\kappa _1d\kappa _2,`$ (63)
$`p_\eta (\eta _1,\eta _2)d\eta _1d\eta _2=p_\eta (\eta _1)p_\eta (\eta _2)(1+b_\eta (\eta _1)\xi _{12}^\eta b_\eta (\eta _2))d\eta _1d\eta _2`$ (64)
In our earlier analysis (Munshi et al. 1999b) we found that $`p^\kappa (\kappa )=\frac{p^\eta (\eta )}{k_{min}}`$ we also noticed that $`\xi _{12}^\kappa =\frac{\xi _{12}^\eta }{\kappa _{min}^{}{}_{}{}^{2}}`$. Using these relations we can now write:
$$b_\kappa (\kappa )=\frac{b_\eta (\eta )}{k_{min}}.$$
(65)
This is one of the main result of our analysis and in the next section we will show that it is indeed a very good approximation to numerical simulations.
## 4 Comparison against Numerical Ray Tracing Simulations
To compare the analytical results with numerical simulations we have smoothed the convergence field or $`\kappa `$\- map generated from numerical simulations using a top-hat filter of suitable smoothing angle $`\theta _0`$. The minimum smoothing radius we have used is $`1^{}`$ which is much larger compared to the numerical resolution length scale. In our earlier studies we found that numerical artifacts are negligible even for angles as small as $`.25^{}`$. The maximum smoothing radius we have studied is $`8^{}`$, which is much smaller compared to the box size of all models and we expect that the finite volume corrections are not significant in our studies. The box size is L = 166.28’ for the EDS models, L =235.68’ for $`\mathrm{\Omega }=0.3`$ open model and L = 209.44’ for $`\mathrm{\Omega }=0.3`$ model with cosmological constant $`\mathrm{\Lambda }=0.7`$. The numerical outputs from more than ten realizations of these cosmological models were used to find the average and the scatter and were tested against the analytical predictions. For LCDM models we have used only one particular relaisation.
For the determination of the bias from the convergence maps we found that the computation of cumulative bias $`b(>\kappa _s)`$ is much more stable compared to the computation of the differential bias $`b(\kappa _s)`$. We have therefore used the analytical predictions for $`b(>\kappa _s)`$ to compare against numerical simulations. The bias we have studied will represent the correlation of those spots in the convergence maps which are above certain threshold and is defined as:
$$b_\kappa (>\kappa _s)=\frac{1}{\sqrt{\kappa _s(\gamma _1)\kappa _s(\gamma _2)_c}}\left(\frac{_{\kappa _s}^{\mathrm{}}𝑑\kappa _1_{\kappa _s}^{\mathrm{}}𝑑\kappa _2p_\kappa (\kappa _1,\kappa _2)}{\left[_{\kappa _s}^{\mathrm{}}p_\kappa (\kappa _1)𝑑\kappa _1\right]^2}1\right)^{1/2}.$$
(66)
The above expression relies on evaluating bias by finding out directly correlation of points in the convergence map which are above certain threshold and does not depend on any factorization property of bias. We have tested the functional form for the bias as predicted from the analytical model against numerical simulations. The analytical prediction tells us that the bias associated with spots where the smoothed convergence field is negative is very small and is almost zero. On the other hand the negative convergence spots are not biased at all, this property also holds for the bias associated with under-dense regions in the density field which are responsible for producing the negative spots in the convergence field. For positive peaks in the convergence map the bias is positive and it increases sharply as we move towards higher thresholds for $`\kappa _s`$. Contributions to positive spots in the convergence maps comes from density peaks in underlying 3D mass distributions and again its remarkable that the bias associated with both of them are actually described by the same functional form. While we change variables from the reduced convergence $`\eta _s`$ to the actual convergence field $`\kappa _s`$ we introduce a multiplicative factor of $`1/k_{min}`$. As we explained before that this term introduces the cosmological dependence for the bias in the convergence field. This multiplicative factor is largest for the OCDM model and is smallest for the SCDM model and hence the bias associated with peaks in the convergence maps also follow the same pattern.
The analytical results and results of numerical simulations are plotted in figures - 3-6. Comparing the analytical results with the results obtained from the numerical simulations we find that these predictions match very accurately. The general trend of increase of bias with $`\kappa _s`$ is well reproduced by the analytical models. Also we find that the analytical model is able to describe the relative ordering of the bias function for different cosmological models quite accurately. We have also computed the scatter in numerical results and found them to increase with $`\kappa _s`$. With increasing the threshold we decrease the number of spots which cross that threshold thereby increasing the sample variance. The sample variance also increases with the smoothing length scale $`\theta _0`$ due to decrease in the number of smoothed patches which carry completely independent information. To improve upon the situation we have increased the sampling by shifting the grid used to bin the convergence map in two perpendicular direction. Although there has been several detailed studies to quantify the effect of finite volume correction on the statistics of galaxy clustering, similar studies are still lacking for weak lensing statistics. However it was found in numerical studies clustering that the computed PDF shows a sharp cutoff at some maximum value of cell count $`n_{max}`$ which is the densest cell in the catalog. Increasing the size of the catalog will increase $`n_{max}`$ up to which one can reliably compute the PDF. The long exponential tail of PDF also shows large fluctuations due to the presence or absence of any rare dense objects in the N-body catalog. In this paper we have shown how intimately statistics of the weak lensing convergence field $`\kappa _s`$ and its density field counterpart are related. This will indicate that the numerical computations of the one-point PDF and the bias of $`\kappa _s`$ and $`\delta `$ are both affected in a very similar manner due to the finite size of the catalog. It is therefore quite remarkable that despite the fact that we have not introduced any involved correction procedure in our measurement of bias or PDF the numerical results match the analytical predictions so accurately. We have not studied the factorization property of the bias as predicted by hierarchical ansatz, as there has not been any parallel study in this direction so far dark matter clustering in 3D. However we plan to extend our results to such analysis elsewhere.
We have studied both analytically and numerically how the bias changes if we change the source redshift $`z_s`$. In our earlier studies of PDF we have shown that for low source redshift when one probes the highly non-linear regime directly, the PDF is charecterised by a long exponential tail originating from collapsed objects. Where as if we increase the source redshift we are adding more slabs of matter in the intervening path which are virtually uncorrelated and hence it makes the distribution more Gaussian. However it also increases the variance of the convergence field. We find a similar result for the bias in $`\kappa _s`$ field too, for low source redshift the highly dense spots which are contributing towards the nearby high $`\kappa _s`$spots may also be physically be very near by and hence they themselves may be correlated more strongly than the underlying mass distribution. On the other hand when we increase the source redshift we increase the number of hight density spots which will produce the high convergence $`\kappa _s`$ spots in the map, however it is to be kept in mind that clearly not all these high density spots are in physical proximity in 3D space and might exist in different slabs of the underlying mass distribution. This explains why although the variance of $`\kappa _s`$ and correlation in $`\kappa _s`$ do increase with the source redshift but the bias associated with high $`\kappa _s`$ spots actually decrease with the source redshift.
In our analytical studies we have assumed that the small angle approximation is valid and we also assumed that the separation angle $`\theta _{12}`$ is much larger than the smoothing angle $`\theta _0`$. This will imply that variance within the smoothed patch is much larger compared to the correlation between the two smoothed patches. However the separation angle should again be small so that we can still replace all spherical harmonics by Fourier modes. In numerical simulations we have placed the smoothed patches separated by an angle at least by thrice the smoothing angle. In almost all cases the separation angle is 3.5 times larger than the smoothing angle. We have also found (as predicted by analytical results) that once the large separation limit is reached the bias function no-longer depends on the separation angle $`\theta _{12}`$.
## 5 Discussion
The hierarchical ansatz provides a framework in which gravitational clustering is generally studied in the highly non-linear regime. However most of such earlier studies are directed towards understanding of galaxy clustering and were tested against numerical N-body simulations. However studies based on clustering properties of galaxies are more difficult to interpret as they always depend on a particular biasing scheme employed to relate galaxy population to the underlying mass fluctuation. Forthcoming weak lensing surveys will provide an unique opportunity to directly study matter distribution in the universe and will help us in understanding their clustering properties. The main motivation of this paper is to extend use of hierarchical ansatz to the study of bias as measured from weak lensing surveys.
There have been several studies in which perturbative techniques were used to relate the convergence statistics measured from weak lensing surveys to the clustering statistics of the underlying mass distribution. However analytical predictions based on perturbative techniques normally requires a large smoothing angle. Most of early observational studies of weak lensing surveys will focus on smaller angular scales and hence will require a completely non-linear treatment. In this paper we have shown that our current level of understanding of gravitationally clustering is enough to make concrete predictions for two-point statistics of the convergence field.
We have used the non-local scaling relations to evolve the matter two-point correlation function and used the hierarchical ansatz to relate the higher order correlation function with two-point correlation function. Combined with the generating function approach these two ansatze provide a powerful tool for the analysis of the weak lensing maps and its multi-point statistics. In accompanying papers we have computed the lower order cumulants and the cumulant correlators for various smoothing angle and for different source redshifts (Munshi & Jain 1999a). We have also studied the complete PDF for the smoothed convergence field (Munshi & Jain 1999b). In this paper we have extended these studies by testing analytical predictions for the bias against numerical simulations. The cumulant correlators are nothing but lower order moments of the bias function which we have studied in this paper. We found a very good agreement between analytical results and ray tracing simulations.
The bias which we have studied here is induced by gravitational clustering. It is known from earlier studies that such a bias occurs naturally both in the quasi-linear (Bernardeau 1994) and the highly non-linear regime for over-dense objects (Bernardeau & Schaeffer 1992, 1999, Munshi et al. 1999a,b). It is factorizable and will only depend on the scaling parameters $`x`$ associated with collapsed objects. We have extended such analytical studies for the weak-lensing convergence field and have shown that similar results do indeed hold even for such cases. Our numerical study confirms such analytical claims and also shows that the functional form for the bias associated with the convergence field is exactly same as that for over-dense spots in underlying mass distribution, so measurement of the bias from weak lensing observations will provide a direct probe into the bias associated with over-dense dark objects in the underlying mass distribution as induced by gravitational clustering. This will help us to separate the contribution to the bias due to gravitational clustering from the bias associated with other non-gravitational sources involved in galaxy formation process.
Although our formalism (which is based on formalism developed by Bernardeau & Schaeffer (1992)) relates the bias of overdense cells in density field with “hot-spots” of the convergence map. Similar results can be obtained by using the extended Press-Schechter formalism to relate the peaks in the convergence field directly with the bias associated with collapsed halos. A detailed analysis will be presented elsewhere.
The hierarchical ansatz not only provides the two-point correlation function (i.e. the bias which we have studied here) for the over-dense objects but it also provides a recipe for computing the higher order correlation functions and the associated cumulants and cumulant correlators. Our formalism which we have developed here can directly be extended to compute such quantities in the highly non-linear regime i.e. at small smoothing angle. However a detailed error analysis is necessary to study the possibility of estimating such quantities from weak lensing observations.
Our analytical results use the top-hat filter, however it was pointed out before that the compensated filter is more suitable for observational studies. We plan to extend our analytical and numerical studies for other filters and results will be presented elsewhere. However it should be kept in mind that although nearby cells do have correlated convergence fields after smoothing with the top-hat window, such correlations are much less significant for the compensated filter (a property which makes it more suitable for the observational purposes for PDF studies).
It is clear that the Born approximation is a necessary ingredient in the analytical computation which we have undertaken in this paper. The Born approximation neglects all the higher order terms in the photon propagation equation. It is then obvious that any statistical quantity such as the higher order cumulants and the cumulant correlators will have correction terms which are neglected in our approach. In the quasi-linear regime i.e. for large smoothing length scales these correction terms can in fact be evaluated using perturbative calculations. However such an approach is not possible to adopt in the highly non-linear regime, because in small angular scales where most contribution comes from length scales with variance well above unity the whole perturbative analysis breaks down and the only way we can check such an approximation is to compare analytical predictions against numerical simulations. Earlier studies of lower order cumulant correlators confirmed that such correction terms are indeed negligible even in small smoothing angle (Munshi & Jain 1999b). A very good match that we have found for the bias function in this study shows that such correction terms are indeed negligible at arbitrary order.
## Acknowledgment
I was supported by a fellowship from Humboldt foundation at MPA when this work was completed. It is pleasure for me to acknowledge many helpful discussions with Bhuvnesh Jain, Francis Bernardeau, Patrick Valageas, Peter Coles and Adrian L. Melott. The complex integration routine I have used to generate $`b(>\eta )`$ was made available to me by Francis Bernardeau. I am grateful to him for his help. The ray tracing simulations were carried out by Bhuvnesh Jain. I am greatly indebted to him for allowing me to use the data which made the present study possible. |
warning/0001/cond-mat0001228.html | ar5iv | text | # Field distributions and effective-medium approximation for weakly nonlinear media
## I Introduction
Numerous papers have been devoted to the problem of determining the effective constitutive law of random nonlinear composites in electrostatics , of importance for understanding various nonlinear phenomena, either entering the class of “weakly nonlinear” (WNL) phenomena, or that of “strongly nonlinear” phenomena such as random fuse-type or dielectric breakdown-type failure of materials .
A convenient approach to the homogeneization problem is the “energetic” one , which can be summarized as follows. Let $`w_x(E)`$ be the local potential (the energy density) from which derives the local nonlinear constitutive law. The homogeneization step consists in computing the effective potential $`W`$ as the volume average
$$W=w_x(E),$$
(1)
from which the effective constitutive law can be deduced (see below). Denoting by $`p_\alpha `$ the volume proportion of component (“phase”) $`\alpha `$ in a composite medium where $`w_x=w_\alpha `$ if $`x`$ is in phase $`\alpha `$, and introducing the volume averages per phase $`_\alpha `$, $`W`$ can be written
$$W=\underset{\alpha }{}p_\alpha w_\alpha (E)_\alpha .$$
(2)
A class of effective-medium theories, mostly concerned with weak power-law nonlinearities, use our ability to compute the second moment of the electric field $`E`$ in the various phases of the composite , completed by a decoupling approximation
$$w_\alpha (E)_\alpha w_\alpha (E^2_\alpha ^{1/2}).$$
(3)
This approximation is exact only when $`w_\alpha `$ is quadratic (the linear theory), or when the field is uniform, which occurs only if all the $`w_\alpha `$ are equal, i.e. in a homogeneous medium.
The purpose of this paper is to present a means to overcome this approximation, starting from the following remark: actually, with an additional hypothesis of ergodicity per phase, a volume phase average can be reinterpreted as an average over some probability distribution for the electric field in the phase under consideration. In this paper, volume averages will henceforth be identified to statistical averages. Then, the problem of computing $`W`$, Equ. (2), reduces to computing the probability distribution $`𝒫_\alpha `$ of the field in each phase, by means of which the phase averages $`_\alpha `$ can be carried out, without appealing to (3).
Note that the full probability distribution $`𝒫(𝐄)`$ of the electric field in the medium is then
$$𝒫(𝐄)=\underset{\alpha }{}p_\alpha 𝒫_\alpha (𝐄).$$
(4)
Thus, the probability distributions of the field per phase are functions of great importance, since they carry all the information needed to build up the effective nonlinear properties. A few analyses of $`𝒫(𝐄)`$ are available , mainly focused on its tail of relevance for breakdown phenomena . However, up to my knowledge, none has been specifically aimed at the individual components $`𝒫_\alpha (𝐄)`$. Chen and Sheng studied numerically the distribution of the local field in binary composites . An analytical model of the same distribution was subsequently proposed by Barthélémy and Orland, with excellent agreement with the latter simulations . Both works were limited to linear composites, in a weak-contrast situation. A first numerical attempt in the continuum, in a strong-contrast situation this time, is due to Cheng and Torquato . They carried out two-dimensional numerical calculations on disordered systems with inclusions of various shapes. An exact calculation for $`𝒫`$ in a linear medium with Hashin-Shtrikman (HS) geometry was recently made using a density of states approach . In these works, the distribution $`𝒫`$ was found to be essentially bimodal for a binary medium (save in the HS geometry where fine-structure peaks are not wiped out by positional disorder), especially at high dielectric contrast . This is understandable in wiew of the decomposition (4).
Still, in parallel with the second moment of the field, the first moment in each phase (the average field) can be computed as well from any linear effective-medium theory, without requiring any a priori knowledge about $`𝒫_\alpha `$. Widely used in mechanics of random continuous media in the framework of the so-called “thermoelastic problem” , this fact seems to have escaped most of the litterature on random dielectric media, excepted in the work by Ponte Castañeda and Kailasam where it is used to compute reference fields around which non-linear potentials are expanded prior to homogeneization. Knowing the first two moments is enough to attempt a modelization of $`𝒫_\alpha `$. The proposal examined in the present paper is to approximate each $`𝒫_\alpha (𝐄)`$ by a gaussian vector distribution, and thereby, to estimate (2) for any potential, without appealing to the strong (and somewhat artificial) decoupling assumption (3).
The scope of the paper is limited however to WNL media, for which it is sufficient to compute the field distribution to linear order only . An extension of these considerations to strongly nonlinear media will be examined elsewhere.
## II General framework
Let us therefore consider a $`d`$-dimensional random medium described by WNL thermodynamic potentials of the type
$$w_x(E)=w_x^{\text{lin}}(E)+w_x^{\text{nl}}(E),$$
(5)
where
$$w_x^{\text{lin}}(E)=\frac{1}{2}\epsilon E^2$$
(6)
($`\epsilon (x)`$ being the local permittivity entering the linear part), and $`w_x^{\text{nl}}(E)`$ is the nonlinear (i.e. non-quadratic) part of the potential, assumed to be small. The local constitutive law is
$$D_i=\frac{w_x(E)}{E_i}.$$
(7)
The potential is prescribed in each phase: $`w_x(E)=w_\alpha (E)`$ if $`x\alpha `$. Under the ergodic hypothesis introduced above, and a “boundary” condition $`𝐄=𝐄^0`$, the effective potential reads
$$W(E^0)=W_{\text{lin}}(E^0)+W_{\text{nl}}(E^0),$$
(8)
where
$`W_{\text{lin}}(E^0)`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha }{}}p_\alpha \epsilon _\alpha E^2_\alpha ,`$ (9)
$`W_{\text{nl}}(E^0)`$ $`=`$ $`{\displaystyle \underset{\alpha }{}}p_\alpha w_\alpha ^{\text{nl}}(E)_\alpha .`$ (10)
The effective constitutive law in the homgeneized medium is then
$$D_i^0=\frac{W(E^0)}{E_i^0}.$$
(11)
To first order in $`w_x^{\text{nl}}`$, the field distributions entering (9) and (10) need only be computed at the linear level . The procedure is the following one. An extended linear anisotropic homogeneization problem is first considered, with an artificial ferroelectric-like part added to the potential :
$$\stackrel{~}{w}_x^{\text{lin}}(E)=\frac{1}{2}E_i\epsilon _{ij}E_j+P_iE_i,$$
(12)
where $`\epsilon _{ij}(x)=\epsilon _{ij}^\alpha `$ and $`P_i(x)=P_i^\alpha `$ if $`x\alpha `$. The corresponding local constitutive law is $`D_i=\epsilon _{ij}E_j+P_i`$. Assuming one knows an estimate for
$$\stackrel{~}{W}_{\text{lin}}(E^0)=\stackrel{~}{w}_x^{\text{lin}}(E),$$
(13)
(cf. below), the following exact relations hold for the distribution $`𝒫_\alpha `$ of the original linear problem (9):
$`E_i_\alpha `$ $`=`$ $`{\displaystyle \frac{1}{p^\alpha }}{\displaystyle \frac{\stackrel{~}{W}_{\text{lin}}}{P_i^\alpha }}|_{\genfrac{}{}{0pt}{}{\epsilon _{ij}=\epsilon \delta _{ij}}{P_i=0}},`$ (15)
$`E_iE_j_\alpha `$ $`=`$ $`{\displaystyle \frac{2}{p^\alpha }}{\displaystyle \frac{\stackrel{~}{W}_{\text{lin}}}{\epsilon _{ij}^\alpha }}|_{\genfrac{}{}{0pt}{}{\epsilon _{ij}=\epsilon \delta _{ij}}{P_i=0}}.`$ (16)
I emphasize that the derivatives are computed at the point $`P_i=0`$, $`\epsilon _{ij}^\alpha =\epsilon ^\alpha \delta _{ij}`$ where $`\epsilon ^\alpha `$ is the scalar permittivity in (5) and (9). There, $`\stackrel{~}{w}_x^{\text{lin}}=w_x^{\text{lin}}`$ and $`\stackrel{~}{W}_{\text{lin}}=W_{\text{lin}}`$. Note that (16) is a slight generalization of the formula
$$E^2_\alpha =\frac{2}{p^\alpha }\frac{W_{\text{lin}}}{\epsilon ^\alpha }.$$
(17)
Both (15) and (16) are particular cases of a more general theorem, the demonstration of which can be found in appendix B of the review by Ponte Castañeda and Suquet on nonlinear composites. This theorem is a consequence of the equality $`D_iE_i=D_iE_i`$ which holds only for particular boundary conditions, or in the infinite-volume limit if there exists an effective permittivity.
Let us set
$`M_i^\alpha `$ $`=`$ $`E_i_\alpha ,`$ (18)
$`C_{ij}^\alpha `$ $`=`$ $`E_iE_j_\alpha M_i^\alpha M_j^\alpha ,`$ (19)
with $`M_i^\alpha M^\alpha \widehat{E}_i^0`$ and $`C_{ij}^\alpha C_{}^\alpha \widehat{E}_i^0\widehat{E}_j^0+C_{}^\alpha (\delta _{ij}\widehat{E}_i^0\widehat{E}_j^0)`$, since the averages only depend on the direction $`\widehat{E}_i^0=E_i^0/E^0`$ of the macroscopic field. I introduce now the gaussian approximation for the probability distributions $`𝒫_\alpha `$
$$𝒫_\alpha (𝐄)=\frac{1}{\left[(2\pi )^ddet(C^\alpha )\right]^{1/2}}e^{\frac{1}{2}(E_iM_i^\alpha )C_{}^{\alpha }{}_{ij}{}^{1}(E_jM_j^\alpha )},$$
(20)
which constitutes the main ingredient of the theory.
There remains to compute an approximation for the quantities $`E_i_\alpha `$ and $`E_iE_j_\alpha `$, in oprder to completely define the above gaussian distributions. Let us therefore estimate $`\stackrel{~}{W}_{\text{lin}}(E^0)`$ through an extension of the Bruggeman self-consistent effective-medium approximation to potentials of the form (12). The formula is given and discussed in Appendix A. Using the scalar permittivities and $`P_i=0`$, this estimate reduces to
$$W_{\text{lin}}(E^0)=\frac{1}{2}\epsilon _0E_{}^{0}{}_{}{}^{2},$$
(21)
where $`\epsilon _0`$ is the usual Bruggeman effective permittivity. In addition, the derivatives (15) and (16) of the estimate for $`\stackrel{~}{W}_{\text{lin}}(E^0)`$ yield
$`M^\alpha `$ $`=`$ $`\mu ^\alpha E^0,`$ (22)
$`E_iE_j_\alpha `$ $`=`$ $`z\mu _{}^{\alpha }{}_{}{}^{2}\left[E_i^0E_j^0{\displaystyle \frac{E_{}^{0}{}_{}{}^{2}}{d+2}}\left(1y\mu ^2\right)\delta _{ij}\right],`$ (23)
where
$`\mu ^\alpha `$ $`=`$ $`{\displaystyle \frac{d\epsilon ^0}{\epsilon ^\alpha +(d1)\epsilon ^0}},`$ (24)
$`y`$ $`=`$ $`{\displaystyle \frac{\epsilon _0}{\mu ^2\epsilon }},`$ (25)
$`z`$ $`=`$ $`{\displaystyle \frac{d+2}{d\mu ^2+2/y}}.`$ (26)
Note that the Bruggeman equation for $`\epsilon _0`$ reads $`\mu =1`$, or alternatively $`\mu \epsilon =\mu \epsilon _0=\epsilon _0`$. From these relations, we check that: (i) $`\epsilon E=\epsilon ^0E`$; (ii) $`\epsilon E^2=\epsilon ^0E^2`$, where $`E=E^0`$, as must be. Expressions for $`C_{}^\alpha `$ and $`C_{}^\alpha `$ are readily obtained from (23).
In Fig. 1 are displayed curves for the probability density $`𝒫_{}(E_{}/E^0)`$ of the normalized longitudinal electric field, $`E_{}=𝐄\widehat{𝐄}^0`$, for a binary medium with dielectric ratio $`\epsilon _2/\epsilon _1=5`$, for various fractions $`p_2`$ of material 2. The space dimension is $`d=2`$. With $`𝒫(E)`$ given by (4), and $`𝐄`$ written as $`𝐄=E_{}\widehat{𝐄}^0+𝐄_{}`$ (which defines $`𝐄_{}`$), $`𝒫_{}`$ is obtained as
$$𝒫_{}(E_{}/E^0)=E^0d^{d1}E_{}𝒫(𝐄)=E^0\underset{\alpha }{}p_\alpha \frac{1}{(2\pi C_{}^\alpha )^{1/2}}e^{\frac{1}{2}\frac{(E_{}M^\alpha )^2}{C_{}^\alpha }},$$
(27)
where $`E_{}`$ is the transverse component of the field: $`𝐄_{}=𝐄𝐄_{}`$. The probability density tends towards a Dirac peak, centered at $`E_{}/E^0=1`$ as $`p_20`$ or $`p_21`$, since the field is then equal to the applied field $`𝐄^0`$. For intermediate volume fractions, the bimodal character of the probability distribution is evident.
Under the same conditions, Fig. 2 displays the probability density of the scaled modulus of the transverse electric field, $`E_{}/E^0`$. This distribution is defined, for $`E_{}>0`$, by
$$𝒫_{}(E_{}/E^0)=E_{}^{0}{}_{}{}^{d1}𝑑E_{}𝑑\mathrm{\Omega }_𝐄_{}𝒫(𝐄)=E^0\underset{\alpha }{}p_\alpha \frac{S_{d1}}{(2\pi C_{}^\alpha )^{(d1)/2}}\left(\frac{E_{}}{E^0}\right)^{d2}e^{\frac{1}{2}\frac{E_{}^2}{C_{}^\alpha }},$$
(28)
where $`S_d=2\pi ^{d/2}/\mathrm{\Gamma }(d/2)`$ is the surface of a $`d`$-dimensional unit sphere. Unlike the previous one, this distribution is not multimodal since the mean value of the transverse field is $`𝐄_{}_\alpha =\mathrm{𝟎}`$ in each phase, whatever the volume fractions.
Fig. 3 displays $`𝒫_{}(E_{}/E^0)`$ for a higher dielectric ratio $`\epsilon _2/\epsilon _1=1000`$. The concentrations are $`p_2=0.12`$, $`0.24`$, $`0.36`$. These plots can be compared to that obtained from hard disks simulations by Cheng and Torquato (CT) in Fig. 3 of Ref. , for the same dielectric contrast and concentrations $`p_2=0.2`$, $`0.4`$, $`0.6`$. These concentrations differ from ours by a factor $`0.6`$. Apart from these differences, the overall features (shape and heights) of the probability distributions are well rendered: a Dirac peak close to $`E_{}=0`$ indicates that the field is almost null in phase $`2`$, and a widely spread contribution from phase 1 for the highest $`p_2`$ indicates an enhancement of the fluctuations in this phase as the percolation threshold (for the Bruggeman theory), or the jamming threshold (for the CT simulations) is approached. The differences in the values for $`p_2`$ can be explained by the fact that the present theory relies on the Bruggeman effective-medium formula, of relevance for cell-materials but not really adequate for hard disks. Moreover, the overall agreement between our distributions and that of CT is not so good at moderate dielectric contrast: Bruggeman’s theory completed by the gaussian approximation somewhat overestimates, because of its percolating nature, the width of the probability distributions. Better adequation with the CT simulations would probably be obtained with an effective-medium theory for dielectric-coated inclusions, which would prevent percolation. Since our main objective however is to discuss an effective-medium therory for WNL composites, such an improvement will not be considered here.
With these results in hand, analytical investigations of the theory can be carried out in some limiting cases, as well as numerical ones in more complicated situations. The next section examines particular WNL potentials.
## III Weakly nonlinear power-law potential
As a first application, a WNL local potential of the form
$$w_x(E)=\frac{1}{2}\epsilon E^2+\frac{1}{4}\chi E^4+O(E^6),$$
(29)
is considered, where the $`E^4`$ term is the first in a series of corrections to the linear behaviour. The corresponding constitutive law is
$$D_i=[\epsilon +\chi E^2+O(E^4)]E_i.$$
(30)
This approximation actually is a weak-field one.
### A Effective-medium formula
The effective potential takes the form
$$W(E^0)=\frac{1}{2}\epsilon ^0E_{}^{0}{}_{}{}^{2}+\frac{1}{4}\chi ^0E_{}^{0}{}_{}{}^{4}+O(E_{}^{0}{}_{}{}^{4}),$$
(31)
where $`\epsilon ^0`$ is the Bruggeman result, and
$$\chi ^0=\chi E^4/E_{}^{0}{}_{}{}^{4}=\underset{\alpha }{}p_\alpha \chi _\alpha E^4_\alpha /E_{}^{0}{}_{}{}^{4}.$$
(32)
The phase averages $`E^4_\alpha `$ are carried out with the help of Wick’s theorem which allows to write down expressions for the integer moments of a centered vector gaussian distribution by mere inspection : with $`A_i=E_iM_i^\alpha `$,
$$A_iA_iA_jA_j_\alpha =A_iA_i_\alpha A_jA_j_\alpha +2A_iA_j_\alpha A_iA_j_\alpha .$$
(33)
We deduce
$`E^4_\alpha `$ $`=`$ $`E_iE_iE_jE_j_\alpha `$ (34)
$`=`$ $`C_{ii}^\alpha C_{jj}^\alpha +2C_{ij}^\alpha C_{ij}^\alpha +2M_{}^{\alpha }{}_{}{}^{2}C_{ii}^\alpha +4M_i^\alpha M_j^\alpha C_{ij}^\alpha +M_{}^{\alpha }{}_{}{}^{4}`$ (35)
$`=`$ $`\left[C_{}^\alpha +(d1)C_{}^\alpha +M_{}^{\alpha }{}_{}{}^{2}\right]^2+2\left[C_{}^{\alpha }{}_{}{}^{2}+(d1)C_{}^{\alpha }{}_{}{}^{2}+2M_{}^{\alpha }{}_{}{}^{2}C_{}^\alpha \right].`$ (36)
A few simplifications yield the simple result
$$\chi ^0=\left[(1+2/d)y^2+(22/d)z^22\right]\mu ^4\chi .$$
(37)
As a first check, we remark that in a non-disordered situation where $`\epsilon `$ and $`\chi `$ are constant in the medium, we have $`\epsilon _0=\epsilon `$ and $`\mu =1`$, so that $`y=z=1`$; whence $`\chi _0=\chi `$, as was expected.
### B Weak-contrast expansion
A further check for the self-consistent formula (37) consists in examining its weak-contrast limit. An exact expression is known for any nonlinear potential, which is first briefly reminded. In the weak-contrast expansion , the local potentials $`w_x(y)`$ are assumed to fluctuate weakly around their mean value $`w^{(0)}(y)=w_x(y)`$. Introducing a bookkeeping parameter $`t`$ to be set to 1 in the final results, the contrast, $`w_x^{(1)}(y)`$, is defined by
$$w_x(y)=w^{(0)}(y)+w_x^{(1)}(y)t,$$
(38)
so that $`w_x^{(1)}(y)=0`$. An expansion for the effective potential is sought for in the form:
$$W(E^0)=W^{(0)}+W^{(1)}t+W^{(2)}t^2+\mathrm{}$$
(39)
Then ,
$$W(E^0)=w(E^0)\frac{1}{2}\frac{n_{}}{\epsilon _{}}\left[_iw_x^{(1)}(E^0)\widehat{E}_{}^{0}{}_{i}{}^{}\right]^2t^2+O(t^3),$$
(40)
where
$`n_{}`$ $`=`$ $`{\displaystyle \frac{d\mathrm{\Omega }_{\widehat{k}}}{S_d}\frac{r(\widehat{}k\widehat{}E^0)^2}{1+(r1)(\widehat{}k\widehat{}E^0)^2}},`$ (41)
$`_i_jw^{(0)}(E^0)`$ $``$ $`\epsilon _{}\widehat{E}_{}^{0}{}_{i}{}^{}\widehat{E}_{}^{0}{}_{j}{}^{}+\epsilon _{}(\delta _{ij}\widehat{E}_{}^{0}{}_{i}{}^{}\widehat{E}_{}^{0}{}_{j}{}^{}),`$ (42)
$`r`$ $`=`$ $`\epsilon _{}/\epsilon _{}.`$ (43)
This result is the exact one . An analogous expansion for the complementary potential $`\stackrel{~}{W}(D^0)`$ defined from the dual homogeneization problem can be written down .
Let us therefore set $`\epsilon =\epsilon +\delta \epsilon t`$ and $`\chi =\chi +\delta \chi t`$. The Bruggeman permittivity $`\epsilon ^0`$ is exact to second order in the constrast:
$$\epsilon ^0=\epsilon \left[1\frac{\delta \epsilon ^2}{d\epsilon ^2}t^2+O(t^3)\right].$$
(44)
Expanding (37) to second order in $`t`$ entails
$$\chi ^0=\chi \left\{1+\left[\frac{2(d+8)}{d(d+2)}\frac{\delta \epsilon ^2}{\epsilon ^2}\frac{4}{d}\frac{\delta \epsilon \delta \chi }{\epsilon \chi }\right]t^2+O(t^3)\right\},$$
(45)
which can directly be obtained from a first-order expansion of (40) in $`\chi `$.
Comparing now formula (37) to the widely used approximation
$$\chi ^0\chi _2^0=\underset{\alpha }{}\chi _\alpha E^2_\alpha ^2/E_{}^{0}{}_{}{}^{4}=y^2\mu ^4\chi ,$$
(46)
we see that the latter is not exact to second order in the contrast, save in $`d=1`$: its expansion indeed reads
$$\chi _2^0=\chi \left\{1+\left[\frac{2(d+2)}{d^2}\frac{\delta \epsilon ^2}{\epsilon ^2}\frac{4}{d}\frac{\delta \epsilon \delta \chi }{\epsilon \chi }\right]t^2+O(t^3)\right\}.$$
(47)
Hence, the gaussian decoupling which accounts for the vector character of the electric field, of importance in nonlinear problems, is superior to the simple approximation $`E^4E^2^2`$.
### C Dilute limit
For a binary medium with two components of constitutive parameters $`(\epsilon _1,\chi _1)`$ and $`(\epsilon _2,\chi _2)`$, the dilute limit is the limiting situation where the volume fraction $`p`$ of (e.g.) component 2 is small. Setting
$$T=\frac{\epsilon _2\epsilon _1}{\epsilon _2+(d1)\epsilon _1},$$
(48)
an expansion, to first order in $`p`$ of (37) yields
$$\chi ^0=\chi _1+\left\{(\chi _2\chi _1)(T1)^4+\chi _1T^2\left[2\frac{d(d+8)}{d+2}4T+T^2\right]\right\}p+O(p^2).$$
(49)
Bergman computed the exact effective nonlinear conductivity of a binary medium in the dilute limit, for spherical inclusions . His result takes the form
$`\chi _{\text{exact}}^0`$ $`=`$ $`\chi _1+\{(\chi _2\chi _1)(T1)^4`$ (50)
$`+`$ $`\chi _1T^2[2{\displaystyle \frac{d(d+8)}{d+2}}+4{\displaystyle \frac{d(d4)}{d+2}}T+{\displaystyle \frac{d(3d^210d+16)}{3(d+2)}}T^2]\}p+O(p^2).`$ (51)
Compared to (51), expression (49) becomes exact when $`\chi _1=0`$ (nonlinear inclusion in a linear host). It also becomes exact for $`d=1`$, and in the limit $`d\mathrm{}`$ where the field is $`E=E_0`$ in each phase so that $`\chi ^0=\chi `$. Moreover, it is exact up to order $`T^2`$, which is consistent with its correct limiting weak-contrast behavior. For $`d=2`$, it is exact up to order $`T^3`$. However, the term of order $`T^4`$ is wrong for $`2d<\mathrm{}`$. The reason for this misbehaviour is that the dilute limit requires an exact computation of $`E^4_\alpha `$ (to linear order) , whereas we approximate it through a gaussian average.
On the other hand, expression (46) yields
$$\chi _2^0=\chi _1+\left\{(\chi _2\chi _1)(T1)^4+\chi _1T^2\left[2(d+2)4T+T^2\right]\right\}p+O(p^2),$$
(52)
once again a result less precise than (49).
At the present time, since it requires an exact solution for the one-body problem, the test of the dilute limit appears to be the most challenging one for nonlinear effective-medium theories.
This test is illustrated in Fig. 4 which displays $`\chi ^0`$, as calculated from (37) and (46), against the concentration $`p`$ of medium 2, for moderate dielectric contrast and $`d=2`$. The thick line segments at $`p=0,1`$ represent exact tangents obtained from (51). The tangent at $`p=1`$ follows from substituting $`(1p)`$ for $`p`$ and interchanging the indices $`1`$ and $`2`$ in (51). Though not exact in the dilute limit, formula (37) yields tangents quite close to the exact ones. The marked inaccuracy of formula (46) near $`p=0`$ and $`p=1`$ results in a lower value for $`\chi ^0`$ in the whole concentration range.
### D Percolative behavior
Before examining the predictions of Equ. (37) near the percolation transition, I first briefly review the critical behaviour of WNL composites, and discuss a flaw of the Bruggeman formula in this context.
An insulator/perfect conductor binary mixture undergoes a percolation transition for a critical metal fraction $`p=p_c`$ . For WNL phases, the critical behaviour of the field fluctuations is now well understood . In the limiting situation where $`\epsilon _1\epsilon _2`$, and for $`pp_c`$, one observes a behaviour
$$\epsilon ^0\epsilon _1(pp_c)^s,\chi ^0\chi _1(pp_c)^{(2s+\kappa ^{})};$$
(53)
on the contrary, for $`p>p_c`$, one has
$$\epsilon ^0\epsilon _2(p_cp)^t,\chi ^0\chi _2(p_cp)^{(2t\kappa )},$$
(54)
where $`s`$ and $`t`$ are the superconductivity and conductivity exponents, and where $`\kappa `$ and $`\kappa ^{}`$ are the noise exponents which characterize an anomalous nonlinear susceptibility enhancement near the threshold . As a consequence of the exact inequality $`E^4E^2^2`$, the exponents $`\kappa `$ and $`\kappa ^{}`$ are necessarily positive.
In the Bruggeman formula, $`p_c=1/d`$, and $`\epsilon _0\epsilon _1/(1dp)`$ for $`p<1/d`$, $`\epsilon _0\epsilon _2(1dp)/(1d)`$ for $`p>1/d`$ ($`\epsilon _1\epsilon _2`$), whence $`s=t=1`$. Reporting these expressions into (37) yields $`\kappa =\kappa ^{}=0`$, a result shared by all effective-medium formulae based on (17).
These values can be compared to exact bounds obtained for $`\kappa `$ and $`\kappa ^{}`$ on the basis of the links-nodes-blob (LNB) model for electric percolation, now commonly accepted as a model for e.g. random resistor networks (RRNs). These are
$`(3d4)\nu 2t+1\kappa 2(d1)\nu t,`$ (56)
$`(4d)\nu 2s+1\kappa ^{}2\nu s,`$ (57)
(for $`d6`$; the values for $`d>6`$ are that of $`d=6`$. Moreover $`t`$ is not defined for $`d=1`$), where $`\nu `$ is the correlation length exponent: $`\xi |pp_c|^\nu `$. Though the latter information is absent from the Bruggeman theory, an inequality $`\nu >0`$ must hold in any situation relevant to percolating systems. In (III D), the lower bounds must not be greater than the upper ones, which leads to lower bounds for the exponents $`s`$ and $`t`$ themselves, namely
$`s(2d)\nu +1,`$ (58)
$`t(d2)\nu +1.`$ (59)
In the Bruggeman theory, $`s=t=1`$ so that only $`\nu =0`$ – a problematic value by itself – is allowed, save for $`d=2`$. Reporting $`\nu =0`$ into (III D) leads to $`\kappa =\kappa ^{}=1`$, which is unacceptable. For $`d=2`$, one obtains $`\kappa =\kappa ^{}=2\nu 1`$. Compared to the values $`\kappa =\kappa ^{}=0`$ of the above effective-medium theory, this gives $`\nu =1/2`$, a non-conflicting value, though lower than the exact result $`4/3`$.
Of course, the fact that Bruggeman’s formula is a poor approximation near the percolation threshold is well-known. However, this discussion enlightens a fundamental inconsistency in the Bruggeman exponent values which goes beyond mere numerical inadequacy, as long as one wishes to estimate the properties of systems obeying the LNB scheme. It is to be noted that for $`d=2`$, where no inconsistency appears, the exact equality $`s=t`$ holds (a consequence of self-duality) and is obeyed by the Bruggeman exponents.
The above argument is to be brought together with another one by Bergman. Discussing the failure of a “non ambiguous” non-linear Bruggeman-type model, he concluded that bounds for fluctuations always refer to some particular type of microstructure, and that no such bounds exist which apply to all materials .
Because of these problems, comparisons between effective-medium theories built on the Bruggeman formula, and simulations on RRNs are expected to be more significant in dimension $`d=2`$ (another reason for preferring the two-dimensional case as a test-bench is that the bond percolation threshold $`p_c=1/2`$ on a square lattice is exactly reproduced by Bruggeman’s formula). In Fig. 5 are displayed formulas (37) and (46) against $`p`$, in a high contrast situation ($`\epsilon _2/\epsilon _1=10000`$). The same trend as in Fig. 4 is observed: formula (37) yields a higher estimate. The values for $`\epsilon _{1,2}`$ and $`\chi _{1,2}`$ are the same as that used in Fig. 3 of a paper by Levy and Bergman , where simulations results on RRNs are reported, and compared to (46). The authors remarked that the height of the peak in $`\chi ^0`$ at $`p_c`$ was badly underestimated by (46), which gives $`\chi _{\text{peak}}^00.5`$ only. In their numerical simulations, for a $`30\times 30`$ network, the peak height is found to be $`\chi _{\text{peak}}^00.86`$. In the infinite-size limit, this value is expected to increase somewhat. In Fig. 5, the height of the peak given by (37) is $`\chi _{\text{peak}}^01.0`$, which thus compares fairly well to simulations.
### E Other power-law nonlinearities
In this section, potentials of the type (5) where
$$w_x^{\text{nl}}(E)=\frac{\chi (x)}{\gamma +1}E^{\gamma +1}$$
(60)
are considered. For simplicity, $`\gamma `$ is assumed to be constant in the material. The gaussian averages over the field in each phase are carried out numerically, with the help of a bi-variate integration routine. One integration variable is the field modulus $`E`$, and the other is the cosine $`𝐄𝐄^0/(EE^0)`$. In fig. 6 are shown plots for
$$\chi ^0=\chi (E/E^0)^{\gamma +1},$$
(61)
computed from gaussian averages, for a moderate dielectric contrast and various powers $`\gamma >1`$ ranging from $`\gamma =1.5`$ to $`\gamma =5`$, against the volume fraction $`p`$ of medium 2 (solid lines). An exponential enhancement of $`\chi ^0`$ is observed with increasing $`\gamma `$. Its origin lies in the existence of a non-zero probability for $`E>E^0`$ in the composite, since $`𝐄^0=𝐄`$ by definition. These fluctuations are amplified by the nonlinearity, while that for which $`E<E^0`$ are reduced. The larger the width of the probability distribution, the larger this enhancement, so that the peak culminates in the region $`pp_c=1/2`$. Once again, the estimates of the present theory are much higher than that predicted by the decoupling assumption
$$\chi _2^0=\underset{\alpha }{}p_\alpha \chi _\alpha (E/E^0)^2_\alpha ^{(\gamma +1)/2},$$
(62)
(dashed lines), especially for large values of $`\gamma `$. This is understandable, since the functions to be averaged can be written $`w_x^{\text{nl}}(E)=\chi (x)f(E^2)`$, where $`f(z)=z^{(1+\gamma )/2}/(1+\gamma )`$. For $`\gamma >1`$, these functions are convex, so that $`f(z)f(z)`$. Therefore, the decoupling assumption always underestimates the fluctuations when applied to such potentials. An overestimation would instead take place with concave potentials.
## IV Conclusion
In this article, a theory for the nonlinear susceptibility of weakly nonlinear composites is proposed. This theory is based on a multimodal gaussian approximation for the overall probability distribution of the electric field, to linear order. The parameters which define this distribution (means and second moments, in each phase of the disordered medium), are obtained from Bruggeman’s theory, whose limitations are discussed in this context. The present model provides for the first time an analytical estimate for the probability distribution of the electric field, for arbitrarily high dielectric contrast, in percolating media. The resulting effective nonlinear susceptibility is exact to second order in the contrast, and close to the exact result in the dilute limit. Significant quantitative improvement is obtained on previous nonlinear effective-medium theories, even in the percolating regime, at least in the two-dimensional case. This study emphasizes the importance of accounting for the vector character of the electric field when averaging the local potentials. Improvements of the present theory could consist in finding a better approximation than the gaussian one for the components of the probability distribution of the field. More realistic distributions could be obtained by, e.g., extending the perturbative theory of Barthélémy and Orland to high dielectric contrasts via self-consistent effective-medium approximations.
###### Acknowledgements.
Stimulating discussions with M. Barthélémy and H. Orland are gratefully acknowledged. I also thank H. E. Stanley and M. Barthélémy for their kind hospitality at the Center for Polymer Studies (Boston University), where part of this work was performed.
## A Self-consistent anisotropic linear theory with permanent polarization
With local potentials of the form (12), the homogeneized potential reads
$$\stackrel{~}{W}_{\text{lin}}(E^0)=\frac{1}{2}E_i^0\epsilon _{ij}^0E_j^0+P_i^0E_j^0+\frac{1}{2}\mathrm{\Delta }P_i\mu _{ij}g_{jk}\mathrm{\Delta }P_j,$$
(A1)
where
$`g_{ij}`$ $`=`$ $`{\displaystyle \frac{d\mathrm{\Omega }_{\widehat{𝐤}}}{S_d}\frac{k_ik_j}{k_l\epsilon _{lm}^0k_m}},`$ (A2)
$`\mu _{ij}^\alpha `$ $`=`$ $`[1g(\epsilon ^\alpha \epsilon ^0)]_{ij}^1,`$ (A3)
$`P_i^0`$ $`=`$ $`P_j\mu _{ji},`$ (A4)
$`\mathrm{\Delta }P_i^\alpha `$ $`=`$ $`P_i^\alpha P_i^0,`$ (A5)
and the Bruggeman condition for $`\epsilon _{ij}^0`$ is $`\mu _{ij}=\delta _{ij}`$.
Formula (A1) has been derived by means of a functional formalism , used with a trial potential of the form $`W^0(E)=(1/2)E_i\epsilon _{ij}^0E_j+P_i^0E_i+c^0`$. Rather than to give a demonstration which would complicate the present article, the validity of (A1) is established by looking at its consequences. First, one can easily check that, for a binary medium, (A1) exactly reduces to formula (2.16) in the work by Ponte Castañeda and Kailasam . Next, the average field per phase, $`E_i_\alpha =M_i^\alpha `$, is (using the symmetry of the tensors $`g`$ and $`\mu g`$; $`\mu `$ itself is not necessarily symmetric)
$$M_i^\alpha =\frac{1}{p_\alpha }\frac{\stackrel{~}{W}_{\text{lin}}(E^0)}{P_i^\alpha }=\mu _{ij}^\alpha (E_j^0+g_{jk}\mathrm{\Delta }P_k^\alpha ).$$
(A6)
From the definition of $`P_i^0`$, one deduces that $`M=E^0`$, in agreement with the boundary conditions. Finally, the macroscopic consitutive relation derived from the effective potential, namely
$$D_i^0=\frac{\stackrel{~}{W}_{\text{lin}}(E^0)}{E_i^0}=\epsilon _{ij}^0E_j^0+P_i^0,$$
(A7)
is consistent with the equation
$$D_i^0=\epsilon _{ij}E_j+P_i=\epsilon _{ij}^\alpha M_j^\alpha +P_i^\alpha ,$$
(A8)
with $`M_j^\alpha `$ computed with (A6). This directly follows from the equivalent form of the Bruggeman equation, $`\epsilon _{ij}\mu _{jk}=\epsilon _{ij}^0\mu _{jk}=\epsilon _{ij}^0`$, and from the identity $`\mu _{ij}^\alpha =\delta _{ij}+\mu _{ik}^\alpha g_{kl}(\epsilon _{li}^\alpha \epsilon _{li}^0)`$. |
warning/0001/hep-th0001148.html | ar5iv | text | # References
1. Introduction
Within the framework of Connes’ noncommutative geometry , the Higgs field and the symmetry breaking mechanism in the standard model have a remarkable geometrical picture. The Higgs field is a connection, which arises from the geometry of the two-point space . (See also , and the references therein.) This suggests that the discrete geometry may play an important role in physics. Differrential calculus and gauge theory on discrete groups was proposed in . The systematic approach towards the differential calculus and the gauge theory on arbitrary finite or countable sets was formulated in . It is natural to ask: what is the explicit form of the action functional of gauge fields on the $`n`$-point space? Here $`n`$ is the natural number ($`n2`$). When $`n=2`$, the action is just the case of Higgs field. The purpose of the present paper is to answer this question for the simplest case, i.e., the gauge group is taken to be $`U(1)`$.
In Section 2 we will review differential calculus on $`n`$-point space which is formulated by Dimakis and Müller-Hoissen , and then in Section 3 we will construct the spectral triple on $`n`$-point space and derive the explicit form of action functional of $`U(1)`$ gauge fields. In section 4 the action funcional over $`\{`$space-time$`\}`$ $`\times `$ $`\{n`$-point space$`\}`$ is obtained. In the concluding section, we comment on the physical meaning of the action functional we derived and its relation to the theory of phase transitions.
2. Differential calculus on $`n`$-point space
In this section we shall review the differential calculus on $`n`$-point space. More detailed account of the construction can be found in .
Let $`M`$ be a set of $`n`$ points $`i_1,\mathrm{},i_n`$ $`(n<\mathrm{})`$, and $`𝒜`$ the algebra of complex functions on $`M`$ with $`(fg)(i)=f(i)g(i)`$. Let $`p_i𝒜`$ defined by
$$p_i(j)=\delta _{ij}.$$
(1)
Then it follows that
$$p_ip_j=\delta _{ij}p_j,\underset{i}{}p_i=1.$$
(2)
In other words, $`p_i`$ is a projective operator in $`𝒜`$. Each $`f𝒜`$ can be written as
$$f=\underset{i}{}f(i)p_i,$$
(3)
where $`f(i)𝐂`$, a complex number. The algebra $`𝒜`$ can be extended to a universal differential algebra $`\mathrm{\Omega }(𝒜)`$. The differentials satisfy the following relations:
$`p_idp_j`$ $`=`$ $`(dp_i)p_j+\delta _{ij}dp_i,`$ (4)
$`{\displaystyle \underset{i}{}}dp_i`$ $`=`$ $`0.`$ (5)
This means that the differential calculus over $`n`$-point space $`M`$ associates with it $`n1`$ linear independent differentials. There is a natural geometrical representation associated with $`M`$. Let the projective operators $`p_i`$ $`(i=1,\mathrm{},n)`$ be the orthonormal base vectors in the Euclidean space $`𝐑^n`$. Then $`M`$ forms the vertices of the $`(n1)`$-dimensional hypertetrahedron embeded in $`𝐑^n`$.
The universal first order differential calculus $`\mathrm{\Omega }^1`$ is generated by $`p_idp_j(ij)`$, $`i,j=1,2,\mathrm{},n`$. Notice that $`p_idp_i`$ is the linear combinations of $`p_idp_j(ij)`$.
Similarly, the compositions of $`p_idp_j(ij)`$, $`i,j=1,2,\mathrm{},n`$ generate the higher order universal differential calculus on $`M`$. For example, the universal second order differential calculus $`\mathrm{\Omega }^2`$ is generated by $`p_idp_jp_jdp_k(ij,jk)`$, $`i,j,k=1,2,\mathrm{},n`$.
A simple calculation shows that
$$dp_i=\underset{j}{}(p_jdp_ip_idp_j).$$
(6)
Furthermore,
$$dp_idp_j=\underset{k}{}(p_kdp_ip_idp_jp_idp_kp_kdp_j+p_idp_jp_jdp_k).$$
(7)
Any $`1`$-form $`\alpha `$ can be written as $`\alpha =_{i,j}\alpha _{ij}p_idp_j`$ with $`\alpha _{ij}𝐂`$ and $`\alpha _{ii}=0`$. One then finds
$$d\alpha =\underset{i,j,k}{}(\alpha _{jk}\alpha _{ik}+\alpha _{ij})p_idp_jp_jdp_k.$$
(8)
Now we consider a connection $`\alpha `$ on $`M`$, $`\alpha `$ is a $`1`$-form, and skew -adjoint, $`\alpha ^{}=\alpha `$. $`\alpha `$ obeys the usual transformation rule,
$$\alpha ^{}=u\alpha u^{}+udu^{}.$$
(9)
Here $`u=_iu(i)p_i𝒜`$, and $`u(i)U(1)`$, the Abelian unitary group. $`\alpha `$ is thus called the $`U(1)`$ gauge field on $`M`$. In order to make the formulae concise, we introduce
$$a=\underset{i,j}{}(1+\alpha _{ij})p_idp_j.$$
(10)
Notice that
$$a_{ii}=1.$$
(11)
One then has
$`a^{}`$ $`=`$ $`uau^{},`$
$`a_{ij}^{^{}}`$ $`=`$ $`u(i)a_{ij}u(j)^{}.`$ (12)
The curvature of the connection $`\alpha `$ is given by
$$\theta =d\alpha +\alpha ^2,$$
(13)
and transforms in the usual way, $`\theta ^{}=u\theta u^{}`$. As a $`2`$-form, $`\theta `$ can be written as
$`\theta `$ $`=`$ $`{\displaystyle \underset{i,j,k}{}}\theta _{ijk}p_idp_jp_jdp_k,`$
$`\theta _{ijk}`$ $`=`$ $`a_{ij}a_{jk}a_{ik}.`$ (14)
3. From spectral triple to action functional over $`M`$
In noncommutative geometry, all the geometrical data is determined by a spectral triple $`(𝒜,,D)`$, where $`𝒜`$ is an involutive algebra, $``$ is a Hilbert space with an involutive representation $`\pi `$ of $`𝒜`$, $`D`$ is a self-adjoint operator acting on $``$.
We now construct the spectral triple $`(𝒜,,D)`$ over the $`n`$-point space $`M`$. In our case, $`𝒜`$ is the algebra on $`M`$ defined in the last section. Without loss of generality, $``$ is taken to be the $`n`$-dimensional linear space over $`𝐂`$, i.e., $``$ is just the direct sum $`=_{i=1}^n_i`$, $`_i=𝐂`$. The action of $`𝒜`$ is given by
$$f𝒜\left(\begin{array}{cccc}f(1)& 0& \mathrm{}& 0\\ 0& f(2)& \mathrm{}& 0\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ 0& 0& \mathrm{}& f(n)\end{array}\right).$$
(15)
Then $`D`$ is the Hermitian $`n\times n`$ matrix with elements $`D_{ij}=D_{ji}^{}`$, and $`D_{ij}`$ is a linear mapping from $`_j`$ to $`_i`$. We have the following equality defines the involutive representation of $`da`$ ($`a𝒜)`$ in $``$,
$$\pi (da)=i[D,\pi (a)].$$
(16)
To ensure the differential $`d`$ satisfies
$$d^2=0,$$
(17)
one has to impose the following condition on $`D`$,
$$D^2=\mu ^2I,$$
(18)
where $`\mu `$ is a real constant and $`I`$ is the $`n\times n`$ unit matrix.
We now take $`D_{ij}0`$ $`(ij).`$ Then the representation $`\pi :\mathrm{\Omega }(𝒜)()`$ is injective on $`\mathrm{\Omega }^1(𝒜)`$.
In Connes’ language , Our spectral triple $`(𝒜,,D)`$ is odd (except the case of $`n=2`$). For the sake of convenience, we omit the symbol $`\pi `$ from now on.
The projective operator $`p_i`$ can be expressed as the $`n\times n`$ matrix now,
$$(p_i)_{\alpha \beta }=\delta _{\alpha i}\delta _{\beta i}.$$
(19)
Notice that $`D`$ commute exactly with the action of $`𝒜`$, for the sake of convenience, we can ignore the diagonal elements of $`D`$, i. e.,
$$D_{ii}=0.$$
(20)
From (16) and (19), one has
$`(p_idp_j)_{\alpha \beta }`$ $`=`$ $`i\delta _{\alpha i}\delta _{\beta j}D_{ij},`$ (21)
$`(p_idp_jp_jdp_kp_kdp_lp_ldp_r)_{\alpha \beta }`$ $`=`$ $`\delta _{\alpha i}\delta _{\beta r}D_{ij}D_{jk}D_{kl}D_{lr}.`$ (22)
One can define an inner product $`<|>`$ in $`\mathrm{\Omega }^{}𝒜`$ by setting
$$<\alpha |\beta >=tr_\omega (\alpha ^{}\beta |D|^p).$$
(23)
where $`tr_\omega `$ is the Dixmier trace. In our case $`p=0`$ and $`tr_\omega `$ is reduced to the ordinary trace,
$$<\alpha |\beta >=tr(\alpha ^{}\beta ).$$
(24)
Since the gauge field $`\alpha `$ satisfies $`\alpha ^{}=\alpha `$, one has $`\theta ^{}=\theta `$. Then the action functional of $`\theta `$ is
$$S=\theta ^2=<\theta |\theta >=tr\theta ^2.$$
(25)
From (14), (22) and (25), we have
$$S=\underset{i,j,k,l}{}\theta _{ijk}\theta _{kli}D_{ij}D_{jk}D_{kl}D_{li}.$$
(26)
Denote
$$a_{ij}D_{ij}=H_{ij},$$
(27)
where $`a_{ij}`$ is defined in (10). Then $`H=(H_{ij})`$ is a Hermitian matrix with
$$H_{ii}=0.$$
(28)
From (14), (18), (26), (27) and (28), one thus has
$$S=trH^42\mu ^2trH^2+n\mu ^4.$$
(29)
From (28), the eigenvalues $`\lambda _i(i=1,2,\mathrm{},n)`$ of $`H`$ satisfy:
$$\underset{i=1}{\overset{n}{}}\lambda _i=0.$$
(30)
(29) can be written as the following:
$$S=\underset{i=1}{\overset{n}{}}\lambda _i^42\mu ^2\underset{i=1}{\overset{n}{}}\lambda _i^2+n\mu ^4.$$
(31)
The eigenvalues of such kind of $`H`$’s generate $`(n1)`$-dimensional Euclidean space $`𝐑^{n1}`$. By the quadric transformation, (31) can be taken the form as
$$S=C_2^{ijkl}\underset{i,j,k,l=1}{\overset{n1}{}}\phi _i\phi _j\phi _k\phi _lC_1\underset{i=1}{\overset{n1}{}}\phi _i^2+n\mu ^4,$$
(32)
where $`(\phi _1,\mathrm{},\phi _{n1})`$ is a vector in $`𝐑^{n1}`$, $`C_2^{ijkl}`$ and $`C_1`$ are real constants.
For the sake of convenience, we identify $`𝐑^{n1}`$ with a subspace embedded in the $`n`$-dimensional geometrical representation space of $`M`$ introduced in Section 2 from now on. In the $`(n1)`$-dimensional rectangular coordinate system, the reference point is taken to be the center of the $`(n1)`$-dimensional hypertetrahedron. $`M`$ can then be represented by a set of $`n`$ vectors in $`𝐑^{n1}`$: $`e_i^\alpha `$ ($`\alpha =1,\mathrm{},n`$; $`i=1,\mathrm{},n1`$), such that
$$\underset{i}{}e_i^\alpha e_i^\beta =\frac{n}{n1}\delta ^{\alpha \beta }\frac{1}{n1}.$$
(33)
In (33) we have chosen the normalization of the vectors to be unity for convenience. This set of $`e`$’s satisfy
$`{\displaystyle \underset{\alpha }{}}e_i^\alpha `$ $`=`$ $`0;`$ (34)
$`{\displaystyle \underset{\alpha }{}}e_i^\alpha e_j^\alpha `$ $`=`$ $`{\displaystyle \frac{n}{n1}}\delta _{ij}.`$ (35)
It should be mentioned that the properties of $`M`$ is encoded in those of the set of spin states in the Potts model . The reason will be discussed in section 5. Using (34), the eigenvalues of $`H`$ is
$$\lambda _\alpha =\underset{i=1}{\overset{n1}{}}\varphi _ie_i^\alpha ,\alpha =1,\mathrm{},n.$$
(36)
Here $`\varphi _i(i=1,\mathrm{},n1)`$ is a real parameter. We call $`𝚽=(\varphi _1,\mathrm{},\varphi _{n1})`$ the order parameter field in $`𝐑^{n1}`$. Finally from (29), (35) and (36), we obtain the explicit form of $`S`$ over the $`n`$-point space $`M`$:
$$S=\underset{i,j,k,l}{}(\underset{\alpha }{}e_i^\alpha e_j^\alpha e_k^\alpha e_l^\alpha )\varphi _i\varphi _j\varphi _k\varphi _l\frac{2n}{n1}\mu ^2(\underset{i}{}\varphi _i^2)+n\mu ^4.$$
(37)
Notice that the constant term $`n\mu ^4`$ can be left out.
When $`n=2`$, (37) changes into
$$S=2(\varphi ^2\mu ^2)^2.$$
(38)
This is just the Hamiltonian density of the Landau phenomenological theory of phase transitions below the critical temperature . Here $`\varphi `$ is known as the order parameter. Notice that the size of the coefficients of $`\varphi ^2`$ and $`\varphi ^4`$ does not affect the values of critical exponents of phase transitions, but it may affect the mass value of the Higgs field when (38) is considered as the Higgs potential: From (18), (27) and (36), we have $`\varphi ^2=\mu ^2|a_{12}|^2`$. One then has
$$S=2\mu ^4(|a_{12}|^21)^2,$$
(39)
which is the form of Connes’ version of Higgs potential.
4. Action functional over $`V\times M`$
Now we construct the $`U(1)`$ gauge field theory over $`\{`$space-time$`\}`$ $`\times `$ $`\{n`$-point space$`\}`$. Denote space-time by $`V`$, $`𝒜`$ the algebra of complex functions on $`V\times M`$. Just as in Section 2, Each $`f𝒜`$ can be written as
$$f=\underset{i}{}f(i)p_i.$$
(40)
Notice that this time $`f(i)`$ is a complex function over $`V_i`$, the $`ith`$ copy of $`V`$. $`𝒜`$ can be also extended to a universal differential algebra $`\mathrm{\Omega }(𝒜)`$. Denote the differential on $`M`$ by $`d_f`$. In other words, The differential $`d`$ in Section 2 and Section 3 is replaced by $`d_f`$. Let $`d_s`$ be the differential on $`V`$, and $`d`$ the total differential on $`V\times M`$. One then has
$$d=d_s+d_f.$$
(41)
The nilpotency of $`d`$ requires that
$$d_sd_f=d_fd_s.$$
(42)
Differentiate (40), we have
$$df=\underset{i}{}(d_sf(i))p_i+\underset{i}{}f(i)d_fp_i.$$
(43)
Any $`1`$-form $`\alpha `$ can be written as
$$\alpha =\underset{i,j}{}\alpha _{ij}p_id_fp_j+\underset{i}{}\alpha _ip_i,$$
(44)
with $`\alpha _{ij}`$, the complex function on $`V`$ and $`\alpha _{ii}=0`$; $`\alpha _i`$, the $`1`$-form on $`V_i`$. One then finds
$`d\alpha `$ $`=`$ $`{\displaystyle \underset{ij}{}}(d_s\alpha _{ij})p_id_fp_j+{\displaystyle \underset{i,j,k}{}}(\alpha _{jk}\alpha _{ik}+\alpha _{ij})p_id_fp_jp_jd_fp_k`$
$`+{\displaystyle \underset{i}{}}(d_s\alpha _i)p_i{\displaystyle \underset{i}{}}\alpha _id_fp_i.`$
Now we consider a connection $`\alpha `$ on $`V\times M`$, $`\alpha `$ is a $`1`$-form and skew-adjoint, i.e., $`\alpha `$ is given by (44) and $`\alpha ^{}=\alpha `$. $`\alpha `$ obeys the usual transformation rule,
$$\alpha ^{}=u\alpha u^{}+udu^{}.$$
(46)
Here $`u=_iu(i)p_i𝒜`$, and $`u(i)U(1)`$, the Abelian unitary group on $`V_i`$. $`\alpha `$ is thus called the $`U(1)`$ gauge field on $`V\times M`$. In this section, we only consider the simpliest case, i.e.,
$`\alpha _i`$ $`=`$ $`A,`$ (47)
$`u(i)`$ $`=`$ $`u(j)(i,j=1,2,\mathrm{},n).`$ (48)
Here $`A`$ is a $`U(1)`$ gauge field on $`V`$. The physical meaning of the above asumptions is: there exist unique gauge field, i.e., the Maxwell electromagnetic field over all copies of $`V`$.
$`\alpha `$ can then be written by
$$\alpha =\underset{i,j}{}\alpha _{ij}p_id_fp_j+A.$$
(49)
As in Section 2, we introduce
$$a=\underset{i,j}{}(1+\alpha _{ij})p_id_fp_j.$$
(50)
Notice that
$$a_{ii}=1.$$
(51)
One then has
$`a^{}`$ $`=`$ $`uau^{},`$
$`a_{ij}^{^{}}`$ $`=`$ $`u(i)a_{ij}u(j)^{}.`$ (52)
$`A`$ obeys the usual $`U(1)`$ gauge transformation rule,
$$A^{}=A+udu^{}.$$
(53)
The curvature of the connection $`\alpha `$ is given by
$$\mathrm{\Theta }=d\alpha +\alpha ^2.$$
(54)
It can be seen that $`\mathrm{\Theta }`$ transforms in the usual way, $`\mathrm{\Theta }^{}=u\mathrm{\Theta }u^{}`$. As a $`2`$-form, $`\mathrm{\Theta }`$ can be written as
$`\mathrm{\Theta }`$ $`=`$ $`d_sA+{\displaystyle \underset{ij}{}}(d_sa_{ij})p_id_fp_j+{\displaystyle \underset{i,j,k}{}}\theta _{ijk}p_id_fp_jp_jd_fp_k,`$
$`\theta _{ijk}`$ $`=`$ $`a_{ij}a_{jk}a_{ik}.`$ (55)
We see that $`\mathrm{\Theta }`$ has a usual differential degree and a finit-difference degree $`(\alpha ,\beta )`$ adding up to $`2`$. Let us begin with the term in $`\mathrm{\Theta }`$ of bi-degree $`(2,0)`$:
$$\mathrm{\Theta }^{(2,0)}=F=d_sA.$$
(56)
$`F`$ is the field strength over the continous space-time $`V`$.
Next, we look at the component $`\mathrm{\Theta }^{(1,1)}`$ of bi-degree $`(1,1)`$:
$$\mathrm{\Theta }^{(1,1)}=\underset{ij}{}(d_sa_{ij})p_id_fp_j.$$
(57)
$`\mathrm{\Theta }^{(1,1)}`$ corresponds to the interaction between $`V`$ and $`M`$. It also obeys the field strength transformation rule:
$$\mathrm{\Theta }^{(1,1)}=u\mathrm{\Theta }^{(1,1)}u^{}.$$
(58)
Notice that there is a peculiar property to the discrete space-time: it is not $`d_s+A`$ but $`d_s`$ appears in (57)! The reason is: we take $`\alpha _i=A`$ at the begining. Otherwise, there will be a contribution from $`\alpha _i`$’s. This coincides with (pp.561-576) and (where we replace $`U(2)`$ by $`U(1)`$ and take $`\omega _a=\omega _b`$ in and take $`A=B`$ in ).
Finally, we have the component $`\mathrm{\Theta }^{(0,2)}`$ of degree $`(0,2)`$:
$$\mathrm{\Theta }^{(0,2)}=\underset{i,j,k}{}\theta _{ijk}p_id_fp_jp_jd_fp_k.$$
(59)
$`\mathrm{\Theta }^{(0,2)}`$ corresponds to the field strength over the finit space $`M`$.
Just as in Section 3, we use the formula (16) to deal with the finite-difference degrees, i.e.,
$$\pi (d_fp_i)=i[D,\pi (p_i)].$$
(60)
We then obtain the action functinal over the discrete space-time $`V\times M`$:
$$S=_V𝑑\nu .$$
(61)
The Lagrangian density is given by the following formulas:
$$=_2+_1+_0,$$
(62)
$$_2=\mathrm{\Theta }^{(2,0)}^2=|F|^2=|d_sA|^2,$$
(63)
$$_1=\mathrm{\Theta }^{(1,1)}^2=tr(d_sH)^2=\frac{n}{n1}\underset{i}{}(d_s\varphi )_i^2,$$
(64)
$`_0`$ $`=`$ $`trH^42\mu ^2trH^2+n\mu ^4`$ (65)
$`=`$ $`{\displaystyle \underset{i,j,k,l}{}}({\displaystyle \underset{\alpha }{}}e_i^\alpha e_j^\alpha e_k^\alpha e_l^\alpha )\varphi _i\varphi _j\varphi _k\varphi _l{\displaystyle \frac{2n}{n1}}\mu ^2({\displaystyle \underset{i}{}}\varphi _i^2)+n\mu ^4.`$
5. Geometric origin of phase transitions
It is well know that the Landau-Ginzburg model is one of the most important models in the theory of phase transitions. It is a kind of ‘metamodel’ - it captures the essence of many models in this field.
The Hamiltonian density of the Landau-Ginzburg model is
$$_{LG}=\frac{1}{2}\alpha ^2|\varphi |^2+\frac{1}{2}\mu ^2\varphi ^2+\frac{1}{4!}\lambda (\varphi ^2)^2,$$
(66)
where $`\varphi `$ is the order parameter, $`\alpha ,\mu ^2,\lambda `$ are phenomenological parameters. Notice that the temperature-dependent coefficient is $`\mu ^2`$. This notation is not intended to imply that $`\mu ^2>0`$.
Consider now the case when $`V\times M=V\times \{2`$point space$`\}`$, it is easy to see from equations (64) and (65) that $`_1`$ and $`_0`$ correspond to the kinetic energy and the potential term of $`_{LG}`$ below the critical temperature respectively.
Moreover, in the continuous-spin formulation of the Ising model, see for example , the effective Hamiltonian can be expressed as $`_{LG}`$ plus additional terms, and these additional terms are of no consequence if all one seeks is critical exponents. Notice that the Ising model is just the $`2`$-state Potts model. One then can imagine that the effective Hamiltonian of the continuous-spin formulation of the $`n`$-state Potts model may be related the action functional of $`U(1)`$ gauge field we derived in the previous section.
The effective Hamiltonian density of the $`n`$-state Potts model is
$`_P`$ $`=`$ $`{\displaystyle \underset{i}{}}(\varphi _i)^2+C_1({\displaystyle \underset{i}{}}\varphi _i^2)+C_2{\displaystyle \underset{i,j,k}{}}({\displaystyle \underset{\alpha }{}}e_i^\alpha e_j^\alpha e_k^\alpha )\varphi _i\varphi _j\varphi _k`$ (67)
$`+C_3{\displaystyle \underset{i,j,k,l}{}}({\displaystyle \underset{\alpha }{}}e_i^\alpha e_j^\alpha e_k^\alpha e_l^\alpha )\varphi _i\varphi _j\varphi _k\varphi _l+C_4({\displaystyle \underset{i}{}}\varphi _i^2)^2,`$
where $`C_i(i=1,2,3,4)`$ is a constant.
From equations (64) and (65), $`_1`$ really corresponds to the kinetic energy term of $`_P`$, $`_0`$ is contained in the potential term of $`_P`$. It might be asked: why there is such a correspondance? This is because that the set of states of the Potts model forms a $`n`$ point space, one then can build the gauge theory on $`\{`$continuous space$`\}`$ $`\times `$ $`\{n`$-point space$`\}`$.
In the Landau-Ginzburg model, the order parameter $`\varphi `$ can be generalized to the vector order paramter $`𝚽`$. Obviously, the action functional of $`U(1)`$ gauge theory over $`\{`$continuous space$`\}`$ $`\times `$ $`\{n`$-point space$`\}`$ is a generalized Landau-Ginzburg Hamiltonian below the critical temperature. This means that there may exist a nontrivial geometrical structure behind a class of phase transitions.
Acknowledgements
The author would like to thank Profs. M.-L. Ge, M.-Z. Guo, Z.-J. Liu, M. Qian and Z.-D. Wang for the helpful discussions. This work was supported in part by the National Science Foundation of China and China Postdoctoral Science Foundation. |
warning/0001/astro-ph0001386.html | ar5iv | text | # Atomic diffusion in metal poor stars.
## 1 Introduction
Atomic diffusion is a basic physical element transport mechanism usually neglected in standard stellar models. It is driven by pressure gradients (or gravity), temperature gradients and composition gradients. Gravity and temperature gradients tend to concentrate the heavier elements toward the center of the star, while concentration gradients oppose the above processes. Overall diffusion acts very slowly in stars, with time scales of the order of $`10^9`$ years, so that the only evolutionary phases where diffusion is efficient are Main Sequence (but see also Michaud, Vauclair & Vauclair 1983 for a discussion about the effect of diffusion in hot Horizontal Branch stars) and the White Dwarf cooling.
The occurrence of diffusion in the Sun has been recently demonstrated by helioseismic studies (see, e.g, Christensen-Dalsgaard, Proffitt & Thompson 1993, Guenther, Kim & Demarque 1996). Solar models including this process can reproduce much better than standard models the solar pulsation spectrum and the helioseismic values of helium surface abundance and depth of the convective envelope. Moreover, it seems that neither turbulence nor other hydrodynamical mixing processes substantially reduce the full efficiency of element diffusion in the Sun, otherwise the helioseismic constraints could not be satisfied (Richard et al. 1996). Since the Sun is a typical Main Sequence (MS) star whose structure closely resembles the one of metal poor MS objects, it appears likely that diffusion should also occur in MS globular cluster (GC) and field Halo stars and be as efficient as in the Sun. Very recently, Lebreton et al. (1999) have shown that diffusion is necessary for reproducing the effective temperatures of Hipparcos subdwarfs in the metallicity range $`1.0`$\[Fe/H\]$``$0.3, belonging mainly to the thick disk of the Galaxy. On the other hand the occurrence of a full efficiency of this process in Halo stars is still a matter of debate, since it appears to be unable to explain the near constancy of the $`{}_{}{}^{7}\mathrm{Li}`$ abundances in metal-poor stars with $`T_{\mathrm{eff}}`$ larger than $``$6000 K (see, e.g., Vauclair & Charbonnel 1998). As reviewed by Vandenberg, Bolte & Stetson (1996), turbulent mixing below the convective zone, rotation, mass loss, have been proposed as additional processes able to partially inhibit atomic diffusion; mass loss in particular (mass loss rates at the level of $`10^{12}`$$`10^{13}M_{}yr^1`$ would be necessary to reproduce the observations) is an interesting candidate (Swenson 1995, Vauclair & Charbonnel 1995), but there are no strong observational constraints at present.
Investigations dealing with the effect of diffusion on Population II stars – and the present work goes along the same line – have generally considered a full efficiency of this process (but see also Proffitt & Michaud 1991), and their results can be regarded as an estimate of the upper limit of the effect of atomic diffusion on the Halo stars evolution (another non-conventional transport mechanism, radiative acceleration, does not appear to appreciably affect the evolutionary properties of low mass stars, at least in the range 1.1 $`M/M_{}`$ 1.3, as investigated by Turcotte, Richer & Michaud 1998).
Due to diffusion (see, e.g., Castellani et al. 1997, Weiss & Schlattl 1999) the stellar surface metallicity and helium content progressively decrease during the MS phase – due to their sinking below the boundary of the convective envelope –, reaching a minimum around the Turn-Off (TO) stage; then, since envelope convection deepens, a large part of the metals and helium diffused toward the center are again engulfed in the convective envelope, thus restoring the surface $`Z`$ (with $`Z`$ we indicate, as usual, the mass fraction of the metals) to almost the initial value and $`Y`$ (helium mass fraction), after the first dredge up, to a value almost as high as for evolution without diffusion. Along the Red Giant Branch (RGB) phase diffusion is basically inefficient because of the much shorter evolutionary time scales. The net effect on the evolutionary tracks is to have the MS (for a given stellar mass and initial chemical composition) colder for fixed value of the luminosity, and a less luminous and colder TO (which is reached earlier), with respect to standard models. The reason for this behaviour is that the inward settling of helium raises the core molecular weight and the molecular weight gradient between surface and center of the star. This increases the stellar radius and the rate of energy generation in the center. The metal diffusion only partially counterbalances this effect by decreasing the opacity in the envelope and increasing the central CNO abundance.
With respect to a standard isochrone of given initial metallicity and reference TO luminosity, isochrones computed accounting for the diffusion of helium and metals give an age lower by $``$1 Gyr, if the same initial metallicity is used. As far as the RGB evolution is concerned, the location of the RGB in the Hertzsprung-Russell (HR) diagram is basically unchanged with respect to standard models, and the level of the Horizontal Branch is also almost negligibly affected (Castellani et al. 1997).
Several papers have been published about the influence of atomic diffusion in old stars (see, e.g., Proffitt & Vandenberg 1991, Chaboyer, Sarajedini & Demarque 1992, D’Antona, Caloi & Mazzitelli 1997, Castellani et al. 1997, Cassisi et al. 1998, Castellani & Degl’Innocenti 1999), with the main goal of studying the influence on GC ages. The approach usually followed is to compute models with diffusion for a certain set of initial metallicities, and then to compare with the observational Colour-Magnitude-Diagram (CMD) of a given GC the isochrones whose initial metal content matches the spectroscopical GC one. Since the chemical composition of a GC is determined by means of observations of its RGB stars (see, e.g., Carretta & Gratton 1997), the spectroscopical GC metallicity truly reflects the initial one.
A very different situation holds for field MS subdwarfs, a point recently raised by Morel & Baglin (1999 - hereinafter MB99). The spectroscopical subdwarfs metallicity is not the original one, because diffusion along the MS decreases the envelope metallicity (and helium content). When comparing theoretical diffusive isochrones with subdwarfs, one must compute models with suitable initial chemical abundances which produce, at the subdwarf age, its observed surface metallicity. This means that subdwarfs models must be computed using a larger initial $`Z`$ with respect to the observed one, the exact value depending on the star age. This occurrence has potentially very important implications for example for MS-fitting distances and subdwarfs GC ages, since it implies that the MS (and TO) of subdwarfs and GC sharing the same observational value of the metallicity are not coincident.
In this paper we present models for the MS phase of low mass metal-poor stars, accounting for atomic diffusion (Sect. 2); we have considered the full efficiency of this process, as in the Sun. We will then analyze for the first time in a consistent way the effect of diffusion on three important quantities for cosmological and Galactic evolution models, namely GC distances derived by means of the MS-fitting technique (Sect. 3), age of field subdwarfs (Sect. 4), and the helium enrichment ratio $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ estimated from the width of the local subdwarfs MS (Sect. 5). A summary follows in the final section. In view of the previously discussed possibility that the efficiency of diffusion could be somewhat reduced in Population II stars, our results may be viewed as upper limits.
## 2 The models
The stellar evolution computations have been performed using the same code and the same input physics as in Salaris & Weiss (1998 - hereinafter SW98). We just recall that we have used the opacities by Iglesias & Rogers (1996) and Alexander & Ferguson (1994) for an $`\alpha `$-enhanced heavy elements mixture ($`[\alpha /\mathrm{Fe}]`$=0.4; the details of the distribution are given in SW98), and that the mixing length calibration allows to reproduce the RGB effective temperatures of a selected sample of GC in the ($`M_{\mathrm{bol}},T_{\mathrm{eff}}`$) plane, as derived by Frogel, Persson & Cohen (1983), which is the observational constraint for Halo stars. In case of models with diffusion, since the RGB position is almost unchanged with respect to standard isochrones, the same mixing length value satisfies the metal-poor RGB observational constraint. Therefore, we did not change the mixing length calibration with respect to the case of standard isochrones. This permits also to clearly show the influence of diffusion on the stellar models without any contribution from the variation of other parameters. We stress also that large part of our results does not depend on the mixing length calibration.
For the present calculations we have included the diffusion of helium and heavy elements following Thoul, Bahcall & Loeb (1994); their formalism and the input physics used in our models have been already successfully tested on the Sun (see, e.g., Ciacio, Degl’Innocenti & Ricci 1997). The variations of the abundances of H, He, C, N, O and Fe are followed all along the structures; all other elements are assumed to diffuse in the same way as fully-ionized iron (see, e.g., Thoul et al. 1994 for a comparison among different diffusion formalisms). The local changes of metal abundance are taken into account also in the opacity computation, by calculating the actual global metallicity at each mesh point, and then interpolating among tables with different $`Z`$. In this way one does not take into account the small changes in the metal distribution due to differences in the diffusion velocities of C,N,O and Fe; however, for the mass range we are dealing with ($`M`$1.0 $`M_{}`$) the differences are small and our procedure does not introduce any significant error in the opacity (see, e.g., the detailed discussion in Turcotte et al. 1998).
We have computed a set of MS models (and isochrones) with diffusion and initial metallicities \[Fe/H\]=$`2.3`$, $`2.0`$, $`1.7`$, $`1.3`$, $`1.0`$, $`0.7`$, $`0.6`$ ($`Y_0`$=0.23 and $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$=3 as in SW98). In addition we have also computed a set of isochrones for 8 and 12 Gyr which displays as actual surface metallicity the set of values previously given (we will call them ‘calibrated’ diffusive isochrones, following the nomenclature by MB99). When computing the latter isochrones we had to ensure that, for the selected ages, all the different evolving masses showed the prescribed surface metallicity. This means that we had to employ an iterative procedure for finding the exact value of the initial metallicity (larger than the final one) to be used for each case (we kept fixed $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$=3 when deriving the initial helium abundance).
In Figure 1 we show a comparison in the HR diagram among the sets of isochrones computed, and the standard ones by SW98, with $`t=`$8 and 12 Gyr and \[Fe/H\]=$``$2.3 and $`1.0`$. For the standard (solid lines) and ‘calibrated’ diffusive ones (dotted lines - hereinafter C) the labelled value of \[Fe/H\] represents the actual surface metallicity, which is constant along the isochrone, while for the non calibrated isochrones with diffusion (dashed line - hereinafter D) it represents only the initial surface metallicity.
Our results closely resemble the results by MB99; specifically, for a fixed age and luminosity the standard isochrones are hotter than the D and C ones. The C isochrones are the reddest among the three sets. The $`T_{\mathrm{eff}}`$ difference at fixed luminosity among the three sets of isochrones increases with increasing metallicity. As in MB99 we find that the shift toward higher $`T_{\mathrm{eff}}`$ of the C isochrones with respect to the standard ones increases for increasing luminosities; this changes slightly the slope of the MS, making it more vertical for the C isochrones. At \[Fe/H\]=$``$2.3 the C and D isochrones are almost coincident all along the lower MS. Table 1 displays a quantitative evaluation of the $`T_{\mathrm{eff}}`$ difference, at selected luminosities along the lower MS, between standard and C isochrones with $`t`$=8 and 12 Gyr and metallicities \[Fe/H\]=$``$1.0, $``$1.7, $``$2.3.
Figure 2 shows the difference $`\mathrm{\Delta }T_{\mathrm{eff}}`$ at fixed age between standard and C isochrones for three selected values of the evolving mass along the MS, metallicities \[Fe/H\]=$``$1.0 and $``$1.7, $`t=`$8 and 12 Gyr. For the \[Fe/H\]=$`1.7`$ isochrones only data with $`t=`$8 Gyr are shown, since stars with M=0.8$`M_{}`$ are evolved off the MS at $`t=`$12 Gyr. The displayed data are comparable with analogous quantities in Fig. 6 of MB99; our results appear consistent with MB99, when taking into account that their $`\mathrm{\Delta }T_{\mathrm{eff}}`$ values correspond to an age of 10 Gyr.
In Figure 3 the differences $`\mathrm{\Delta }`$\[Fe/H\] ($`\mathrm{\Delta }Y`$) between the initial metallicity (helium content) and the TO surface values of D isochrones are displayed for two selected initial metallicities and various ages. As a general trend, $`\mathrm{\Delta }`$\[Fe/H\] and $`\mathrm{\Delta }Y`$ increase for decreasing age (in the age range we are dealing with) and for decreasing initial metallicity (at fixed age). In the age range 6-14 Gyr $`\mathrm{\Delta }`$\[Fe/H\] is generally between 0.1 and 1.0 – apart for the most metal poor isochrones which show a larger metal depletion ($`\mathrm{\Delta }`$\[Fe/H\]$``$1.8 at $`t`$=6 Gyr for an initial \[Fe/H\]=$``$2.3) – , while $`\mathrm{\Delta }Y`$ ranges between $``$0.05 and $``$0.2; the surface $`Y`$ is practically zero for the \[Fe/H\]=$``$2.3 isochrones when $`t`$ is less than about 8 Gyr.
This increase of the TO surface metallicity (and helium) depletion with decreasing age could appear at first surprising, since diffusion has less time to work when age is lower; however, one must also take into account that convective envelopes are progressively thinner for stars populating the TO at decreasing age, thus increasing the rate of depletion of helium and metals (notice that more metal poor models have thinner surface convective regions). It is the competition between these two factors which determines the final trend of the TO surface abundances with time.
In Fig. 4 we compare our C isochrones (bolometric luminosities were transformed to $`M_\mathrm{V}`$ using the bolometric corrections described in Weiss & Salaris 1999; we just recall that, after the necessary calibration of the zero point to reproduce the solar $`M_\mathrm{V}`$ value, our adopted bolometric corrections from Buser & Kurucz 1978, 1992 agree quite well with the empirical determinations by Alonso, Arribas & Martinez-Roger 1996a) with a sample of metal-poor subdwarfs with accurate Hipparcos parallaxes ($`\sigma (\pi )/\pi <0.12`$) listed by Carretta et al. (1999), and $`T_{\mathrm{eff}}`$ derived from the Infrared Flux Method (IRFM – Alonso, Arribas & Martinez-Roger 1996b). The goal is to check if isochrones with full efficiency of diffusion are compatible with the HR diagram of field subdwarfs with accurate parallaxes and empirical $`T_{\mathrm{eff}}`$ determinations (therefore eliminating the influence of the adopted colour-transformations).
Lebreton et al. (1999) recently performed this kind of comparison for subdwarfs with metallicities in the range $``$1.0$`<`$\[Fe/H\]$`<`$0.3, and found that for $``$1.0$`<`$\[Fe/H\]$`<`$0.5 the inclusion of the full efficiency of diffusion is necessary for reproducing observational data. We extend their analysis by studying the case of lower metallicities. Our adopted subdwarfs spectroscopic \[Fe/H\] values come from Carretta et al. (1999), and are homogeneous with the Carretta & Gratton (1997) metallicity scale for GC; even if the metallicities adopted by Alonso et al. (1996b) for applying the IRFM method are different (generally lower by $``$ 0.2), the sensitivity of the derived $`T_{\mathrm{eff}}`$ to the input metallicity is so low (Alonso et al. 1996b) that no appreciable inconsistency is introduced by our choice of the \[Fe/H\] scale. For two stars we found differences as high as $``$1.0 between the metallicity used by Alonso et al. (1996b) and the Carretta et al. (1999), and we did not consider them. The \[Fe/H\] values for the subdwarfs displayed in the pictures are in the ranges, respectively, between $``$1.55 and $``$1.69 (top panel), $``$1.24 and $``$1.28 (middle panel), $``$0.98 and$``$1.02 (bottom panel). The error bar on \[Fe/H\] is typically by 0.10-0.15, while the errors on $`T_{\mathrm{eff}}`$ and $`M_\mathrm{V}`$ are shown in the figure.
The displayed MS isochrones have ages equal to 8 and 12 Gyr respectively, corresponding to approximately the upper and lower limit of the ages determined by SW98 for a large sample of Galactic GC. As it is evident from the figure, the agreement between observations and theoretical models is satisfactory. Our calibrated diffusive C isochrones (which are the correct ones to be compared with field subdwarfs of a given metallicity) reproduce satisfactorily the observations. The best agreement is for subdwarfs with average metallicity \[Fe/H\]=$``$1.0 and $``$1.6; at the intermediate metallicity $`1.25`$ the models appear to be slightly too hot, but when taking into account the error bar on the temperature and metallicity of the subdwarfs the discrepancy does not appear to be significant. In the case of \[Fe/H\]=$``$1.0 standard isochrones are also shown; it is clear that in spite of the good agreement between C models and data, one cannot use these results as a definitive proof that diffusion is fully efficient in Halo subdwarfs. Standard isochrones can also reasonably reproduce the observational data (even if for different ages), at least given the present sample of objects and observational uncertainties on $`M_\mathrm{V}`$ and $`T_{\mathrm{eff}}`$.
Regarding this last point one should notice that Cayrel et al (1997) have shown how the MS location of standard isochrones appears to be too hot by $``$120-140 K in comparison with a sample of metal poor subdwarfs with accurate Hipparcos parallaxes, when using the empirical $`T_{\mathrm{eff}}`$ determinations by Alonso et al. (1996b), and \[Fe/H\] values from the compilations by Cayrel de Strobel et al. (1997). This difference with respect to our conclusion that at \[Fe/H\]=$``$1.0 standard isochrones are still able to reproduce within the errors the position of local subdwarfs, is due mainly to the different \[Fe/H\] scale we have employed. We used the Carretta et al (1999) metallicities which are on a scale homogeneous with the GC metallicities we will use in the next section. For the stars with \[Fe/H\]$``$1.0 our adopted \[Fe/H\] values are about 0.2-0.3 larger than the corresponding values given by Cayrel de Strobel et al. (1997 - we averaged the high S/N data); these larger metallicities reduce the discrepancy between standard isochrones and observations.
## 3 Atomic diffusion and MS-fitting distances
The MS-fitting technique is the ‘classical’ method to derive distances to GC (see, e.g., Sandage 1970). The basic idea is very simple. Suppose that precise parallaxes of neighbouring subdwarfs are available; for a given GC metallicity $`Z_{\mathrm{cl}}`$, it is possible to construct an empirical template MS by considering subdwarfs with metallicity $`Z_{\mathrm{sbdw}}`$ close to $`Z_{\mathrm{cl}}`$ and applying to their colours small shifts (obtained using the derivative $`\mathrm{\Delta }(colour)_{\mathrm{MS}}/\mathrm{\Delta }`$\[Fe/H\] as derived from theoretical isochrones) for reducing them to a mono-metallicity sequence with $`Z=Z_{\mathrm{cl}}`$. The fit of this empirical MS to the observed GC one (reddening-corrected) provides the cluster distance modulus.
The main underlying assumption is that the MS of subdwarfs with a certain value of \[Fe/H\] is coincident with the MS of GC with the same metallicity. If atomic diffusion is at work in Halo stars, the underlying assumption of this technique is no longer rigorously satisfied. The point is that (as previously discussed) the spectroscopical metallicity of a GC is determined from its RGB stars; this metallicity is very close to the primordial GC chemical composition, but is not (due to the effect of diffusion) the MS one. Therefore, when fitting the local subdwarfs MS to the MS of a GC with the same observed metallicity, one is introducing an error in the derived distance modulus.
The release of the Hipparcos catalogue has enlarged the number of metal poor subdwarfs with precise parallaxes which can be now used for applying this technique to the Galactic GC, and several authors (see, e.g, Reid 1997, Gratton et al. 1997, Chaboyer et al. 1998, Carretta et al. 1999) have recently derived distances (and ages) of GC using the MS-fitting and subdwarfs with Hipparcos parallaxes. In particular, Carretta et al. (1999) have carefully analyzed the total error budget associated with the MS-fitting, but the effect of atomic diffusion in subdwarfs is nowhere mentioned.
We display in Fig. 5 standard, C and D isochrones transformed to the $`M_\mathrm{V}(BV)`$ plane according to the transformations described in Weiss & Salaris (1999), for \[Fe/H\]=$`2.3`$ and $`1.0`$. In the case of D isochrones, when computing the bolometric corrections (already used in the previous section) and the $`T_{\mathrm{eff}}`$-colour conversion, we have taken into account the fact that the surface \[Fe/H\] value is not constant along the MS (it is actually decreasing). It must be noticed that the surface helium content for stars around the TO of the isochrones with diffusion is always much lower (generally $`Y`$ is in the range $``$0.1-0.2 for ages between 8 and 12 Gyr, as discussed in the previous section) than the helium content used in the model atmospheres producing the adopted colour transformations. However, this does not introduce a substantial systematic error since, according to Carney (1981), the He abundance does not appreciably affect the flux distribution at temperatures appropriate to GC dwarfs and subgiants.
The behaviour of the isochrones in the observational $`M_\mathrm{V}(BV)`$ plane closely follows the results in the theoretical HR diagram. Standard isochrones (solid line) are systematically bluer than the diffusive ones. The C isochrones are the reddest ones, and they are progressively redder than the standard or the D ones for increasing age (the effect is stronger for large metallicities). In Fig. 6 the lower MS for \[Fe/H\]=$`1.0`$ and $`t=`$8,12 Gyr is shown; when $`M_\mathrm{V}`$6 standard isochrones are unaffected by age (a fact that is well known), while D isochrones are insensitive to the age only for $`M_\mathrm{V}`$7 and C isochrones are always affected by the age, at least down to $`M_\mathrm{V}`$=7.3 (because of the metal decrease with time due to diffusion).
The MS of a GC with an observed RGB metallicity \[Fe/H\]=$`1.0`$ is given by the D isochrones in Fig. 6, while the MS of local subdwarfs with the same observed metallicity is given by the C isochrones. As it is evident the two MS are generally not coincident.
Another difference with respect to the standard case (and another potential source of systematic errors on the actual MS-fitting distances) is the value of the derivative $`\mathrm{\Delta }(BV)_{\mathrm{MS}}`$/$`\mathrm{\Delta }`$ \[Fe/H\]. This is the only information needed from theoretical isochrones to be employed in the MS-fitting technique; it is used for shifting the subdwarfs to a mono-metallicity sequence corresponding to the observed cluster \[Fe/H\]. Since the difference in colour between the standard MS and the C MS depends on the metallicity (see Fig. 5) this will have an impact on $`\mathrm{\Delta }(BV)_{\mathrm{MS}}/\mathrm{\Delta }`$\[Fe/H\] for the subdwarfs.
Are these differences large enough to affect substantially the MS-fitting GC distances? It depends on the subdwarf sample. To explain this point let’s consider, as an example, subdwarfs with \[Fe/H\]$``$1.3 and $`M_\mathrm{V}6`$, which hypothetically have to be employed for deriving the distance to a GC with \[Fe/H\]$``$0.7. The variation of $`\mathrm{\Delta }(BV)_{\mathrm{MS}}/\mathrm{\Delta }`$\[Fe/H\] due to diffusion causes a shift of the empirical subdwarfs MS at the cluster metallicity by $`\mathrm{\Delta }(BV)`$+0.02 with respect to the standard case. This, by itself, would induce a GC distance modulus larger by $``$0.1 mag, since the MS slope $`\mathrm{\Delta }M_\mathrm{V}/\mathrm{\Delta }(BV)`$ is equal to about 5.5. However, one must correct for the vertical $`M_\mathrm{V}`$ difference between subdwarfs and GC MS, which tends to reduce the derived distance modulus by $``$0.05–0.08 mag in the age range 8–12 Gyr. The final combined effect is to have distances unchanged or increased at most by 0.05 mag with respect to the standard case. However, in the hypothesis that for determining the MS-fitting distance to a GC with \[Fe/H\]$``$1.3 one can use only subdwarfs with \[Fe/H\]$``$0.7, the situation is quite different, since the use of the diffusive C isochrones would cause a decrease of the distance modulus by $``$0.10-0.13 mag.
In the following we will study the effect of diffusion on the MS-fitting distances obtained using subdwarfs with accurate Hipparcos parallaxes. We have considered, as a test (the results are summarised in Table 2), four clusters included in the analysis by Gratton et al. (1997), namely M92 (\[Fe/H\]$`2.15`$), M5 and NGC288 (\[Fe/H\]$`1.1`$), 47Tuc (\[Fe/H\]$`0.7`$). The subdwarfs $`M_\mathrm{V}`$, $`(BV)_0`$ and \[Fe/H\] values come from Table 2 of Gratton et al. (1997); the clusters reddenings and metallicities are from the quoted paper, as well as the observational clusters MS lines. For each cluster we have considered only bona fide single stars fainter than $`V`$=6 (to avoid evolutionary effects for the standard isochrones, as well as the influence of the mixing-length calibration), with $`\sigma (\pi )/\pi <0.12`$ and in the same metallicity range as in Gratton et al. (1997).
In the case of the standard models by SW98 we recover basically the same distances by Gratton et al (1997), whose results were obtained by using a value for $`\mathrm{\Delta }(BV)_{\mathrm{MS}}/\mathrm{\Delta }`$ \[Fe/H\] derived from different isochrones, and considering subdwarfs also in the range 5.5$`<M_\mathrm{V}<`$6.0. When deriving the MS-fitting distances taking into account diffusion, we have (as outlined in the previous example) corrected the subdwarfs colours by using the $`\mathrm{\Delta }(BV)_{\mathrm{MS}}/\mathrm{\Delta }`$ \[Fe/H\] values derived from the C models, and we have also accounted for the difference in brightness at fixed colour between the subdwarfs MS (C isochrones) and the clusters one (D isochrones). Since there are small evolutionary effects for the D isochrones (representing the GC) even when 6$`M_\mathrm{V}`$7, we have taken into account 4 different possibilities. In the first two cases we have assumed for the clusters age $`t_{\mathrm{cl}}`$=8 Gyr with subdwarfs ages $`t_{\mathrm{sbdw}}`$=8 and 12 Gyr, and in the second two cases we considered $`t_{\mathrm{cl}}`$=12 Gyr and again $`t_{\mathrm{sbdw}}`$=8 and 12 Gyr.
As it is clear from Table 2, there are no appreciable modifications to the distance moduli derived from standard isochrones. The differences with respect to the standard case are small and generally within the small formal error bars associated to the fit (the error bar takes into account only the error on the fit due to the uncertainties on the subdwarfs $`M_\mathrm{V}`$ and $`(BV)`$). This is a quite important point, since it confirms the robustness of the published Hipparcos MS-fitting distances which did not take into account the effect of atomic diffusion on GC and field subdwarfs evolution.
The reason for this occurrence is that – thanks to the Hipparcos results – the sample of lower MS metal poor subdwarfs with accurate parallaxes has substantially increased with respect to the recent past. In performing the MS fitting we have used objects whose metallicity is close to the actual GC metallicity; in this case, as it is evident, the colour correction to be applied to the subdwarfs is small, and even the occurrence of a sizeable change of $`\mathrm{\Delta }(BV)_{\mathrm{MS}}/\mathrm{\Delta }`$ \[Fe/H\] does not modify appreciably the final distance. Moreover, the subdwarfs are all sufficiently faint so that the difference between the GC (D isochrones) and subdwarfs (C isochrones) MS is generally kept at the lowest possible value (this difference generally increases for increasing luminosity).
In conclusion, the effect of diffusion on the two main distance determination methods for GC stars, namely MS-fitting and HB fitting, is practically negligible, since also the HB luminosities are negligibly influenced by diffusion. The final effect on the GC age estimates is therefore just a reduction by about 1 Gyr due to the change of the TO brightness.
## 4 Atomic diffusion and subdwarfs ages
The age of field subdwarfs is, together with the age of stellar clusters, an important piece of information needed for understanding time scales and formation mechanism of the Galaxy. As repeatedly stressed before, the isochrones used for studying field subdwarfs of a certain surface metallicity are different from the ones to be employed when dealing with GC with the same observational value of \[Fe/H\]. This is clearly at odds with the usual procedure, when using standard isochrones, to use the same models for field and GC MS stars.
In Fig. 7 (upper panel) we display, as an example, standard and C isochrones with \[Fe/H\]=$``$2.3 and selected ages ($`t=`$8,12,13 Gyr in the case of standard models, while for C isochrones $`t=`$8,12 Gyr). It appears clear how the influence of diffusion is extremely relevant for the derived subdwarfs ages. The TO absolute $`V`$ brightness for a standard isochrone with $`t=`$13 Gyr is coincident with the TO brightness of a C isochrone of only 8 Gyr; both the TO $`M_\mathrm{V}`$ and $`(BV)`$ differences corresponding to an age between 8 and 12 Gyr are strongly reduced (by $``$50%) when passing from standard to C isochrones. This undoubtedly causes a strong reduction in the derived subdwarfs age, and has an important effect also on the determination of the age dispersion.
The reason for such a big difference with respect to standard calculations is that the C isochrone is basically an isochrone of much larger initial metal abundance than the standard one (by how much larger depends on the age) and with diffusion. It is well known that for a fixed age the effect of diffusion at a fixed initial metallicity is to decrease the TO luminosity (and colour); in the case of C isochrones, there is in addition the effect of the larger initial abundance which further lowers the TO luminosity (and colour); the amount of the cumulative effect depends on the selected age and initial metallicity.
In the lower panel of Fig. 7 a comparison between the TO region of C and D isochrones with \[Fe/H\]=$``$2.3 and $`t=`$8,12 Gyr is shown. The use of the D isochrones (suitable for GC) for deriving subdwarfs ages does not introduce a too large error for $`t=`$12 Gyr, while the difference with respect to C isochrones is very large for $`t=`$8 Gyr. This is due to the fact that, due to the lower age, the TO-mass is higher and therefore the depth of the convective envelope is smaller in TO stars, with the consequent larger depletion (and increase of the initial metallicity for the calibrated models) of the metal and helium abundances due to diffusion.
Diffusion has further important consequences when trying to determine the age-metallicity relation for old Halo subdwarfs. We have already seen that the use of diffusive C isochrones in place of standard ones strongly reduces the derived subdwarfs ages. Once the age of a sample of subdwarfs is obtained, one can study the age-metallicity relation for deriving information about Galactic formation mechanism and time scales. Of course the relevant quantity is the relation between the age of the stars and the metallicity from which they formed, which, in case of diffusion, is different from the actual one. The difference between initial and surface \[Fe/H\] at the TO is always within 0.1-0.3 for ages of about 10 Gyr or larger. But if one derives subdwarfs ages of the order of 6-8 Gyr (using the appropriate C isochrones), the observed \[Fe/H\] of TO stars must be increased by $``$0.2-1.0 (somewhat larger corrections are found for the most metal poor models with \[Fe/H\]=$``$2.3) for obtaining the initial value. This has to be taken into account in the analysis.
It is interesting to notice that if, for example, TO very metal poor subdwarfs with \[Fe/H\] around $``$3.0 (a sample of them can be found in Schuster et al. 1996) are found to be relatively young ($`t=`$8 Gyr) when employing the appropriate $`\alpha `$-enhanced diffusive C isochrones, their initial metallicity had to be \[Fe/H\]$``$2.2, very similar to the initial metallicity of the most metal poor GC.
By considering the fact that the \[Fe/H\] depletion is larger at lower ages (at least for ages larger than $``$6 Gyr), it is interesting to notice that if one finds in a given observed metallicity range that metal poor subdwarfs are younger than metal rich ones, this could correspond – for a particular combination of ages and width of the metallicity range – to an age spread at a constant value of the initial metallicity.
## 5 Atomic diffusion and the value of $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$
The ratio $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ of helium supplied to the interstellar medium by stars relative to their supply of heavy elements is an important quantity to test theoretical stellar yields and for deriving the slope of the relation between helium and oxygen in extra-galactic Hii regions, a fundamental ingredient for determining the primordial helium abundance.
One possible approach to the determination of this quantity is the study of the MS width of local subdwarfs. Since changes of the initial values of $`Y`$ and $`Z`$ push the MS in opposite directions (increasing $`Y`$ makes the MS bluer, while increasing $`Z`$ makes it redder), the width of the local subdwarfs MS for a fixed metallicity range is a function of the $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ ratio in the interstellar medium. One can therefore consider two $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ indicators: either the vertical (usually in $`M_{\mathrm{bol}}`$) width at a fixed value of $`T_{\mathrm{eff}}`$, or the horizontal width (in $`(BV)`$ or log$`(T_{\mathrm{eff}}`$)) at a fixed value of $`M_\mathrm{V}`$ (see, e.g., Castellani, Degl’Innocenti & Marconi 1999 and references therein). Usually the lower MS (corresponding to subdwarfs with $`M_\mathrm{V}>`$5.5-6.0) is used for the analysis to avoid (as in the MS-fitting technique) evolutionary effects and the influence of the mixing length calibration.
As it has been shown before, one of the effects of atomic diffusion on MS subdwarfs is to increase the MS width for a fixed metallicity interval and assumed initial $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ value. This is due to the fact that the colour difference between the diffusive C isochrones and the standard ones is metallicity dependent, and is larger at larger metallicities.
As an example, we have considered a value for subdwarfs effective temperature log($`T_{\mathrm{eff}}`$)=3.70 (corresponding to $`M_\mathrm{V}>`$6); we then computed the MS $`\mathrm{\Delta }M_{\mathrm{bol}}`$ broadening due to diffusion, in the interval between \[Fe/H\]=$``$2.3 and $``$0.7 – a metallicity range typical of Halo subdwarfs – and for subdwarfs ages equal to 8 and 12 Gyr, by means of comparisons with the SW98 models. As expected, $`\mathrm{\Delta }M_{\mathrm{bol}}`$ results to be larger for C isochrones with respect to standard ones, the exact value depending on the subdwarfs age since the entire MS location of C isochrones does depend on age; this means that standard isochrones underestimate $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ with respect to the calibrated diffusive ones.
The amount of this systematic difference was derived by computing additional C isochrones and varying the initial $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ ratio in the range between 1 and 5. We found that C isochrones (in the explored $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ range) lead to initial $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ ratios larger by $``$1-2, the exact amount depending on the subdwarfs ages. Moreover, we found that the dependence on the initial helium abundance of the values of $`M_{\mathrm{bol}}`$ at a fixed log$`(T_{\mathrm{eff}})`$ along the lower MS, is in broad agreement with the results from standard models by Castellani et al. (1999).
## 6 Summary
We have analyzed the influence of heavy elements and helium diffusion on the MS of metal poor low mass stars in connection with the determination of GC distances via MS-fitting technique, field subdwarfs ages, and the helium enrichment ratio $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ derived from the width of the subdwarfs MS. These three quantities are all of paramount importance for cosmological and Galactic evolution issues.
The necessity of this analysis was prompted by the recognition that isochrones for MS subdwarfs and GC with direct spectroscopical determinations of the metallicity are not the same if diffusion is taken properly into account; moreover, differences in the MS location and TO position between subdwarfs standard and diffusive isochrones are metallicity and age dependent.
We have considered the full effect of atomic diffusion, without any allowance for possible hydrodynamical mixing phenomena which could reduce the efficiency of diffusion in Population II stars, as some observations appear to suggest (see the discussion in Sect. 1).
Our main results are:
i) $`\alpha `$-enhanced calibrated diffusive isochrones reproduce well within the observational errors the position in the $`M_\mathrm{V}T_{\mathrm{eff}}`$ plane of metal poor field subdwarfs with accurate parallaxes and empirical values of $`T_{\mathrm{eff}}`$ (from the IRFM method), for reasonable assumptions about their age. However, with the observational sample considered here and taking into account the existing observational uncertainties, it is impossible from this comparison to demonstrate that diffusion is fully efficient in Population II stars.
ii) MS-fitting distances obtained using current samples of Hipparcos subdwarfs and standard isochrones are negligibly affected by atomic diffusion.
iii) The estimated subdwarfs ages and age dispersion are strongly modified when diffusion is properly considered for the subdwarfs isochrones (absolute ages are significantly reduced). Since the actual metallicity of subdwarfs in the TO region can be very different from the initial one, one must take into account this effect when deriving age-metallicity relations.
iv) The value of $`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$ in the Galactic Halo metallicity range turns out to be systematically underestimated (by $`\delta `$($`\mathrm{\Delta }Y/\mathrm{\Delta }Z`$)$``$1-2) if standard isochrones are employed.
## Acknowledgments
We thank H. Ritter for a preliminary reading of the manuscript and the referee, Dr. Y. Lebreton, for valuable comments which improved the presentation of our results. |
warning/0001/cond-mat0001340.html | ar5iv | text | # Ordering Dynamics of Heisenberg Spins with Torque : Crossover, Spin waves and Defects
## I Introduction
When a many-body system like a magnet or a binary fluid is quenched from its disordered high temperature phase to its ordered configuration at low temperatures, the slow annealing of “defects” (interfaces in binary fluids, vortices (hedgehogs) in XY (Heisenberg) magnets) separating competing domains, makes the dynamics very slow. The system organizes itself into a self similar spatial distribution of domains characterized by a single diverging length scale which typically grows algebraically in time $`L(t)t^{1/z}`$. This spatial distribution of domains is reflected in the scaling behavior of the equal-time correlation function $`C(r,t)f(r/L(t))`$. The autocorrelation function, $`A(t)L(t)^\lambda `$ is a measure of the memory of the initial configurations. The exponents $`z`$ and $`\lambda `$ and the scaling function $`f(x)`$ characterize the dynamical universality classes at the zero temperature fixed point (ZFP).
The above phenomenology suggests that the asymptotic dynamics of the order parameter is dominated by the dynamics of its defects, and that bulk fluctuations (concentration waves in a binary fluid, spin waves in a magnet) relax fast and decouple from the dynamics of defects at late times. This picture is at the heart of recent approximate theories such as the Gaussian Closure Scheme .
But is this picture accurate ? In this paper we shall study the very realistic example of the conservative dynamics of a Heisenberg magnet driven by a torque induced by the local molecular field, and show that the longer-lived spin waves couple to the defects even at late times, driving the system to a new fixed point. The new ‘torque-driven’ fixed point characterized by $`z=2`$ and $`\lambda 5.05`$, is accessed after a crossover time $`t_c1/g^2`$ (where $`g`$ is the strength of the torque). Crossover scaling forms describe physical quantities at late times (like the domain size $`L(t,g)`$ and correlation functions $`C(r,t,g)`$ and $`A(t,g)`$) for all values of $`g`$. In the absence of the torque, the spin-waves decay faster, but even so we find that the Gaussian Closure Scheme is internally inconsistent. This inconsistency may however be rectified by including leading corrections to the gaussian distribution (as suggested by Mazenko for the dynamics of the conserved scalar (Ising) order parameter).
For completeness we also study the effects of including the torque in the dynamics following a quench to the critical point $`T_c`$. As reported in earlier studies , the torque is relevant with exponents $`z=4\epsilon /2`$ and $`\lambda =d`$ (where $`\epsilon =6d`$). We show to all orders in perturbation theory that $`\lambda =d`$ which follows as a consequence of the conservation of total magnetization .
## II Heisenberg Magnet and Precessional Dynamics
The order parameter $`\stackrel{}{\varphi }`$ (whose components are $`\varphi _\alpha `$ with $`\alpha =1,2,3`$) describing a coarse-grained spin density in a Heisenberg ferromagnet in three dimensions experiences a torque from the joint action of the external field (if present) and the local molecular field. In response the spins precess with a Larmour frequency $`\mathrm{\Omega }_L`$ about the total magnetic field. Coupling to various faster degrees of freedom like lattice vibrations or electrons, causes a dissipation in energy and an eventual relaxation towards equilibrium.
This dynamics follows from the generalized Langevin equation and the Poisson algebra ,
$$\frac{\varphi _\alpha }{t}=\mathrm{\Gamma }^2\frac{\delta F}{\delta \varphi _\alpha }+\mathrm{\Omega }_Lϵ_{\alpha \beta \gamma }\varphi _\beta \frac{\delta F}{\delta \varphi _\gamma }+\eta _\alpha .$$
(1)
The noise $`\stackrel{}{\eta }`$ arising from the heat bath has mean zero and is conservative,
$$\eta _\alpha (𝐱,t)\eta _\beta (𝐱^{},t^{})=2k_BT\mathrm{\Gamma }\delta _{\alpha \beta }^2\delta (𝐱𝐱^{})\delta (tt^{}).$$
(2)
The free-energy functional $`F`$ is taken to be of the Landau-Ginzburg form,
$`F[\stackrel{}{\varphi }]`$ $`=`$ $`{\displaystyle d^3x\left[\frac{\sigma }{2}(\stackrel{}{\varphi })^2\frac{r}{2}(\stackrel{}{\varphi }\stackrel{}{\varphi })+\frac{u}{4}(\stackrel{}{\varphi }\stackrel{}{\varphi })^2\right]}.`$ (3)
The second term in Eq. (1) is clearly the torque $`\stackrel{}{M}\times \stackrel{}{H}`$, where $`\stackrel{}{H}\delta F/\delta \stackrel{}{\varphi }`$ is the local molecular field.
Both the inertial term (by virtue of $`F`$ being rotationally invariant in spin space) and the dissipation conserve the total spin, and so the full equations of motion (1) also conserve the total spin.
Since the noise correlator is proportional to temperature, we may drop it in our discussion of zero temperature quenches. We then scale space $`𝐱`$, time $`t`$ and the order parameter $`\stackrel{}{\varphi }`$ as
$`𝐱\sqrt{{\displaystyle \frac{r}{\sigma }}}𝐱,t{\displaystyle \frac{\mathrm{\Gamma }r^2t}{\sigma }},\stackrel{}{\varphi }\sqrt{{\displaystyle \frac{u}{r}}}\stackrel{}{\varphi }`$
to obtain the equation of motion in dimensionless form,
$$\frac{\stackrel{}{\varphi }}{t}=^2\left(^2\stackrel{}{\varphi }\stackrel{}{\varphi }+(\stackrel{}{\varphi }\stackrel{}{\varphi })\stackrel{}{\varphi }\right)+g\left(\stackrel{}{\varphi }\times ^2\stackrel{}{\varphi }\right).$$
(4)
The dimensionless parameter $`g=(\mathrm{\Omega }_L\sigma /\mathrm{\Gamma })(ru)^{1/2}`$ is the ratio of the precession frequency to the relaxation rate. Setting $`\mathrm{\Omega }_L10^7`$ Hz and $`\mathrm{\Gamma }10^610^{10}`$ Hz, gives $`g`$ in the range of $`10^310`$.
## III Phase Ordering Dynamics at $`T=0`$
Let us now prepare the system initially in the paramagnetic phase and quench to zero temperature. We study the time evolution of the spin configurations as they evolve according to Eq. (4). We calculate the equal time correlator,
$$C(𝐫,t)\stackrel{}{\varphi }(𝐱,t)\stackrel{}{\varphi }(𝐱+𝐫,t),$$
(5)
and the autocorrelator,
$$C(\mathrm{𝟎},t_1=0,t_2=t)A(t)=\stackrel{}{\varphi }(𝐫,0)\stackrel{}{\varphi }(𝐫,t),$$
(6)
where the angular brackets are averages over the random initial conditions and space. At late times these correlators should attain their scaling forms
$$C(𝐫,t)f(r/L(t))$$
(7)
and
$$A(t)L(t)^\lambda .$$
(8)
The length scale $`L(t)`$, which is a measure of the distance between defects, may be evaluated either from the first zero of $`C(r,t)`$ or from the scaling of the energy density, $`\epsilon =\frac{1}{V}𝑑𝐫(\stackrel{}{\varphi }(𝐫,t))^2L(t)^2`$, and grows with time as $`L(t)t^{1/z}`$. We compute the scaling function $`f(x)`$, the dynamical exponent $`z`$ and the autocorrelation exponent $`\lambda `$ by simulating the Langevin Eq. (4).
### A Langevin Simulation
The Langevin simulation is performed by discretizing Eq. (4) on a simple cubic lattice (with size $`N`$ ranging from $`50^3`$ to $`60^3`$) and adopting an Euler scheme for the derivatives . The space and time intervals have been chosen to be $`\mathrm{}x=2.5`$ and $`\mathrm{}t=0.2`$. With this choice of parameters, we have checked that the resulting coupled map does not lead to any instability. We have also checked that the results remain unchanged on slight variations of $`\mathrm{}x`$ and $`\mathrm{}t`$. Throughout our simulation we have used periodic boundary conditions.
The correlation functions Eqs. (5), (6) are calculated for values of $`g`$ ranging from $`0`$ to $`1`$. Measured quantities are averaged over $`510`$ initial configurations. The initial configurations are taken from two ensembles, both in the disordered phase. In ensemble A, $`\stackrel{}{\varphi }(t=0)`$ is uniformly distributed within the volume of a unit sphere centered at the origin. $`\stackrel{}{\varphi }`$ at different spatial points are uncorrelated. In ensemble B, $`\stackrel{}{\varphi }(t=0)`$ is uniformly distributed on the surface of a unit sphere centered at the origin. $`\stackrel{}{\varphi }`$ at different spatial points are again uncorrelated. We consider these two initial conditions to check if the late time dynamics is insensitive to the choice of initial ensemble (as long as they do not introduce any long-range correlations).
We first report simulation results for ensemble $`𝐀`$.
Figure $`1`$ is a scaling plot of $`C(r,t)`$ versus $`r/L(t)`$ for various values of the parameter $`g`$, where $`L(t)`$ is extracted from the first zero of $`C(r,t)`$. Note that the scaling function for $`g=0`$ is very different from those for $`g>0`$ ; further the $`g>0`$ scaling functions do not seem to depend on the value of $`g`$. This suggests that the dynamics crosses over to a new ‘torque-driven’ ZFP. This is also revealed in the values of the dynamical exponent $`z`$. In Fig. 2, a plot of $`L(t)`$ versus $`t`$ gives the expected value of $`z=4`$ when $`g=0`$. For $`g>0`$, we see a distinct crossover from $`z=4`$ when $`t<t_c(g)`$ to $`z=2`$ when $`t>t_c(g)`$. The crossover time $`t_c(g)`$ decreases with increasing $`g`$. The same $`z`$ exponent and crossover are obtained from the scaling behaviour of the energy density $`\epsilon `$.
To make sure that our results are not affected by finite size, we compute 3 relevant time scales (shown in Table 1 below) — (1) $`t_c(g)`$, the crossover time from a $`t^{1/4}`$ to a $`t^{1/2}`$ growth, (2) $`t_s(g)`$, the time at which asymptotic scaling begins, (3) $`t_{fs}`$, the time at which finite size effects become prominent. It is clear from the Table that $`t_c<t_s<t_{fs}`$, as it should be if our data is to be free of finite size artifacts. A general rule-of-thumb is that finite size effects start becoming prominent when the domain size gets to be of order $`1/3`$ the system size, and we see from Table 1 that $`L_{max}/N`$ is comfortably less than $`1/3`$.
Table 1
| $`g`$ | $`t_c(g)`$ | $`t_s(g)`$ | $`t_{fs}`$ | $`L_{max}/N`$ | $`f_{min}`$ |
| --- | --- | --- | --- | --- | --- |
| $`0`$ | $``$ | $`900`$ | $`>7650`$ | $`1/10`$ at $`t=7650`$ | $`0.14`$ |
| $`0.1`$ | $`3150`$ | $`7650`$ | $`>7650`$ | $`1/6`$ at $`t=7650`$ | $`0.08`$ |
| $`0.3`$ | $`900`$ | $`1350`$ | $`>7650`$ | $`1/4`$ at $`t=7650`$ | $`0.06`$ |
| $`0.5`$ | $`450`$ | $`900`$ | $`4950`$ | $`1/3.7`$ at $`t=4950`$ | $`0.06`$ |
FIG. 1. Scaling plot of $`C(r,t)`$ for $`N=50^3`$. The scaling function $`f(x)`$ changes as $`g`$ is varied from $`g=0()`$ to $`g0(g=0.1(),\mathrm{\hspace{0.17em}0.3}(\mathrm{}),\mathrm{\hspace{0.17em}0.5}(\mathrm{}))`$.
FIG. 2. Log-Log plot of $`L(t)`$. At $`g=0()`$ we find that $`z=4`$ (line of slope $`0.25`$ drawn at the bottom for comparison). At $`g0(g=0.1(),\mathrm{\hspace{0.17em}0.3}(\mathrm{}),\mathrm{\hspace{0.17em}0.5}(\mathrm{}))`$, $`z`$ crosses over from $`4`$ to $`2`$ (line of slope $`0.5`$ drawn at the top).
The last column in Table 1 shows $`f_{min}`$, the value of the scaling function evaluated at the first minimum as a function of $`g`$. It is easy to see why $`f_{min}(g)<f_{min}(g=0)`$, since the precession of the spins about the local molecular field would cause spins from neighboring ”domains” to be less anti-correlated. This is borne out by computing the spin-wave correction to an approximate form of $`C(r,t;g=0)`$ (given in Eq. (34), more on this later) to quadratic order in the spin-wave amplitude .
The autocorrelation function $`A(t)`$ is calculated for $`g=0,0.2`$ and $`0.3`$ (Fig. 3). The simulations have been done on a lattice of size $`60^3`$ and averaged over $`10`$ initial configurations (we have to average over a large number of initial configurations for smoother data). The $`\lambda `$ exponent extracted from the asymptotic decay of $`A(t)`$ clearly suggests a crossover from $`\lambda =2.2`$ to $`\lambda 5.05`$. The numerical determination of $`\lambda `$ is subject to large errors and is very sensitive to finite size effects, and so we have to go to very late times and hence large system sizes to obtain accurate results.
FIG. 3. Log-Log plot of $`A(t)`$ vs $`t`$ for $`g=0(),\mathrm{\hspace{0.17em}0.2}(+),\mathrm{\hspace{0.17em}0.3}(\mathrm{})`$. Solid line on top has the form $`a/t^{\lambda /z}`$ where $`\lambda =2.19`$ and $`z=4`$ (corresponding to the $`g=0`$ fixed point) while the one below has a $`\lambda =5.05`$ and $`z=2`$ (corresponding to the ‘torque-driven’ fixed point).
To make sure that we collect asymptotic data untainted by finite size, we compute two time scales (Table 2) — (i) $`t_{fit}(g)`$, the time beyond which $`A(t)`$ can be fit with a power law $`a(t+t_0)^{\lambda /z}`$, (ii) $`t_{fs}`$, the time at which finite size effects on $`A(t)`$ become prominent. The crossover time $`t_c`$ was displayed in Table 1.
To determine $`t_{fs}`$ we plot an effective exponent $`\lambda _{eff}=td(\mathrm{log}A(t))/dt`$ as a function of $`1/t`$. The derivative is calculated numerically with a $`\delta t=15`$ (in units of the time discretisation $`\mathrm{\Delta }t`$). We see from Fig. 4, that at late times $`t>t_{fs}`$, $`\lambda _{eff}`$ crosses over to being a decreasing function of time, clearly a finite size effect. This estimate of $`t_{fs}`$ is not very sensitive to the choice of $`\delta t`$ changing by $`1\%`$ (for $`g=0.2`$) and $`3.5\%`$ (for $`g=0.3`$) as $`\delta t`$ changes by $`5`$ units. Note that finite size effects in $`A(t)`$ appear earlier than in $`C(r,t)`$.
It is seen from Table 2 that $`t_{fit}<t_{fs}`$, as it should if we are to have an accurate determination of $`\lambda `$.
Table 2
| $`g`$ | $`t_{fit}(g)`$ | $`t_{fs}`$ | $`\lambda `$ |
| --- | --- | --- | --- |
| $`0.0`$ | $`900`$ | $`>9000`$ | $`2.199\pm 7.5\times 10^3`$ |
| $`0.2`$ | $`1500`$ | $`5376`$ | $`5.100\pm 6.1\times 10^3`$ |
| $`0.3`$ | $`900`$ | $`5181`$ | $`5.010\pm 2.3\times 10^3`$ |
FIG. 4. $`\lambda _{eff}`$ versus $`1/t`$ for $`g=0.0(),\mathrm{\hspace{0.17em}0.2}(),\mathrm{\hspace{0.17em}0.3}(+)`$. Finite size effects set in when $`\lambda _{eff}`$ starts becoming a decreasing function of time. For $`g=0`$ we do not see any finite size effects in $`\lambda `$ within our simulation times.
The last column of Table 2 lists the value of $`\lambda `$ as a function of $`g`$. The data presented and the plot in Fig. 3 clearly support a crossover from $`\lambda =2.2`$ at $`g=0`$ to $`\lambda =5.05`$ at $`g0`$. The values of $`\lambda `$ satisfy the bound derived in .
We now present results of the Langevin simulation for initial conditions taken from ensemble B. We find that the value of $`z`$, the form of the scaling functions $`f(x)`$ (Fig. 5) and the decay of the autocorrelation function $`A(t)`$ (Fig. 6) are insensitive to the choice of initial conditions.
FIG. 5. Scaling function $`f(x)`$ vs $`x`$ for $`g=0.3`$ using ensembles $`𝐀()`$ and $`𝐁(\mathrm{})`$.
Since the initial condition B sets the magnitude of the spins to its $`T=0`$ equilibrium value, the crossover time $`t_c`$ is smaller than for ensemble A. For the same reason the domain sizes computed using ensemble B are larger than that of A.
FIG. 6. Log-Log plot of $`A(t)`$ for $`g=0.3`$ using ensembles $`𝐀()`$ and $`𝐁(+)`$. A power law $`a/t^{\lambda /z}`$ with $`\lambda =5.05`$ is displayed for comparison.
### B Crossover Phenomenon
It is clear from the last section, that though the asymptotic dynamics is governed by the new ‘torque-driven’ fixed point, the dynamics at earlier times $`t<t_c`$ follows the $`g=0`$ behavior. This suggests that the dynamics for arbitrary $`g`$, may be analyzed as a crossover from the $`g=0`$ fixed point characterized by ($`z=4`$, $`\lambda 2`$) to the torque-driven fixed point where ($`z=2`$, $`\lambda 5`$).
A simple scaling argument encourages us to think of such a crossover scenario. On restoring appropriate dimensions, the dynamical equation Eq.(4) can be rewritten as a continuity equation,
$$\stackrel{}{\varphi }(𝐫,t)/t=\stackrel{}{j}$$
(9)
where the “spin current” is
$$\stackrel{}{j}_\alpha =\mathrm{\Gamma }\left(\frac{\delta F[\stackrel{}{\varphi }]}{\delta \varphi _\alpha }+\frac{\mathrm{\Omega }}{\mathrm{\Gamma }}ϵ_{\alpha \beta \gamma }\varphi _\beta \varphi _\gamma \right).$$
(10)
From a dimensional analysis where we replace $`j_\alpha `$ by the ‘velocity’ $`dL/dt`$, we find
$$\frac{dL}{dt}=\mathrm{\Gamma }\frac{\sigma }{L^3}+\mathrm{\Omega }\frac{\sigma M_0}{L},$$
(11)
where $`M_0`$, $`\sigma `$ and $`\mathrm{\Gamma }^1`$ are the equilibrium magnetization, surface tension and spin mobility respectively. Beyond a crossover time given by $`t_c(g)(\mathrm{\Gamma }/M_0\mathrm{\Omega })^21/g^2`$, simple dimension counting shows that the dynamics crosses over from $`z=4`$ to $`z=2`$ in conformity with our numerical simulations.
FIG. 7. Scaling plot of $`y=L(t,g)/t^{1/4}`$ versus $`x=tg^2`$ for $`g=0.03(),\mathrm{\hspace{0.17em}0.05}(\mathrm{}),\mathrm{\hspace{0.17em}0.07}(\mathrm{}),\mathrm{\hspace{0.17em}0.09}(),\mathrm{\hspace{0.17em}0.10}()`$. The solid line of slope $`0.25`$ is the theoretical estimate of the asymptotic form of the scaling function as $`x\mathrm{}`$ (see text).
The crossover physics is best highlighted by numerically demonstrating crossover scaling of the domain size $`L(t,g)`$ and the correlation functions $`C(r,t,g)`$ and $`A(t,g)`$.
For instance, Eq. (11) suggests that the domain size obeys the scaling form $`L(t,g)=t^{1/4}s_m(tg^2)`$ where the crossover function $`s_m(x)`$ is determined from the transcendental equation,
$$x^{1/2}s_m(x)\mathrm{ln}(1+x^{1/2}s_m^2)2x=0.$$
(12)
We shall now argue (and then confirm numerically) that the above scaling form holds in general. Scaling $`rr/b`$, $`tt/b^z`$ and $`gg/b^{y_g}`$, scales the domain size by
$$L(t,g)=bs(t/b^z,g/b^{y_g})$$
(13)
where $`y_g`$ is the scaling dimension of $`g`$. We choose $`b`$ such that $`t/b^z=1`$, which implies
$$L(t,g)=t^{1/z}s(g/t^{y_g/z}).$$
(14)
Setting $`g=0`$ gives $`L(t,g=0)=t^{1/z}s(0)`$, telling us that $`z=4`$. Thus the scaling form Eq. (14) is governed by the $`g=0`$ fixed point. We therefore need to evaluate $`y_g`$ at this $`g=0`$ fixed point. We determine $`y_g`$ by noting the $`g`$ contribution to Eq. (4)
$`{\displaystyle \frac{d\stackrel{}{\varphi }}{dt}}`$ $``$ $`g\stackrel{}{\varphi }\times \delta F[\stackrel{}{\varphi }]/\delta \stackrel{}{\varphi }`$ (15)
$`=`$ $`g\stackrel{}{\varphi }\times \stackrel{}{\mu }`$ (16)
$``$ $`g/L^2`$ (17)
where the last relation is obtained by demanding local equilibrium (Gibbs-Thomson) on the chemical potential $`\stackrel{}{\mu }`$. Thus equating dimensions, $`[g]=[t^1][L^2]=[L^{z+2}]=[L^2]`$ leading to $`y_g=2`$. The crossover scaling form for the domain size can now be read out from Eq. (14),
$$L(t,g)=t^{1/4}s(g^2t).$$
(18)
The $`x\mathrm{}`$ asymptote of $`s(x)`$ can be obtained by demanding that we recover the ‘torque-driven’ fixed point behavior, which forces $`s(x\mathrm{})x^{1/4}`$.
We will now check whether this crossover scaling form is seen in our Langevin simulation. If the above proposal is true, then the data should collapse onto the scaling curve $`s(x)`$ when plotted as $`L(t,g)/t^{1/4}`$ versus $`tg^2`$. Figure 7 shows the results of the numerical simulation — the data collapse is not good away from the asymptotic regimes. To see a better data collapse away from either fixed point, it is necessary to include corrections to scaling.
Corrections to scaling come from two sources — (i) finite time effects and (ii) nonlinear corrections to the scaling fields . Finite time corrections can be incorporated by introducing finite-time shift factors $`ttt_0`$, which can be neglected in the $`t\mathrm{}`$ limit. Nonlinear corrections to scaling are incorporated by constructing a nonlinear, analytic function $`\stackrel{~}{g}(g)`$ of the physical fields $`g`$, such that it reduces to $`g`$ in the limit $`g0`$. The simplest choice of such a function is
$$\stackrel{~}{g}(g)=\frac{g+cg^2}{1+cg^2},$$
(19)
leading to a nonlinear scaling variable
$$\stackrel{~}{x}=(\stackrel{~}{g}(g))^2(tt_0).$$
(20)
The data plotted with respect to this nonlinear scaling variable shows a much better collapse (Fig. 8) when $`c`$ is chosen to be around $`1.5`$ (in the Figs. 8 - 10, the finite time shift $`t_0`$ was taken to be $`0`$). The simple mean-field estimate $`s_m(\stackrel{~}{x})`$ plotted for comparison ( Eq. (12)), is exact only at the asymptotes.
FIG. 8. Plot of $`y=L(t,g)/(tt_0)^{1/4}`$ versus $`\stackrel{~}{x}`$ when $`c1.5`$. The point $``$ on the $`y`$ axis, represents the value of $`y`$ as $`\stackrel{~}{x}0`$.
We have seen in the last section that the equal time correlation function $`C(r,t,g)`$ is unaltered when scaled with the domain size $`L`$, and so we expect it to have the following scaling behavior
$$C(r,t,g)=f(r/L,t/L^z,g/L^{y_g}),$$
(21)
where $`z`$ is the dynamical exponent at the $`g=0`$ fixed point and $`y_g`$ is the scaling dimension of $`g`$. $`L`$ is the size of the domain, given by Eq. (18). This readily leads to a two variable scaling form ,
$$C(r,t,g)=f(\frac{r}{t^{1/4}},tg^2),$$
(22)
with scaling variables $`\rho =r/t^{1/4}`$ and $`x=tg^2`$. When $`x=0`$ and $`x\mathrm{}`$ then $`f(\rho ,x)=f_0(\rho )`$ and $`f(\rho ,x)=f_T(\rho )`$ respectively, where $`f_0(\rho )`$, $`f_T(\rho )`$ are the asymptotic scaling functions at $`g=0`$ and $`g0`$. Again in terms of the nonlinear scaling variables $`\stackrel{~}{x}`$ and $`\stackrel{~}{\rho }=r/(tt_0)^{1/4}`$, we find a very good collapse of the data for $`c1.2`$ (Fig. 9).
FIG. 9. $`C(r,t,g)`$ versus $`\stackrel{~}{x}`$ at $`\stackrel{~}{\rho }=0.50,\mathrm{\hspace{0.17em}0.82},`$and $`1.50`$ for $`g=0.03(),\mathrm{\hspace{0.17em}0.05}(),\mathrm{\hspace{0.17em}0.07}(\mathrm{}),\mathrm{\hspace{0.17em}0.09}(\mathrm{}),\mathrm{\hspace{0.17em}0.1}(+),\mathrm{\hspace{0.17em}0.3}()`$ showing data collapse for $`c1.2`$.
Similar arguments suggest that the autocorrelation function satisfies the scaling form
$$A(t,g)=t^{\lambda _0/4}a(tg^2),$$
(23)
where $`a(x=0)=a_0`$ is a constant, and $`\lambda _02.2`$ is the value of the autocorrelation exponent at $`g=0`$. As $`x\mathrm{}`$, the scaling function $`a(x)`$ should asymptote to $`a(x)x^{\lambda _0/4\lambda _T/2}`$, where $`\lambda _T5.05`$ is the exponent at the ‘torque-driven’ fixed point. This expectation is borne out by the numerical simulation (Fig. 10), where we have again used the nonlinear scaling variable $`\stackrel{~}{x}`$ for better collapse.
FIG. 10. Log-Log plot of $`y=A(t,\stackrel{~}{g})/t^{\lambda _0/4}`$ versus $`\stackrel{~}{x}`$ for $`g=0.1(),\mathrm{\hspace{0.17em}0.2}(+),\mathrm{\hspace{0.17em}0.3}(\mathrm{})`$ showing data collapse for $`c1.1`$. The scaling function aysmptotes to a line of slope $`\lambda _0/4\lambda _T/2=1.95`$ as $`\stackrel{~}{x}\mathrm{}`$.
The above discussion clearly indicates that for times $`tt_c(g)1/g^2`$, the dynamics is affected by the $`g=0`$ fixed point while for $`tt_c(g)1/g^2`$, it follows the ‘torque-driven’ fixed point. Our scaling analysis suggests the following renormalization group flow diagram,
### C Failure of Mazenko Closure Scheme : Interaction of Defects with Spin Waves
We would like to know if the crossover phenomenon described in the last section can be understood from certain approximate theories of phase ordering of conserved vector order parameters. In particular could we use such theories to calculate the crossover scaling functions and the correlation functions at the ‘torque-driven’ fixed point. The Gaussian Closure Scheme introduced by Mazenko has been considered a very successful theory to compute scaling functions of conserved vector order parameters, and it is to this we turn our attention.
The method consists of trading the order parameter $`\stackrel{}{\varphi }(𝐫,t)`$ which is singular at defect sites, for an everywhere smooth field $`\stackrel{}{m}(𝐫,t)`$, defined by a nonlinear transformation,
$$\stackrel{}{\varphi }(𝐫,t)=\stackrel{}{\sigma }\left(\stackrel{}{m}(𝐫,t)\right).$$
(24)
The choice for the nonlinear function $`\stackrel{}{\sigma }`$ is dictated by the expectation that at late times, the magnitude of $`\stackrel{}{\varphi }`$ saturates to its equilibrium value almost everywhere except near the defect cores. This suggests that the appropriate choice for $`\stackrel{}{\sigma }`$ is an equilibrium defect profile,
$$\frac{1}{2}_m^2\stackrel{}{\sigma }\left(\stackrel{}{m}(𝐫,t)\right)=V^{}\left(\stackrel{}{\sigma }(\stackrel{}{m}(𝐫,t))\right),$$
(25)
where $`V^{}(\stackrel{}{x})\stackrel{}{x}+(\stackrel{}{x}\stackrel{}{x})\stackrel{}{x}`$. The auxiliary field $`\stackrel{}{m}`$ now has the natural interpretation as the position vector from the nearest defect core. Implicit in this choice is that smooth configurations such as spin waves relax fast and so decouple from defects at late times. The simplest nontrivial solution of Eq. (25) is the hedgehog configuration,
$$\stackrel{}{\sigma }\left(\stackrel{}{m}(𝐫,t)\right)=\frac{\stackrel{}{m}(𝐫,t)}{|\stackrel{}{m}(𝐫,t)|}g(|\stackrel{}{m}|),$$
(26)
where $`g(0)=0`$ and $`g(\mathrm{})=1`$.
Equation (4) can be used to derive an equation for the correlation function $`C(12)\stackrel{}{\varphi }(𝐫_1,t_1)\stackrel{}{\varphi }(𝐫_2,t_2)`$. Substituting for $`\stackrel{}{\varphi }`$ (Eqs. (24), (26)) in the right hand side of the resulting equation, we get
$`_tC(12)`$ $`=`$ $`_1^2\left[_1^2C(12)\stackrel{}{\sigma }(\stackrel{}{m}(2))V^{}(\stackrel{}{\sigma }(\stackrel{}{m}(1)))\right]`$ (28)
$`+g\stackrel{}{\sigma }(\stackrel{}{m}(2))\stackrel{}{\sigma }(\stackrel{}{m}(1))\times _1^2\stackrel{}{\sigma }(\stackrel{}{m}(1)).`$
The Gaussian Closure Scheme assumes that each component of $`\stackrel{}{m}(𝐫,t)`$ is an independent gaussian field with zero mean at all times. This implies that the joint probability distribution $`P(12)P(\stackrel{}{m}(1),\stackrel{}{m}(2))`$ is a product of separate distributions for each component and is given by ,
$`{\displaystyle \underset{\alpha }{}}𝒩\mathrm{exp}\{{\displaystyle \frac{1}{2\left(1\gamma ^2\right)}}({\displaystyle \frac{m_\alpha ^2(1)}{S_0(1)}}+{\displaystyle \frac{m_\alpha ^2(2)}{S_0(2)}}`$ (29)
$`{\displaystyle \frac{2\gamma m_\alpha (1)m_\alpha (2)}{\sqrt{S_0(1)S_0(2)}}})\},`$ (30)
where
$$𝒩=\frac{1}{2\pi \sqrt{(1\gamma ^2)S_0(1)S_0(2)}}$$
and
$$\gamma \gamma (12)=\frac{C_0(12)}{\sqrt{S_0(1)S_0(2)}}.$$
(31)
The joint distribution has been written in terms of the second moments $`S_0(1)=m_\alpha (1)^2`$ and $`C_0(12)=m_\alpha (1)m_\alpha (2)`$.
With this assumption, the right hand side of Eq. (28) simplifies to,
$`{\displaystyle \frac{C(12)}{t_1}}`$ $`=`$ $`^2\left[^2C(12)+{\displaystyle \frac{\gamma }{2S_0(1)}}{\displaystyle \frac{C(12)}{\gamma }}\right]`$ (33)
$`+g\stackrel{}{\sigma }(\stackrel{}{m}(2))\stackrel{}{\sigma }(\stackrel{}{m}(1))\times ^2\stackrel{}{\sigma }(\stackrel{}{m}(1)),`$
where the laplacian is taken with respect to $`𝐫_1`$. With the joint probability distribution given by Eq. (30), it is clear that the last term in the above equation vanishes, implying that the torque is irrelevant at late times. This result of the Gaussian Closure Scheme, is in direct contradiction to the results of the last two sections.
What has gone wrong ? There are two possible sources of error
1. The Gaussian assumption for the probability distribution of $`\stackrel{}{m}`$ is invalid.
We show below that while the Gaussian assumption leads to an internal inconsistency, it may be remedied by considering corrections to the gaussian distribution. This however does not solve the above contradiction.
2. The order parameter $`\stackrel{}{\varphi }`$ cannot be written in terms of the defect field $`\stackrel{}{m}`$ alone.
We will first question the Gaussian assumption, on the lines suggested by Yeung et. al. in the case of a conserved scalar (Ising) order parameter. We will do this for the case when $`g=0`$, the $`g0`$ analysis follows similarly.
The equal time correlation function may be derived from Eqs. (33), (30) and takes the form
$$C(r,t)=\frac{3\gamma }{2\pi }\left[B(2,\frac{1}{2})\right]^2F(\frac{1}{2},\frac{1}{2},\frac{5}{2};\gamma ^2)$$
(34)
where $`B(x,y)`$ and $`F(a,b,c;z)`$ are the Beta and hypergeometric functions respectively and $`\gamma `$ is given in Eq. (31) . We may expand the hypergeometric function as a power series in $`\gamma `$ and then take its fourier transform,
$`S(𝐤,t)`$ $`=`$ $`{\displaystyle \underset{p=0}{\overset{\mathrm{}}{}}}{\displaystyle }d𝐤_1\mathrm{}d𝐤_{2p+1}[a_p\gamma _{𝐤_1}(t)\gamma _{𝐤_2}(t)\mathrm{}\gamma _{𝐤_{2p+1}}(t)`$ (36)
$`\delta (𝐤+𝐤_1+\mathrm{}+𝐤_{2p+1})]`$
where the spectral density $`\gamma _𝐤`$ is the fourier transform of $`\gamma (r,t)`$ and the expansion coefficients,
$$a_p=\frac{9}{8\pi ^{3/2}}\frac{\left[\mathrm{\Gamma }(p+1/2)\right]^2}{\mathrm{\Gamma }(p+5/2)p!}\left[B(2,\frac{1}{2})\right]^2,$$
(37)
are strictly positive for $`p0`$. If Eq.(36) has to satisfy the conservation law $`S(k=0,t)=0`$, it is clear that $`\gamma _𝐤(t)`$ should be $`negative`$ at some values of $`𝐤`$. This is inconsistent with the definition Eq. (31) which implies $`\gamma _𝐤(t)0`$ for all $`𝐤`$. This definition is a consequence of the Gaussian approximation.
FIG. 11. The spectral density $`\gamma (k,t)`$ at $`t=3600`$ becomes negative for $`0k/k_m<0.5`$ and for $`1.5<k/k_m<3.0`$ (inset).
To determine the range of values of $`𝐤`$ for which $`\gamma _𝐤`$ is negative, we numerically evaluate the fourier transform of $`\gamma (r,t)`$ after inverting Eq. (34). This is prone to numerical errors because of statistical errors in our computed $`C(r,t)`$. For instance, a numerical integration of $`𝑑𝐫C(r,t)`$ gives a nonzero value whereas it should be identically zero because of the conservation law. This is reflected in large errors in $`\gamma (𝐤,t)`$ at small $`𝐤`$. We therefore adopt the following procedure. We fit a function $`C_f(x)`$ to the equal time correlation function $`C(r,t)`$ and use this to extract $`\gamma (𝐤,t)`$ from the Eq. (34). The fitting function has been taken to be
$$C_f(x)=\frac{\mathrm{sin}(x/L)}{(x/L)}\left[1+a\left(\frac{x}{L}\right)^2\right]\mathrm{exp}[b(x/L)^2]$$
(38)
which is similar to the analytic form given in Ref. . Note that only $`b`$ and $`L`$ are independent fitting parameters, $`a`$ is determined from the condition $`S_f(k=0)=0`$. This function with $`L=1.5106\pm 1.01\times 10^4`$ and $`b=0.0202\pm 2.14\times 10^4`$ gives a very good fit to $`C(r,t)`$ upto the fourth zero of the function. We observe (Fig. 11) that the spectral density, which should be a strictly positive function of its arguments, becomes negative for $`k/k_m<0.5`$ ($`\gamma (k,t)`$ is peaked at $`k_m`$) and in the range $`1.5<k/k_m<3.0`$.
Our demonstration suggests that a purely gaussian theory for the distribution of $`\stackrel{}{m}`$ is internally inconsistent. This may however be remedied by considering corrections to the purely gaussian distribution, as suggested by Mazenko for the scalar (Ising) order parameter.
In order to help us understand the nature of the corrections, let us first numerically evaluate the probability distribution of $`\stackrel{}{m}`$. We determine $`\stackrel{}{m}`$ by choosing $`g(|\stackrel{}{m}|)`$ in such a way as to make Eq. (26) invertible. A convenient choice is
$$\stackrel{}{\varphi }=\stackrel{}{\sigma }(\stackrel{}{m})=\frac{\stackrel{}{m}}{\sqrt{1+|\stackrel{}{m}|^2}}.$$
(39)
We now compute the asymptotic single point probability density $`P(m_1(𝐫,t))`$ on a $`50^3`$ lattice averaged over $`18`$ initial configurations for $`g=0,\mathrm{\hspace{0.17em}0.3},\mathrm{\hspace{0.17em}0.4}`$ and $`0.5`$. The probability density obeys a scaling form at late times (Figs. 12 and 13), $`P(m_1,t)=P(m_1/L(t))`$, where the length scale $`L(t)=\sqrt{m_1^2}t^{1/z}`$. Moreover Fig. 14 shows that the scaled distribution of $`\stackrel{}{m}`$ is identical for $`g=0`$ and $`g0`$ (the joint probability distributions are however very different). It is clear from Figs. 12 - 14, that the asymptotic distributions show marked deviations from a simple gaussian. To highlight these deviations, we plot the scaled $`\mathrm{log}(\mathrm{log}(P(m_1)))`$ versus $`\mathrm{log}(m_1^2)`$ (Fig. 15), a gaussian distribution would have given a straight line with slope $`1`$.
FIG. 12. Scaling plot of the un-normalized $`P(x=m_1/L(t))`$ for $`g=0`$ at different times $`t=900(),\mathrm{\hspace{0.17em}3600}(\mathrm{}),\mathrm{\hspace{0.17em}6300}(\mathrm{})`$. Solid line is a fit to Eq. (40).
FIG. 13. Scaling plot of the un-normalized $`P(x=m_1/L(t))`$ for $`g=0.3`$ at different times $`t=1350(),\mathrm{\hspace{0.17em}3600}(+),\mathrm{\hspace{0.17em}5400}(\mathrm{})`$. Solid line is a fit to Eq. (40).
FIG. 14. Scaling plot of the un-normalized $`P(x=m_1/L(t,g))`$ for $`g=0()`$ and $`g=0.3(+)`$ at $`t=4500`$ showing that the distributions are identical within error bars.
FIG. 15. Deviation of $`P(x=m_1/L)`$ from gaussian (straight line) for $`g=0`$. Data have been collected at times $`t=900(),\mathrm{\hspace{0.17em}3600}(+),\mathrm{\hspace{0.17em}6300}(\mathrm{})`$.
Figures 12 - 14 suggest that the deviations from gaussian can be computed by expanding $`P(m)`$ in a Hermite polynomial basis $`H_n`$ (a strategy advocated in Ref. for the scalar (Ising) dynamics),
$$P(x)=\underset{n=0}{\overset{\mathrm{}}{}}p_nH_n(x)e^{x^2},$$
(40)
where $`x=m_1(r,t)/\sqrt{S_0(r,t)}`$ and $`H_0(x)=1`$, $`H_1(x)=2x`$ and $`H_{n+1}(x)=2xH_n(x)2nH_{n1}(x)`$. The dark line in Figure 12 is an accurate fit to the $`g=0`$ data, with $`p_0=1`$, $`p_1=1.33\times 10^3\pm 6.0\times 10^5`$, $`p_2=0.2352\pm 3.8\times 10^5`$, $`p_3=1.55\times 10^4\pm 1.5\times 10^5`$, $`p_4=5.542\times 10^3\pm 7.0\times 10^6`$. Similarly in Fig. 13, the dark line is an accurate fit to the $`g=0.3`$ data with $`p_0=1`$, $`p_1=3.95\times 10^3\pm 5.5\times 10^5`$, $`p_2=0.2899\pm 1.3\times 10^5`$, $`p_3=5.35\times 10^4\pm 1.3\times 10^5`$, $`p_4=1.1913\times 10^2\pm 7.0\times 10^6`$. Indeed the odd coefficients are zero to within numerical accuracy, indicating that the distribution is even.
It is conceivable that such corrections would be able to salvage the inconsistency issue, since an additive term to the right hand side of Eq. (36) would not allow us to assert that $`\gamma _k`$ should be negative for some values of $`k`$.
Though the remedy suggested cures the inconsistency problem, it will still give a zero value to the torque contribution in Eq. (33), as long as the probability distribution of each component of $`\stackrel{}{m}`$ is even and independent. We have already demonstrated that the single point distribution is even, now we shall show that each cartesian component of $`\stackrel{}{m}`$ is independently distributed.
We numerically calculate $`P(m_1(1),m_2(2))`$ (which we label $`P(x,y)`$) at equal times $`t_1=t_2=t`$ and arbitrary separation, say $`|𝐫_1𝐫_2|=4\sqrt{3}`$ for $`g=0.3`$ (Fig. 16).
FIG. 16. Normalized joint probability distribution $`P(x,y)`$ where $`x=m_1(1),y=m_2(2)`$ for $`g=0.3`$ at $`t=2250`$ and $`|𝐫_1𝐫_2|=4\sqrt{3}`$ (averaged over $`18`$ initial configurations).
To show that the joint distribution is independent in each component, we plot the difference $`\mathrm{\Delta }(x,y)=P(x,y)P(x)P(y)`$ for $`g=0.3`$ (Fig. 17) and find it to be zero within the accuracy of our numerical computation.
FIG. 17. Plot of $`\mathrm{\Delta }(x,y)`$ where $`x=m_1(1),y=m_2(2)`$ at $`t=2250`$ and $`|𝐫_1𝐫_2|=4\sqrt{3}`$ for $`g=0.3`$. The maximum magnitude of $`\mathrm{\Delta }`$ is of the order of errors in $`\mathrm{\Delta }(x,y)`$.
We are thus forced to admit the second possibility, namely that the order parameter $`\stackrel{}{\varphi }`$ cannot be written in terms of $`\stackrel{}{m}`$ alone. For in transforming the spins $`\stackrel{}{\varphi }`$ exclusively to $`\stackrel{}{m}`$ we have implicitly ignored spin waves. A most direct demonstration of this is to compare $`C_{3\stackrel{}{\varphi }}=\stackrel{}{\varphi }(1)(\stackrel{}{\varphi }(2)\times _2^2\stackrel{}{\varphi }(2))`$ with the defect-only contribution $`C_{3\widehat{m}}=\widehat{m}(1)(\widehat{m}(2)\times _2^2\widehat{m}(2))`$ (where $`\stackrel{}{m}`$ is computed by inverting Eq. (39)).
We find that for $`g=0`$ both $`C_{3\stackrel{}{\varphi }}`$ and $`C_{3\widehat{m}}`$ are zero within error bars (Fig. 18). This is true even at very early times which implies that in the absence of the torque the spinwaves decay very fast compared to the relaxation timescale of the defects. On the other hand, when $`g0`$, we find that the two correlators behave very differently. Figure 19 clearly shows that even at late times, $`C_{3\stackrel{}{\varphi }}`$ is non zero while the defect-only contribution $`C_{3\widehat{m}}`$ is zero within errorbars.
This suggests the following decomposition in terms of defect fields (singular part) and spin-waves (smooth part), $`\stackrel{}{\varphi }=\stackrel{}{\sigma }(\stackrel{}{m})+\stackrel{}{u}`$, when $`g0`$. Such a decomposition gives rise to contributions to $`C_{3\stackrel{}{\varphi }}`$ reflecting the interaction between defects and spin-waves.
FIG. 18. $`y=C_{3\stackrel{}{\varphi }}(r)`$ $`(\mathrm{})`$ and $`y=C_{3\widehat{m}}(r)()`$ at $`t=3600`$ and $`r=|𝐫_1𝐫_2|`$ for $`g=0`$ are zero within the error bars (averaged over $`5`$ initial configurations).
FIG. 19. $`y=C_{3\stackrel{}{\varphi }}(r)`$ $`(\mathrm{})`$ and $`y=C_{3\widehat{m}}(r)()`$ at $`t=3600`$ and $`r=|𝐫_1𝐫_2|`$ for $`g=0.3`$ are distinctly different (averaged over $`5`$ initial configurations). $`C_{3\widehat{m}}(r)(+)`$, which has contributions from defects alone, is zero (within errorbars), whereas $`C_{3\stackrel{}{\varphi }}(r)`$, which in addition involves spin-wave excitations, is non zero.
We conclude this long meandering section by recounting its salient results. When $`g0`$, typical spin configurations at late times consist of slowly moving defects and long-lived spin-waves which interact with each other. The asymptotic spin distribution cannot be written in terms of the distribution of defects alone. When $`g=0`$, the spin waves decay faster, leading to a decoupling of the spin waves and defects at late times.
## IV Ordering Dynamics at $`T=T_c`$
We end this study with a brief discussion of the ordering dynamics Eq. (1) of Heisenberg spins quenched to the critical point. The critical dynamics of this model (called Model J in this context) was investigated some time ago by Ma and Mazenko . On the other hand, the dynamical renormalization group formalism for quench dynamics set up by Janssen $`et.al.`$, has been used to study Models A - C . In this section we use the dynamical renormalization technique to study the quench dynamics of Model J (given by Eq. (1) at the critical point). Though this section does not contain anything new of a fundamental nature, it does however compute exponents to all orders in perturbation.
We first demonstrate that the precession term is relevant for the quench dynamics to $`T_c`$. We will then calculate the $`z`$ and $`\lambda `$ exponents at this new fixed point. We will show, to all orders in perturbation, that $`\lambda `$ is exactly equal to the spatial dimension $`d`$. This latter fact, a consequence of the conservation law (and indeed true for Model B dynamics too), may also be arrived at by the general arguments presented in Ref. .
In the absence of the torque term, the nontrivial fixed point is given by the Wilson-Fisher value (WF), $`u^{}=(8/11)\pi ^2ϵ`$, where $`ϵ=4d`$. Power counting shows that the scaling dimension of $`g`$ is $`d/2+1z+\eta /2`$, where the exponents take their WF values $`z=4\eta `$ and $`\eta =(5/242)ϵ^2`$. This implies that the torque $`g`$ is relevant at the WF fixed point when $`d<6`$ .
We now have to determine this new torque driven fixed point and calculate the dynamical exponents $`z`$ and $`\lambda `$. Both these exponents can be obtained readily using general arguments, which we briefly discuss. At the new fixed point it is clear that $`g`$ does not get renormalized, which implies that $`z=(d+2+\eta )/2`$. Thus a calculation of $`z`$ within perturbation theory reduces to a calculation of $`\eta `$ at this fixed point . Likewise $`\lambda `$ can be obtained from the general arguments outlined in Ref. . A crucial ingredient in this argument (valid only for quenches to $`T_c`$) is the demonstration that $`S(k,t)`$ obeys a scaling form at $`k=0`$, a feature that was proved in Ref. to all orders in perturbation for Model B dynamics. Here we directly calculate both $`z`$ and $`\lambda `$ using diagrammatic perturbation theory, and show that $`\lambda =d`$ to all orders in perturbation.
This is done within the Martin-Siggia-Rose (MSR) formalism . For our problem, the MSR generating functional is,
$`𝒵[\stackrel{}{h},\stackrel{}{\stackrel{~}{h}}]={\displaystyle }𝒟(\stackrel{}{\stackrel{~}{\varphi }})𝒟(\stackrel{}{\varphi })\mathrm{exp}\{J[\stackrel{}{\varphi },\stackrel{}{\stackrel{~}{\varphi }}]H_0[\stackrel{}{\varphi }_0]`$ (42)
$`+{\displaystyle _0^{\mathrm{}}}dt{\displaystyle }d𝐤(\stackrel{}{\stackrel{~}{h}}_𝐤\stackrel{}{\stackrel{~}{\varphi }}_𝐤+\stackrel{}{h}_𝐤\stackrel{}{\varphi }_𝐤)\}`$
with the MSR action written as
$`J[\stackrel{}{\varphi },\stackrel{}{\stackrel{~}{\varphi }}]={\displaystyle _0^{\mathrm{}}}dt{\displaystyle }d𝐤\{\stackrel{}{\stackrel{~}{\varphi }}_𝐤[_t\stackrel{}{\varphi }_𝐤+\mathrm{\Gamma }k^2{\displaystyle \frac{\delta F[\stackrel{}{\varphi }]}{\delta \stackrel{}{\varphi }_𝐤}}`$ (45)
$`+{\displaystyle }d𝐤_1({\displaystyle \frac{g\mathrm{\Gamma }}{2}}(k_1^2(𝐤𝐤_1)^2)\stackrel{}{\varphi }_{𝐤_1}\times \stackrel{}{\varphi }_{𝐤𝐤_1})]`$
$`\mathrm{\Gamma }k^2\stackrel{}{\stackrel{~}{\varphi }}_𝐤\stackrel{}{\stackrel{~}{\varphi }}_𝐤\}.`$
In the expression for the generating functional, the initial distribution of the order parameter (gaussian with the width $`=\tau _0^1`$) enters the form of $`H_0=𝑑𝐤\frac{\tau _0}{2}(\stackrel{}{\varphi }_𝐤(0)\stackrel{}{\varphi }_𝐤(0))`$ .
Power counting reveals the presence of two different upper critical dimensions coming from the quatric term ($`d_c^u=4`$) and the cubic torque term ($`d_c^g=6`$) in the action $`J`$. This implies we have to evaluate the fixed points and exponents in a double power series expansion in $`ϵ=4d`$ and $`\epsilon =6d`$ .
The unperturbed correlation $`C_𝐤^0(t_1,t_2)`$ and response $`G_𝐤^0(t_1,t_2)`$ functions, and the bare $`u`$ and $`g`$ vertices are shown in Fig. 20. Again power counting shows that at $`d=3`$, our perturbation expansion does not generate additional terms other than those already contained in $`J`$, i.e. the theory is renormalizable. However the perturbation theory gives rise to ultraviolet divergences which can be removed by adding counter-terms to the action.
To remove these divergences, we introduce renormalization factors (superscripts $`R`$ and $`B`$ denote renormalized and bare quantities respectively), $`\stackrel{}{\stackrel{~}{\varphi }}_𝐤^R(0)=(\stackrel{~}{Z}Z_0)^{1/2}\stackrel{}{\stackrel{~}{\varphi }}_𝐤^B(0)`$, $`\stackrel{}{\varphi }_𝐤^R(t)=Z^{1/2}\stackrel{}{\varphi }_𝐤^B(t)`$, $`\stackrel{}{\stackrel{~}{\varphi }}_𝐤^R(t)=\stackrel{~}{Z}^{1/2}\stackrel{}{\stackrel{~}{\varphi }}_𝐤^B(t)`$, $`u^R=Z_u^1u^B`$, $`g^R=Z_g^1g^B`$, $`\mathrm{\Gamma }^R=Z_\mathrm{\Gamma }^1\mathrm{\Gamma }^B`$ and $`\tau _0^R=Z_{\tau _0}^1\tau _0^B`$.
Since the dynamics obeys detailed balance, the renormalization factors $`Z`$ and $`Z_u`$ are the same as in statics. Further the conservation of the order parameter forces $`Z\stackrel{~}{Z}=1`$ to all orders.
The new fixed point is given by the zeroes of the $`\beta `$ functions of the theory. The $`\beta `$ functions, calculated from the $`Z`$ factors, get contributions from all diagrams containing the primitively divergent diagrams $`\mathrm{\Gamma }_{\varphi \stackrel{~}{\varphi }}^{(2)}`$, $`\mathrm{\Gamma }_{\varphi \varphi \stackrel{~}{\varphi }}^{(3)}`$ and $`\mathrm{\Gamma }_{\varphi \varphi \varphi \stackrel{~}{\varphi }}^{(4)}`$ (Fig. 20).
The new fixed point, to one loop, is given by $`g^{}=\pm \sqrt{192\pi ^3\epsilon }+𝒪(\epsilon ^{3/2})`$, $`u^{}=(8/11)\pi ^2ϵ+𝒪(ϵ^2)`$ (note $`u^{}`$ does not change from its WF value to all loops) and the dynamical exponent $`z=4\epsilon /2+𝒪(ϵ^2)`$ .
FIG.20. Unperturbed (a) response function $`G_𝐤^0`$, (b) correlation function $`C_𝐤^0`$, and the (c) two bare vertices $`u`$ and $`g`$. Wavy and straight lines represent the $`\stackrel{}{\stackrel{~}{\varphi }}_𝐤(t)`$ and $`\stackrel{}{\varphi }_𝐤(t)`$ fields respectively. (d) Primitively divergent diagrams $`\mathrm{\Gamma }_{\varphi \stackrel{~}{\varphi }}^{(2)}`$, $`\mathrm{\Gamma }_{\varphi \varphi \stackrel{~}{\varphi }}^{(3)}`$ and $`\mathrm{\Gamma }_{\varphi \varphi \varphi \stackrel{~}{\varphi }}^{(4)}`$.
The $`\lambda `$ exponent can be computed from the response function $`G_𝐤(t,0)\stackrel{}{\stackrel{~}{\varphi }}_𝐤(0)\stackrel{}{\varphi }_𝐤(t)`$ since this is equal to the autocorrelation function $`\tau _{0}^{}{}_{}{}^{1}\stackrel{}{\varphi }_𝐤(t)\stackrel{}{\varphi }_𝐤(0)`$, as can be seen from the first term in $`J`$ on integrating by parts. The response function gets renormalized by
$$G_k^R(t,0)=Z_0^{1/2}G_k^B(t,0).$$
(46)
The divergent contributions to $`G_B`$ could come from two sources. Each term in the double perturbation series could contain the primitively divergent subdiagrams $`\mathrm{\Gamma }^{(2)}`$, $`\mathrm{\Gamma }^{(3)}`$ or $`\mathrm{\Gamma }^{(4)}`$, which we have already accounted for by replacing these by their renormalized counterparts. The other divergent contribution could arise from the primitive divergences of the 1-particle reducible vertex function $`\mathrm{\Gamma }^{(2)}(𝐤,t,0)`$, defined by $`G_𝐤(t,0)G_𝐤(tt^{})\mathrm{\Gamma }^{(2)}(𝐤,t^{},0)𝑑t^{}`$. The superficial divergence of the diagrams contributing to $`G_𝐤(t,0)`$ is $`D=V_u(d4)+\frac{V_g}{2}(d6)2`$ (where $`V_u(V_g)`$ is the number of $`u`$ ($`g`$) vertices respectively). This is negative for all $`d`$. For (a) when $`d>6`$, the only stable fixed point is the gaussian fixed point and so $`D=2`$, (b) when $`4<d6`$, $`u`$ is irrelevant and so $`D=\frac{V_g}{2}(d6)2<0`$ and (c) when $`d4`$, $`D`$ is clearly negative. This implies that $`G_𝐤^B(t,0)`$ does not get renormalized and $`Z_0=1`$. Consequently $`\lambda `$ stays at its mean-field value of $`d`$ for this conserved Heisenberg dynamics both with and without the torque.
## V Conclusions
Traditional analysis of the asymptotic ordering dynamics of vector order parameters focuses on the dynamics of defects, and ignores the bulk excitations like spin waves which most often decay faster. In this work we have looked at the very realistic model of Heisenberg spins with precessional dynamics and have shown that the longer lived spin waves couple to the defects even at late times, driving the system to a new fixed point. The new ‘torque-driven’ fixed point characterized by $`z=2`$ and $`\lambda 5.05`$, is accessed after a crossover time $`t_c1/g^2`$ (where $`g`$ is the strength of the torque). Crossover scaling forms describe physical quantities like domain size and equal/unequal time correlation functions for all values of $`g`$. In the absence of the torque, the spin-waves decay faster and so do not contribute to the asymptotic dynamics.
We also studied the effects of the torque on the dynamics following a quench to the critical point $`T_c`$. The torque is relevant with exponents $`z=4\epsilon /2`$ and $`\lambda =d`$ (where $`\epsilon =6d`$). We found to all orders in perturbation theory that $`\lambda =d`$ which follows as a consequence of the conservation of total magnetization.
We hope we have provided strong evidence that in order to go beyond the present approximate theories of the asymptotic dynamics of conserved order parameters, we need to systematically evaluate contributions coming from the interaction of defects with spin-waves.
We thank D. Dhar for interesting discussions and Y. Hatwalne for a critical reading of the manuscript. |
warning/0001/cond-mat0001301.html | ar5iv | text | # Decoherence in Bose-Einstein Condensates: Towards Bigger and Better Schrödinger Cats
## I Motivation and Summary of Results
Microscopic quantum superpositions are an everyday physicist’s experience. Macroscopic quantum superpositions, despite nearly a century of experimentation with quantum mechanics, are still encountered only very rarely. Fast decoherence of macroscopically distinct states is to be blamed . In spite of that, recent years were a witness to an interesting quantum optics experiment on decoherence of a few photon superpositions. Moreover, matter-wave interference in fullerene $`\mathrm{C}_{60}`$ has been observed . Another recent experiment has succeeded in “engineering” the environment in the context of trapped ions . This success plus rapid progress in Bose-Einstein condensation (BEC) of alkali metal atomic vapors tempt one to push similar investigations even further into the macroscopic domain. The condensates can contain up to $`10^7`$ atoms in the same quantum state. What is more, it is possible to prepare condensates in two different internal states of the atoms. Some of these pairs of internal levels are immiscible, and their condensates tend to phase separate into distinct domains with definite internal states . The immiscibility seems to be a prerequisite to prepare a quantum superposition in which all atoms are in one or the other internal state, $`|\psi =(|N,0+|0,N)/\sqrt{2}`$, where $`N`$ is the total number of condensed atoms. There are at least two theoretical proposals how to prepare a macroscopic quantum superposition in this framework . Neither of them addresses the crucial question of decoherence.
We find the quantum superposition state involving significant number of BEC atoms to be practically impossible in the standard harmonic trap. Our master equation (derived in Section IV), when applied to the standard harmonic trap of frequency $`\omega `$, gives for the decoherence rate due to the environment of noncondensed atoms
$$t_{\mathrm{dec}}^1\mathrm{\hspace{0.33em}16}\pi ^3\left(4\pi a^2\frac{N_\mathrm{E}}{V}v_T\right)N^2,$$
(1)
where $`N`$ is a number of condensed atoms, $`v_T=\sqrt{2k_\mathrm{B}T/m}`$ is a thermal velocity in the noncondensed thermal cloud, $`a`$ is a scattering length, $`V`$ is a volume of the trap, and $`N_\mathrm{E}`$ is a number of atoms in the thermal cloud,
$$N_\mathrm{E}e^{\frac{\mu }{k_\mathrm{B}T}}\left(\frac{k_\mathrm{B}T}{\mathrm{}\omega }\right)^3.$$
(2)
Here $`\mu <0`$ is a chemical potential.
Equation (1) gives the rate of decoherence to the leading order in the fugacity $`z=\mathrm{exp}(\beta \mu )`$ and also to the leading order in the condensate size, $`N`$. Next-to-leading order terms are $`O(z^2)`$ and $`O(N)`$, in agreement with Refs., so they were neglected here. We also assume that $`N`$ is small enough for a condensate to live in a single mode as opposed to the large $`N`$ Thomas-Fermi limit. Given all these assumptions, in the standard harmonic trap setting Eq.(1) is a lower estimate for the decoherence rate.
Even without going into details of our derivation, which are given in Section IV, it is easy to understand where a formula like Eq.(1) comes from. $`N^2`$ is the main factor which makes the decoherence rate large. It comes from the master equation of the Bloch-Lindblad form $`\dot{\rho }[N_\mathrm{A}N_\mathrm{B},[N_\mathrm{A}N_\mathrm{B},\rho ]]`$, with $`\mathrm{A}`$ and $`\mathrm{B}`$ the two internal states of the atoms. $`N^2`$ is the distance squared between macroscopically different components of the superposition $`(|N,0+|0,N)/\sqrt{2}`$ \- the common wisdom reason why macroscopic objects are classical . The factor in brackets is a scattering rate of a condensate atom on noncondensate atoms - the very process by which the thermal cloud environment learns the quantum state of the system.
Let us estimate the decoherence rate for a set of typical parameters: $`T=1\mu `$K, $`\omega =50`$Hz, and $`a=3\mathrm{}5`$nm. The thermal velocity is $`v_\mathrm{T}10^2`$m/s. The volume of the trap can be approximated by $`V=4\pi a_{\mathrm{return}}^3/3`$, where $`a_{\mathrm{return}}=\sqrt{2k_\mathrm{B}T/m\omega ^2}`$ is a return point in a harmonic trap at the energy of $`k_\mathrm{B}T`$. We estimate the decoherence time as $`t_{\mathrm{dec}}10^5\mathrm{sec}/(N_\mathrm{E}N^2)`$. For $`N_\mathrm{E}=10^0\mathrm{}10^4`$ and $`N=10\mathrm{}10^7`$ it can range from $`1000`$s down to $`10^{13}`$s. For $`N=10`$ our (over-)estimate for $`t_{\mathrm{dec}}`$ is large. However, already for $`N=1000`$ and $`N_\mathrm{E}=10`$ (which are still below the Thomas-Fermi regime) $`t_{\mathrm{dec}}`$ shrinks down to milliseconds. Given that our $`t_{\mathrm{dec}}`$ is an upper estimate and that big condensates are more interesting as Schrödinger cats, it is clear that for the sake of cat’s longevity, one must go beyond the standard harmonic trap setting.
From Eqs.(1,2) it is obvious that the decoherence rate depends a lot on temperature and on chemical potential. The two factors influence strongly both $`N_\mathrm{E}`$ and $`v_\mathrm{T}`$. Both can be improved by the following scenario, which is a combination of present day experimental techniques. In the experiment of Ref. a narrow optical dip was superposed at the bottom of a wide magnetic trap. The parameters of the dip were tuned so that it had just one bound state. The gap between this single condensate mode and the first excited state was $`1.5\mu `$K, which at $`T=1\mu `$K gives a fugacity of $`z=\mathrm{exp}(1.5)`$. We need the gap so that we can use the single mode approximation. At low temperatures, the gap results in a small fugacity, which is convenient for calculations. We propose to prepare a condensate inside a similar combination of a wide magnetic and a narrow optical trap (or more generally: a wide well plus a narrow dip with a single bound mode) and then to open the magnetic trap and let the noncondensed atoms disperse. The aim is to get rid of the thermal cloud as much as possible. A similar technique was used in the experiment of Ref..
Let us estimate the ultimate limit for the efficiency of this technique. At the typical initial temperature of $`1\mu `$K the thermal velocity of atoms is $`10^2`$m/s. An atom with this velocity can cross a $`1\mu `$m dip in $`10^4`$s. If we wait for, say, $`1`$s after opening the wide trap, then all atoms with velocities above $`10^6`$m/s will disperse away from the dip. A thermal velocity of $`10^6`$m/s corresponds to the temperature of $`10^8\mu `$K. As the factor $`N_Ev_TT^{7/2}`$ in Eq.(1), then already $`1`$s after opening the wide trap the decoherence rate due to non-condensed atoms is reduced by a factor of $`10^{28}`$!
It is not realistic to expect such a “cosmological” reduction factor. The “dip” which is left after the wide harmonic trap is gone could be, for example, a superposition of an ideal dip plus a wide shallow well (which was a negligible perturbation in presence of the wide harmonic trap). The well would have a band of width $`\mathrm{\Delta }E`$ of bound states which would not disperse but preserve their occupation numbers from before the opening of the wide trap. They would stay in contact with the condensate and continue to “monitor” its quantum state. Even if such a truncated environment happens to be already relatively harmless, there are means to do better than that.
Further reduction of the decoherence rate can be achieved by “symmetrization” of the environmental state. Perfect symmetrization can be obtained provided that:
i) Atoms have two internal states $`|A`$ and $`|B`$.
ii) $`|A`$ and $`|B`$ feel the same trap potential.
iii) $`AA`$ and $`BB`$ scattering lengths are the same.
iv) A Hamiltonian has a term which drives coherent transitions $`AB`$ with a frequency of $`\lambda /\mathrm{}`$.
v) $`\mathrm{\Delta }E\lambda `$.
The term in (iv) can be realized by driven coherent transitions like in the experiments of Refs..
Given an ideal symmetry between $`A`$ and $`B`$ (assumptions i-iii), the eigenmodes of the dilute environment have annihilation operators $`S_s(a_s+b_s)`$ and $`O_s(a_sb_s)`$, which are symmetric and antisymmetric respectively. $`a_s`$ and $`b_s`$ are annihilation operators for the two internal states of an atom in the trap eigenstate numbered as “$`s`$” with trap eigenenergy of $`ϵ_s`$. $`S_s`$ has energy $`(ϵ_s\lambda )`$ and $`O_s`$ has energy $`(ϵ_s+\lambda )`$. The symmetric and antisymmetric $`\mathrm{\Delta }E`$-bands of states can be visualized as two ladders shifted with respect to each other by $`2\lambda `$. In other words, the two sets of states feel the same, but shifted, trapping potentials. If (v) is satisfied, then the antisymmetric $`O_s`$’s are nearly empty since they can evaporate into symmetric states and then leave the trap. The symmetric $`S_s`$’s cannot distinguish between $`A`$ and $`B`$ so they do not destroy the quantum coherence between the Schrödinger cat’s components. After symmetrization, $`N_\mathrm{E}`$ in Eq.(1) has to be replaced by the final number of atoms in the antisymmetric states only;
$$N_\mathrm{E}^\mathrm{O}\{\begin{array}{cc}n_\lambda \mathrm{exp}((\mu \lambda )/k_\mathrm{B}T),\hfill & \text{for }2\lambda <\mathrm{\Delta }E\hfill \\ 0,\hfill & \text{for }2\lambda >\mathrm{\Delta }E\hfill \end{array}$$
(3)
Here $`T`$ is the temperature before opening the wide trap. $`n_\lambda `$ is the number of antisymmetric bound states which remain within the $`\mathrm{\Delta }E`$-band of symmetric states. Atoms in these antisymmetric bound states cannot disperse away. For $`2\lambda >\mathrm{\Delta }E`$ this number $`n_\lambda `$ is zero and there is no decoherence from the thermal cloud.
In this perfect symmetrization limit the states $`|\pm (|N,0\pm |0,N)/\sqrt{2}`$ exist within a decoherence-free pointer subspace of the Hilbert space, since they have degenerate eigenvalues of the interaction hamiltonian $`V`$ . Any state of that subspace can be written as $`\alpha |N,0+\beta |0,N`$, with $`\alpha `$ and $`\beta `$ complex numbers. If $`𝒫_{[\alpha |N,0+\beta |0,N]}`$ denotes a projector onto that subspace, then
$$[V,𝒫_{[\alpha |N,0+\beta |0,N]}]=0,$$
(4)
which means that any quantum superposition $`\alpha |N,0+\beta |0,N`$ in the subspace is an eigenstate of the system operators in the interaction hamiltonian (a perfect pointer state), and as such will retain its phase coherence and last forever. The interaction Hamiltonian between the condensate and the thermal cloud is a sum of products of condensate operators and environmental operators. Only terms with symmetric environmental operators are relevant because the antisymmetric states are empty. The total Hamiltonian is symmetric with respect to $`AB`$ so, to preserve this symmetry, the relevant terms with symmetric environmental operators also contain symmetric condensate operators. The argument simplifies a lot for small fugacity where there is only one leading term with the $`N_A+N_B`$ condensate operator. The states $`|\pm `$ are its eigenstates with the same eigenvalue $`N`$. They are also (almost) degenerate eigenstates of the condensate Hamiltonian build out of $`N_{A,B}`$. The coherent transitions $`AB`$ break this degeneracy of $`|\pm `$ but the difference of their eigenfrequencies is negligible as compared to the usual condensate lifetime of $`10`$s. In the next-to-leading order in fugacity there are symmetric interaction terms which change the number of condensed atoms. These terms drive the $`|N,0`$ and $`|0,N`$ states into slightly “squeezed-like” states $`|S,0`$ and $`|0,S`$ respectively . There are also terms which exchange $`A`$ with $`B`$. They give each state a small admixture of the opposite component. Superpositions of these are still decoherence-free pointer subspaces - there are no relevant antisymmetric operators to destroy their quantum coherence. When the antisymmetric environmental states begin to be occupied (see Eq.(3)), then the commutation relation Eq.(4) is only approximate and states within the subspace will decohere. To leading order in fugacity and condensate size, the decoherence rate is given by Eqs. (1) and (3).
This pointer subspace is not perfect – its existence is in apparent contradiction with the finite lifetimes of the condensates which can last for at most $`10\mathrm{}20`$s. The reason is that the thermal cloud is not the only source of decoherence. The condensate loses atoms because of Rayleigh scattering, external heating, and three-body decay. The atoms which escape from the condensate carry information about its quantum state. They destroy its quantum coherence. Three body decay, the last process of the three above is the most important one . In the experiment of Ref. the measured loss rate per atom was $`4/`$s for $`N=10^7`$ or around $`1`$ atom per $`10^7`$s. The last rate scales like $`N^3`$ so already for $`N=10^4`$ just one atom is lost per second; decoherence time is $`1`$s. Another possibility is to increase slightly the dip radius. The loss rate scales like density squared so an increase in the dip width by a factor of $`2`$ reduces the loss rate by a factor of $`2^6=64`$.
Ambient magnetic fields are yet another source of coherence loss . The condensed atoms have magnetic moments. If the magnetic moments of $`A`$ and $`B`$ were different the magnetic field would distinguish between them and would introduce an unknown phase into the quantum superposition, thus rendering its underlying coherent nature undetectable. Fortunately the much used $`|F,m_F=|2,1,|1,1`$ states of <sup>87</sup>Rb have the same magnetic moments. For them the magnetic field is a “symmetric” environment .
One more source of decoherence is the typical $`1\%`$ difference between the $`AA`$ and $`BB`$ scattering lengths which violates the assumption (iii) above. The Hamiltonian is not perfectly symmetric under $`AB`$. Even for a perfectly symmetrized environment symmetric environmental operators couple to not fully symmetric condensate operators. This means that for the $`1\%`$ difference of scattering lengths symmetrization can improve decoherence time by at most two orders of magnitude as compared to the unsymmetrized environment.
In summary, we outlined a BEC scenario for an experimental realisation of a decoherence-free Schrödinger cat. This scenario has two ingredients:
1) opening of the wide trap followed by an evaporation of the thermal cloud,
2) and symmetrization of the environment.
The Schrödinger cat is expected to be a quantum superposition of number eigenstates, $`|N,0`$ and $`|0,N`$.
More details can be found in Sections II-IV, where we study the decoherence rate and the idea of symmetrized environment. Finally, Section V contains discussions.
## II Symmetrized Environment
We introduce annihilation operators for $`|A`$ and $`|B`$,
$`\varphi _A(\stackrel{}{x})=ag(\stackrel{}{x})+{\displaystyle \underset{s}{}}a_su_s(\stackrel{}{x}),`$ (5)
$`\varphi _B(\stackrel{}{x})=bg(\stackrel{}{x})+{\displaystyle \underset{s}{}}b_su_s(\stackrel{}{x}).`$ (6)
where $`g(\stackrel{}{x})`$ is the ground state wave-function localized in the dip with energy $`ϵ_g<0`$, and $`s`$ is an index running over excited (environmental) states with ortonormal wave-functions $`u_s(\stackrel{}{x})`$. The Hamiltonian of the system is
$$H=d^3x[v(\varphi _A^{}\varphi _A)(\varphi _B^{}\varphi _B)+\{\frac{u}{2}\varphi _A^{}\varphi _A^{}\varphi _A\varphi _A+\varphi _A^{}\varphi _A+U(r)\varphi _A^{}\varphi _A\lambda \varphi _A^{}\varphi _B\}+\{AB\}].$$
(7)
Here $`u=4\pi \mathrm{}^2a_{AA}/m`$ ($`a_{AA}=a_{BB}`$ are the inter-scattering lengths) and $`v=4\pi \mathrm{}^2a_{AB}/m`$ ($`a_{AB}`$ is the intra-scattering length). The immiscibility assumption implies $`v>u`$. $`U(r)`$ is the trap potential and $`\lambda `$ is the strength of the coherent driving transitions $`AB`$. Substitution of (5) into (7) and subsequent linearization gives a Hamiltonian for the dilute environment
$$H_\mathrm{E}=\underset{s}{}\left[ϵ_s(a_s^{}a_s+b_s^{}b_s)\lambda (a_s^{}b_s+b_s^{}a_s)\right].$$
(8)
$`ϵ_s`$ is the single particle energy of level $`s`$. We take the lowest enviromental energy $`\mathrm{min}[ϵ_s]=0`$. A transformation
$$S_s=\frac{a_s+b_s}{\sqrt{2}},O_s=\frac{a_sb_s}{\sqrt{2}}$$
(9)
brings $`H_E`$ to a diagonal form
$$H_\mathrm{E}=\underset{s}{}\left[(ϵ_s\lambda )S_s^{}S_s+(ϵ_s+\lambda )O_s^{}O_s\right].$$
(10)
Symmetric $`S`$’s and antisymmetric $`O`$’s form two identical ladders of states but shifted with respect to each other by $`2\lambda `$. There is a gap $`(\lambda ϵ_g)>0`$ between the lowest $`S_s`$ state and the ground state. In equilibrium the state $`S_s`$’s occupation number is $`n_s^\mathrm{S}=1/[\mathrm{exp}(\beta (ϵ_s\lambda ϵ_g))1]z\mathrm{exp}(\beta (ϵ_s\lambda ))`$, where the fugacity $`z=\mathrm{exp}(\beta |ϵ_g|)`$ is assumed to be small. The $`O_s`$’s occupation number is $`n_s^\mathrm{O}n_s^\mathrm{S}\mathrm{exp}(2\beta \lambda )`$. If $`2\lambda \beta 1`$, then $`O`$’s would be nearly empty. The condensate two-mode Hamiltonian is
$$H_\mathrm{C}=ϵ_g(a^{}a+b^{}b)\lambda (a^{}b+b^{}a)+\frac{u_\mathrm{c}}{2}(a^{}a^{}aa+b^{}b^{}bb)+v_\mathrm{c}(a^{}b^{}ab),$$
(11)
where e.g. $`u_\mathrm{c}=ud^3xg^4(\stackrel{}{x})`$ and $`v_\mathrm{c}>u_\mathrm{c}`$. This Hamiltonian was studied in detail in . For a purity factor $`ϵ(\lambda /(v_\mathrm{c}u_\mathrm{c})N)^N1`$ the lowest energy subspace contains two macroscopic superpositions
$`|+={\displaystyle \frac{1}{\sqrt{2N!}}}[(a^{})^N+(b^{})^N]|0,0{\displaystyle \frac{1}{\sqrt{2}}}(|N,0+|0,N),`$ (12)
$`|={\displaystyle \frac{1}{\sqrt{2N!}}}[(a^{})^N(b^{})^N]|0,0{\displaystyle \frac{1}{\sqrt{2}}}(|N,0|0,N).`$ (13)
The lower $`|+`$ and the higher $`|`$ states are separated by a small energy gap of $`N(u_\mathrm{c}v_\mathrm{c})ϵ\mathrm{ln}ϵ`$. If we just have $`ϵ<1`$, the $`|\pm `$ states contain an admixture of intermediate states $`|N1,1,\mathrm{},|1,N1`$ such that their overlap is $`+|=ϵ`$. For $`ϵ1`$ they shrink to
$$(a^{}\pm b^{})^N|0,0.$$
(14)
¿From now on we assume the pure case, $`ϵ1`$.
Finally, the interaction Hamiltonian $`V`$, which contains all condensate-noncondensate vertices, is a sum of a symmetric
$`V_\mathrm{S}=`$ $`\left[(4u+2v)(a^{}a+b^{}b)+(2v)(a^{}b+b^{}a)\right]\left[{\displaystyle \underset{s_1,s_2}{}}S_{s_1}^{}S_{s_2}\alpha _{s_1^{}s_2}\right]+`$ (18)
$`\left[(u+v)(a^{}+b^{})\right]\left[{\displaystyle \underset{s_1s_2s_3}{}}S_{s_1}^{}S_{s_2}S_{s_3}\beta _{s_1^{}s_2s_3}\right]+\mathrm{h}.\mathrm{c}.+`$
$`\left[(4u+2v)(a^{}a+b^{}b)(2v)(a^{}b+b^{}a)\right]\left[{\displaystyle \underset{s_1,s_2}{}}O_{s_1}^{}O_{s_2}\alpha _{s_1^{}s_2}\right]+`$
$`\left[a^{}+b^{}\right]\left[(uv){\displaystyle \underset{s_1s_2s_3}{}}S_{s_1}^{}O_{s_2}O_{s_3}\beta _{s_1^{}s_2s_3}+(4u){\displaystyle \underset{s_1s_2s_3}{}}O_{s_1}^{}O_{s_2}S_{s_3}\beta _{s_1^{}s_2s_3}\right]+\mathrm{h}.\mathrm{c}.`$
and an antisymmetric part
$`V_\mathrm{O}`$ $`=`$ $`[(4u2v)(a^{}ab^{}b)+(2v)(a^{}bb^{}a)][{\displaystyle \underset{s_1,s_2}{}}S_{s_1}^{}O_{s_2}\alpha _{s_1^{}s_2}+\mathrm{h}.\mathrm{c}.]+`$ (20)
$`\left[a^{}b^{}\right]\left[{\displaystyle \underset{s_1s_2s_3}{}}\beta _{s_1^{}s_2s_3}((u+v)O_{s_1}^{}O_{s_2}O_{s_3}+(uv)O_{s_1}^{}S_{s_2}S_{s_3}+(2u)S_{s_1}^{}S_{s_2}O_{s_3})\right]+\mathrm{h}.\mathrm{c}.`$
The coefficients are given by integrals
$$\alpha _{s_1^{}s_2}=\frac{1}{4}d^3xg^2u_{s_1}^{}u_{s_2},\beta _{s_1^{}s_2^{}s_3}=\frac{1}{2\sqrt{2}}d^3xgu_{s_1}^{}u_{s_2}^{}u_{s_3}.$$
(21)
In $`V_\mathrm{S}`$ and $`V_\mathrm{O}`$ we neglected vertices $`\mathrm{C}+\mathrm{C}\mathrm{C}+\mathrm{NC}`$ and $`\mathrm{C}+\mathrm{C}\mathrm{NC}+\mathrm{NC}`$, where $`\mathrm{C}`$ is a condensate and $`\mathrm{NC}`$ is a noncondensate particle, and their hermitian conjugates. They are forbiden by energy conservation due to the gap between the condensate mode and the lowest environmental state.
All the terms in $`V_{\mathrm{S},\mathrm{O}}`$ were arranged in the form \[condensate operator\]$``$\[environment operator\]. $`V_\mathrm{S}`$ contains only \[c.o.\]’s which are symmetric under $`ab`$. They act in precisely the same way on both components of the superposition and as such they do not destroy coherence between the macroscopic components. To illustrate this let us calculate a commutator of the leading order term in $`V_S`$ with the projector onto the subspace of states $`\alpha |N,0+\beta |0,N`$,
$$[(4u+2v)(a^{}a+b^{}b)\underset{s_1,s_2}{}S_{s_1}^{}S_{s_2}\alpha _{s_1^{}s_2},𝒫_{[\alpha |N,0+\beta |0,N]}]=0$$
(22)
This commutator vanishes because states of the form $`\alpha |N,0+\beta |0,N`$ are eigenstates of $`N_\mathrm{C}=a^{}a+b^{}b`$. Therefore, this subspace would be a pointer subspace if this leading term were the only term in the interaction Hamiltonian. However, $`V_\mathrm{S}`$ has other terms that are not simply functions of the total condensate number operator. Coherent states of an annihilation operator $`a+b`$ are exact eigenstates of the \[c.o.\] $`a+b`$ and approximate eigenstates of $`a^{}+b^{}`$. These coherent states, however, are combinations of (14) and as such they are not in the lowest energy subspace of $`H_\mathrm{C}`$. What is more, the matrix elements of $`a+b`$ are of the order $`O(\sqrt{N})`$ which is negligible as compared to the matrix elements of the number operator. If we project $`a+b`$ and its h.c. on the subspace (12), then their approximate eigenstates for large $`|z|`$ are coherent states $`\alpha |z,0+\beta |0,z`$. The decoherence effects of $`a+b`$ and $`N_\mathrm{C}`$ put together lead to a superposition of macroscopic “squeezed-like” states $`\alpha |S,0+\beta |0,S`$. This result is by now well established for a single component condensate . Finally, the \[c.o.\] $`a^{}b+b^{}a`$ drives the state out of the pointer subspace (12). Its effect is supressed by the purity condition $`ϵ1`$ and is also negligible as compared to the direct effect of the $`\lambda `$-term in $`H_C`$. Similar comment applies to the “out of the subspace” action of $`a^{}+b^{}`$. Therefore, the effect of all these terms in $`V_\mathrm{S}`$ imply that the subspace spanned by $`|N,0`$ and $`|0,N`$ is no longer an exact decoherence-free pointer subspace, i.e. $`[V_\mathrm{S},𝒫_{[\alpha |N,0+\beta |0,N]}]0`$. The correct pointer subspace would be one spanned by those “squeezed-like” states. However, to leading order in fugacity and in the condensate size, $`N`$, the dominant terms are those depending on $`N_\mathrm{C}`$; the subspace of $`\alpha |N,0+\beta |0,N`$ is (approximately) the exact decoherence-free pointer subspace.
The antisymmetric \[c.o.\]’s in $`V_\mathrm{O}`$ act in opposite way on both components; they destroy their quantum coherence. To illustrate this let us calculate a commutator of the leading term in $`V_\mathrm{O}`$ with the projector operator onto the subspace $`\alpha |N,0+\beta |0,N`$,
$`[(4u2v)(a^{}ab^{}b)({\displaystyle \underset{s_1,s_2}{}}S_{s_1}^{}O_{s_2}\alpha _{s_1^{}s_2}+\mathrm{h}.\mathrm{c}.),𝒫_{[\alpha |N,0+\beta |0,N]}]=`$ (23)
$`2N(4u2v)(\alpha \beta ^{}|N,00,N|\alpha ^{}\beta |0,NN,0|)({\displaystyle \underset{s_1,s_2}{}}S_{s_1}^{}O_{s_2}\alpha _{s_1^{}s_2}+\mathrm{h}.\mathrm{c}.)`$ (24)
The commutator is $`O(N)`$. The occupation numbers of $`O`$’s are supressed by a Boltzmann factor $`n^\mathrm{S}\mathrm{exp}(2\lambda \beta )`$; for $`2\lambda \beta 1`$ the $`O`$’s are unoccupied. As such they cannot scatter into condensate particles. It is also impossible to scatter $`S`$’s into $`O`$’s thanks to energy conservation. The unoccupied $`O`$’s are irrelevant for decoherence and as such can be neglected in $`V`$ and in the above commutator. This effectively sets the dangerous $`V_\mathrm{O}`$ to zero and leaves us only with $`V_\mathrm{S}0`$. In the absence of $`O`$’s the states $`\alpha |N,0+\beta |0,N`$ are a pointer subspace (or a decoherence-free subspace). The symmetrized environment of just $`S`$’s defines the quantum states of the components but it does not destroy their mutual coherence.
## III Master Equation
Symmetrized environment is a robust idea whose validity does not depend on detailed calculations. Nevertheless, for the sake of illustration we derived (by a perturbative expansion in $`V`$) an approximate Bloch-Lindblad form master equation for the reduced density matrix $`\rho (t)`$ of the condensate modes. The calculations are long but rather straightforward, their details can be found in Section IV. Here we just give the final result.
$$\dot{\rho }=\frac{i}{\mathrm{}}[\rho ,H_C^{\mathrm{ren}}]+\dot{\rho }_\mathrm{S}+\dot{\rho }_\mathrm{O}.$$
(25)
$`H_C^{\mathrm{ren}}`$ is a renormalized condensate Hamiltonian,
$$H_C^{\mathrm{ren}}=H_C+H_C+c_1\mathrm{\Delta }_1^{}\mathrm{\Delta }_1+c_2\mathrm{\Delta }_2^{}\mathrm{\Delta }_2+c_3\mathrm{\Delta }_3^{}\mathrm{\Delta }_3+c_4\mathrm{\Delta }_4^{}\mathrm{\Delta }_4.$$
(26)
$`\mathrm{}`$ means an average over an initial environment thermal density matrix and $`\dot{\rho }_\mathrm{S}`$ and $`\dot{\rho }_\mathrm{O}`$ are contributions from $`V_\mathrm{S}`$ and $`V_\mathrm{O}`$ respectively. They read
$`\dot{\rho }_\mathrm{S}`$ $`=`$ $`d_1D[\mathrm{\Delta }_1]\rho +d_2D[\mathrm{\Delta }_2]\rho ,`$ (27)
$`\dot{\rho }_\mathrm{O}`$ $`=`$ $`d_3D[\mathrm{\Delta }_3]\rho +\stackrel{~}{d}_3D[\mathrm{\Delta }_3^{}]\rho +d_4D[\mathrm{\Delta }_4]\rho `$ (28)
with $`D[\mathrm{\Delta }]\rho \{\mathrm{\Delta }^{}\mathrm{\Delta },\rho \}2\mathrm{\Delta }\rho \mathrm{\Delta }^{}`$ the Lindblad operator. The operators introduced in this expression are defined as
$`\mathrm{\Delta }_1`$ $`=`$ $`(4u+2v)(a^{}a+b^{}b)+(2v)(a^{}b+b^{}a),`$ (29)
$`\mathrm{\Delta }_2`$ $`=`$ $`(u+v)(a^{}+b^{}),`$ (30)
$`\mathrm{\Delta }_3`$ $`=`$ $`(4u2v)(a^{}ab^{}b)+(2v)(a^{}bb^{}a),`$ (31)
$`\mathrm{\Delta }_4`$ $`=`$ $`a^{}b^{}.`$ (32)
The coefficients $`c_i`$ and $`d_i`$ are proportional to fugacity $`z`$. What is more $`c_{3,4}/c_{1,2}e^{2\lambda \beta }`$ and $`d_{3,4}/d_{1,2}e^{2\lambda \beta }`$. As anticipated, the antisymmetric contributions are supressed by a Boltzmann factor due to $`O`$’s. This approximate master equation confirms our heuristic arguments.
Any asymmetry between $`A`$ and $`B`$ is a source of decoherence. If, for example, the $`AA`$ and $`BB`$ scattering lengths were different, $`u_{AA}=u+\delta u`$ and $`u_{BB}\delta u`$, then $`V_\mathrm{O}`$ would acquire extra terms $`\delta u`$. They would show up in $`\dot{\rho }_\mathrm{O}`$ with coefficients $`dzu\delta u`$. The ratio of these to the “symmetric” coefficients $`d_{1,2}`$ is $`\delta u/u`$ (which is approximately $`0.03`$ for $`|F=1;m_F=1`$ and $`|F=2;m_F=1`$ states of <sup>87</sup>Rb ) as compared to $`d_{3,4}/d_{1,2}\mathrm{exp}(2\lambda \beta )`$. If $`\delta u/u`$ is less than the Boltzmann factor, then this asymmetry is not a leading source of decoherence.
## IV Derivation of the Master Equation
We derive a master equation by perturbative expansion in $`V`$ or in the coupling constants $`u,v`$ which we regard to be of the same order. We assume the initial density matrix at $`t=0`$ to be a product $`\stackrel{~}{\rho }(0)=\rho (0)\rho _E`$ of the system (condensate) and environment (noncondensate) density matrices. To take a more accurate starting point we make a rearrangement $`VVV`$ and $`H_\mathrm{C}H_\mathrm{C}+V`$ where $`\mathrm{}=\mathrm{Tr}_\mathrm{E}[\mathrm{}\rho _\mathrm{E}]`$ is a trace over the environment thermal density matrix at the initial time. The new $`H_\mathrm{C}`$ differs from the old one by a renormalization
$`ϵ_gϵ_g^{\mathrm{eff}}=ϵ_g+(4u+2v){\displaystyle \underset{s}{}}n_s^\mathrm{S}\alpha _{s^{}s},`$ (33)
$`\lambda \lambda ^{\mathrm{eff}}=\lambda (2v){\displaystyle \underset{s}{}}n_s^\mathrm{S}\alpha _{s^{}s}.`$ (34)
We compute $`\dot{\stackrel{~}{\rho }}(t)`$ up to second order in the perturbation Hamiltonian $`V=_i\mathrm{\Delta }_iE_i`$, with $`\mathrm{\Delta }_i`$ an operator for the condensate and $`E_i`$ one for the environment. The master equation for $`\rho (t)`$ alone is obtained by tracing $`\dot{\stackrel{~}{\rho }}(t)`$ over the non-condensed modes. The result is
$$\dot{\rho }(t)=\frac{i}{\mathrm{}}[\rho ,H_\mathrm{C}]\frac{1}{2\mathrm{}^2}\underset{ij}{}_0^t𝑑\tau g_{ij}^{\mathrm{sym}}(\tau )[\mathrm{\Delta }_i(0),[\mathrm{\Delta }_j(\tau ),\rho ]]\frac{1}{2\mathrm{}^2}\underset{ij}{}_0^t𝑑\tau g_{ij}^{\mathrm{asym}}(\tau )[\mathrm{\Delta }_i(0),\{\mathrm{\Delta }_j(\tau ),\rho \}],$$
(35)
where
$`g_{ij}^{\mathrm{sym}}(\tau )`$ $`=`$ $`\{E_i(\tau ),E_j(0)\},`$ (36)
$`g_{ij}^{\mathrm{asym}}(\tau )`$ $`=`$ $`[E_i(\tau ),E_j(0)].`$ (37)
We have used that after the rearrangement $`VVV`$ and $`H_\mathrm{C}H_\mathrm{C}+V`$, there are no linear terms in the perturbative expansion because $`V=0`$. We first start to study the contribution of the symmetric terms $`V_\mathrm{S}`$ in the interaction Hamiltonian, and later we shall deal with the antisymmetric ones $`V_\mathrm{O}`$. It is easy to show that the are no cross terms S-O in the calculation of the master equation.
### A Free Hamiltonian
The next step in the calculation is to solve the dynamics of the free condensate Hamiltonian
$$H_\mathrm{C}=(u_\mathrm{c}v_\mathrm{c})J_z^2+\lambda ^{\mathrm{eff}}J_x,$$
(38)
where we introduced angular momentum operators
$`J_x`$ $`=`$ $`{\displaystyle \frac{1}{2}}(a^{}b+b^{}a),`$ (39)
$`J_y`$ $`=`$ $`{\displaystyle \frac{i}{2}}(b^{}aa^{}b),`$ (40)
$`J_z`$ $`=`$ $`{\displaystyle \frac{1}{2}}(a^{}ab^{}b),`$ (41)
and discarded constant terms proportional to the total number of particles $`N=a^{}a+b^{}b`$ . The Heisenberg equations of motion for the condensate operators $`a(t)`$ and $`b(t)`$ are
$`i\mathrm{}\dot{a}=ϵ_g^{\mathrm{eff}}a\lambda ^{\mathrm{eff}}b+(u_\mathrm{c}a^{}a+v_\mathrm{c}b^{}b)a,`$ (42)
$`i\mathrm{}\dot{b}=ϵ_g^{\mathrm{eff}}b\lambda ^{\mathrm{eff}}a+(u_\mathrm{c}b^{}b+v_\mathrm{c}a^{}a)b.`$ (43)
These equations cannot be solved exactly for $`\lambda ^{\mathrm{eff}}0`$, when numbers of $`a`$’s and $`b`$’s are not conserved independently due to the coherent transfer of particles from one state to the other. In their mean-field version, these equations correspond to the well-known macroscopic self-trapping equation, which have the feature that below a critical “purity” the oscillations between states $`A`$ and $`B`$ are not complete; for $`ϵ1`$ the systems is self-locked in either of these states . In this limit we can set $`\lambda ^{\mathrm{eff}}`$ to zero, and for the immiscible case $`v_\mathrm{c}>u_\mathrm{c}`$ the ground state of $`H_C`$ corresponds to maximum eigenstates of $`J_z^2`$, i.e. pure cat states (12). Also in this case the mean-field solutions are
$$a(t)=a(0)e^{\frac{it}{\mathrm{}}[ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2]},b(t)=b(0)e^{\frac{it}{\mathrm{}}[ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2]}.$$
(44)
We believe that these solutions are qualitatively correct as long as the purity factor $`ϵ1`$.
### B Symmetric Interaction Hamiltonian
The symmetric Hamiltonian can be written as
$$V_\mathrm{S}=\mathrm{\Delta }_1E_1+\stackrel{~}{\mathrm{\Delta }}_1\stackrel{~}{E}_1+[\mathrm{\Delta }_2E_2+\stackrel{~}{\mathrm{\Delta }}_2\stackrel{~}{E}_2+\mathrm{h}.\mathrm{c}.],$$
(45)
where we have defined condensate operators
$`\mathrm{\Delta }_1`$ $`=`$ $`(4u+2v)(a^{}a+b^{}b)+(2v)(a^{}b+b^{}a),`$ (46)
$`\stackrel{~}{\mathrm{\Delta }}_1`$ $`=`$ $`(4u+2v)(a^{}a+b^{}b)(2v)(a^{}b+b^{}a),`$ (47)
$`\mathrm{\Delta }_2`$ $`=`$ $`(u+v)(a^{}+b^{}),`$ (48)
$`\stackrel{~}{\mathrm{\Delta }}_2`$ $`=`$ $`(u+v)(a^{}+b^{}),`$ (49)
and the corresponding non-condensate operators
$`E_1`$ $`=`$ $`{\displaystyle \underset{s_1,s_2}{}}\alpha _{s_1^{},s_2}S_{s_1}^{}S_{s_1}{\displaystyle \underset{s}{}}|\alpha _{s^{},s}|^2S_s^{}S_s,`$ (50)
$`\stackrel{~}{E}_1`$ $`=`$ $`{\displaystyle \underset{s_1,s_2}{}}\alpha _{s_1^{},s_2}O_{s_1}^{}O_{s_1}{\displaystyle \underset{s}{}}|\alpha _{s^{},s}|^2O_s^{}O_s,`$ (51)
$`E_2`$ $`=`$ $`{\displaystyle \underset{s_1,s_2,s_3}{}}\beta _{s_1^{},s_2,s_3}(S_{s_1}^{}S_{s_2}S_{s_3}+O_{s_1}^{}S_{s_2}O_{s_3}),`$ (52)
$`\stackrel{~}{E}_2`$ $`=`$ $`{\displaystyle \underset{s_1,s_2,s_3}{}}\beta _{s_1^{},s_2,s_3}(O_{s_1}^{}O_{s_2}S_{s_3}+S_{s_1}^{}O_{s_2}O_{s_3}).`$ (53)
It is clear that the terms $`\mathrm{\Delta }_i`$ and $`\stackrel{~}{\mathrm{\Delta }}_i`$ and their corresponding environment operators have the same structure, and therefore will give the same qualitatively contribution to the master equation. In the following we shall keep only the $`\mathrm{\Delta }_i`$ terms.
The expectation values of multiple-points noncondensate operators are written in terms of the two-point functions
$`S_{s_1}^{}(t_1)S_{s_2}(t_2)=\delta _{s_1s_2}n_{s_1}^\mathrm{S}e^{\frac{i}{\mathrm{}}(ϵ_{s_1}\lambda )(t_1t_2)}e^{\gamma _{s_1}^\mathrm{S}|t_1t_2|},`$ (54)
$`O_{s_1}^{}(t_1)O_{s_2}(t_2)=\delta _{s_1s_2}n_{s_1}^\mathrm{O}e^{\frac{i}{\mathrm{}}(ϵ_{s_1}+\lambda )(t_1t_2)}e^{\gamma _{s_1}^\mathrm{O}|t_1t_2|},`$ (55)
$`S_{s_1}^{}(t_1)\mathrm{O}_{s_2}(t_2)=0`$ (56)
via Wick’s theorem. Here $`n_s^{\mathrm{S},\mathrm{O}}=[ze^{\beta (ϵ_s\pm \lambda )}1]^1`$ are Bose occupation numbers, and the $`\gamma _s`$’s are inverse finite lifetimes of the environmental states. We also expand these expectation values to leading order in fugacity $`z`$. The kernels are
$`E_1(\tau )E_1(0)`$ $`=`$ $`E_1(0)E_1(\tau )^{}={\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{S}e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2})}e^{\tau (\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S})}+O(z^2),`$ (58)
$`E_2(\tau )E_2^{}(0)`$ $`=`$ $`E_2(0)E_2^{}(\tau )^{}={\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda )}[2n_{s_1}^\mathrm{S}e^{\tau (\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S})}+`$ (60)
$`n_{s_1}^\mathrm{O}e^{\tau (\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{O})}]+O(z^2),`$
$`E_2^{}(\tau )E_2(0)`$ $`=`$ $`E_2^{}(0)E_2(\tau )=O(z^2).`$ (61)
Given the kernels and the free evolution of condensate operators, we have to perform the time integrals in (35). We calculate the master equation for late times $`t\gamma _s^1`$. Hence the exponential decay of the propagators dominates their behavior, and we can apply the Markovian approximation replacing the upper limit in the time integration by infinity. This calculation results in two different contributions, one being a renormalization of the free Hamiltonian
$$\delta H_\mathrm{C}=c_1\mathrm{\Delta }_1^{}\mathrm{\Delta }_1+c_2\mathrm{\Delta }_2^{}\mathrm{\Delta }_2,$$
(62)
where
$`c_1`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{S}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2})}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}))^2}},`$ (63)
$`c_2`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2[2n_{s_1}^\mathrm{S}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2)}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}+`$ (65)
$`n_{s_1}^\mathrm{O}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda ϵ_\mathrm{g}^{\mathrm{eff}}N(u_\mathrm{c}+v_\mathrm{c})/2)}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda ϵ_\mathrm{g}^{\mathrm{eff}}N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}].`$
The other contribution has the Lindblad form $`D[\mathrm{\Delta }]\rho \{\mathrm{\Delta }^{}\mathrm{\Delta },\rho \}2\mathrm{\Delta }\rho \mathrm{\Delta }^{}`$ and reads
$$\dot{\rho }_\mathrm{S}=d_1D[\mathrm{\Delta }_1]\rho +d_2D[\mathrm{\Delta }_2]\rho ,$$
(66)
where
$`d_1`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{S}{\displaystyle \frac{\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}))^2}},`$ (67)
$`d_2`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2[2n_{s_1}^\mathrm{S}{\displaystyle \frac{\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S}}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}+`$ (69)
$`n_{s_1}^\mathrm{O}{\displaystyle \frac{\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S}}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}].`$
### C Antisymmetric Interaction Hamiltonian
The antisymmetric Hamiltonian is
$$V_\mathrm{O}=\mathrm{\Delta }_3E_3+\mathrm{\Delta }_4E_4+\mathrm{h}.\mathrm{c}.,$$
(70)
where now
$`\mathrm{\Delta }_3`$ $`=`$ $`(4u2v)(a^{}ab^{}b)+(2v)(a^{}bb^{}a),`$ (71)
$`\mathrm{\Delta }_4`$ $`=`$ $`a^{}b^{},`$ (72)
and the corresponding environment operators
$`E_3`$ $`=`$ $`{\displaystyle \underset{s_1,s_2}{}}\alpha _{s_1^{},s_2}O_{s_1}^{}S_{s_2},`$ (73)
$`E_4`$ $`=`$ $`{\displaystyle \underset{s_1,s_2,s_3}{}}\beta _{s_1^{},s_2,s_3}\left[(u+v)O_{s_1}^{}O_{s_2}O_{s_3}(uv)O_{s_1}^{}S_{s_2}S_{s_3}+2uS_{s_1}^{}S_{s_2}O_{s_3}\right].`$ (74)
The kernels are
$`E_3(\tau )E_3^{}(0)`$ $`=`$ $`E_3(0)E_3^{}(\tau )^{}={\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{O}e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2}+2\lambda )}e^{\tau (\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S})}+O(z^2),`$ (75)
$`E_3^{}(\tau )E_3(0)`$ $`=`$ $`E_3^{}(0)E_3(\tau )^{}={\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{O}e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2}2\lambda )}e^{\tau (\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{O})}+O(z^2),`$ (76)
$`E_3^{}(\tau )E_3(0)`$ $`=`$ $`E_3^{}(0)E_3(\tau )=O(z^2),`$ (77)
$`E_4(\tau )E_4^{}(0)`$ $`=`$ $`E_4(0)E_4^{}(\tau )^{}={\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2[2(u+v)^2n_{s_1}^\mathrm{O}e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}\lambda )\tau (\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{O}+\gamma _{s_3}^\mathrm{O})}+`$ (79)
$`2(uv)^2n_{s_1}^\mathrm{O}e^{\frac{i\tau }{\mathrm{}}(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+3\lambda )\tau (\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{O}+\gamma _{s_3}^\mathrm{O})}],`$
$`E_4^{}(0)E_4(\tau )`$ $`=`$ $`E_4^{}(\tau )E_4(0)=O(z^2).`$ (80)
The renormalization terms are
$$\delta H_\mathrm{C}=c_3\mathrm{\Delta }_3^{}\mathrm{\Delta }_3+c_4\mathrm{\Delta }_4^{}\mathrm{\Delta }_4$$
(81)
with
$`c_3`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2\left[n_{s_1}^\mathrm{O}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}2\lambda )}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}2\lambda ))^2}}n_{s_1}^\mathrm{O}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}+2\lambda )}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}+2\lambda ))^2}}\right],`$ (82)
$`c_4`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2[2(u+v)^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2)}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{O}+\gamma _{s_3}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}+`$ (84)
$`2(uv)^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+\lambda ϵ_\mathrm{g}^{\mathrm{eff}}N(u_\mathrm{c}+v_\mathrm{c})/2)}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+3\lambda ϵ_\mathrm{g}^{\mathrm{eff}}N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}].`$
Finally, the Lindblad part is
$$\dot{\rho }_\mathrm{O}=d_3D[\mathrm{\Delta }_3]\rho +\stackrel{~}{d}_3D[\mathrm{\Delta }_3^{}]\rho +d_4D[\mathrm{\Delta }_4]\rho ,$$
(85)
where
$`d_3`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{O}}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}2\lambda ))^2}},`$ (86)
$`\stackrel{~}{d}_3`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2}{}}|\alpha _{s_1^{},s_2}|^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}}{(\gamma _{s_1}^\mathrm{S}+\gamma _{s_2}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}+2\lambda ))^2}},`$ (87)
$`d_4`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}{\displaystyle \underset{s_1,s_2,s_3}{}}|\beta _{s_1^{},s_2,s_3}|^2[2(u+v)^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{O}+\gamma _{s_3}^\mathrm{O}}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{O}+\gamma _{s_3}^\mathrm{O})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}+`$ (89)
$`2(uv)^2n_{s_1}^\mathrm{O}{\displaystyle \frac{\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S}}{(\gamma _{s_1}^\mathrm{O}+\gamma _{s_2}^\mathrm{S}+\gamma _{s_3}^\mathrm{S})^2+(\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}ϵ_{s_3}+3\lambda +ϵ_\mathrm{g}^{\mathrm{eff}}+N(u_\mathrm{c}+v_\mathrm{c})/2))^2}}].`$
### D Estimate of the Decoherence Rate
To estimate the decoherence rate for our BEC quantum superposition, let us consider one of the terms in (85), namely $`\dot{\rho }=16v^2d_3[J_z,[J_z,\rho ]]`$, which arises form the $`\mathrm{\Delta }_3`$ term. This term corresponds to elastic two-body collisions between condensate and non-condensate atoms, and induces phase decoherence. To this end we first make an estimation for the coefficient $`\alpha _{s_1^{},s_2}`$ that enters in $`d_3`$. We assume that we have a big harmonic trap plus the dip located in its center. We take as the wave function for the ground state a gaussian $`g(\stackrel{}{x})=(m\omega _{\mathrm{dip}}/(\pi \mathrm{}))^{3/2}\mathrm{exp}(m\omega _{\mathrm{dip}}r^2/\mathrm{})`$, where $`\omega _{\mathrm{dip}}`$ is the frequency of the optical trapping potential in the dip. In principle, we should use thermal wave packets for the wave functions of the non-condensed particles . However, due to the strong localization of the condensate in the dip, we can approximate the non-condensed particles as plane waves. Then
$$\alpha _{s_1^{},s_2}=\frac{1}{4}d^3xg^2(\stackrel{}{x})\frac{e^{i\stackrel{}{k}_1\stackrel{}{x}}}{\sqrt{V}}\frac{e^{i\stackrel{}{k}_2\stackrel{}{x}}}{\sqrt{V}}=\frac{1}{4V}\mathrm{exp}\left(\frac{\mathrm{}}{4m\omega _{\mathrm{dip}}}|\stackrel{}{k}_1\stackrel{}{k}_2|^2\right)$$
(90)
On the other hand, the decay rate $`\gamma `$ is much smaller than $`w`$. This lets us replace the last factor in $`d_3`$ by $`\pi \delta (\mathrm{}^1(ϵ_{s_1}ϵ_{s_2}))`$, where we have used the identity $`lim_{\gamma 0}\frac{\gamma }{\gamma ^2+x^2}=\pi \delta (x)`$. Finally, replacing the sum over momenta by the continuum expression $`_\stackrel{}{k}Vd^3k=V𝑑\mathrm{\Omega }𝑑kk^2`$, and using the free dispersion relation $`ϵ_\stackrel{}{k}=\mathrm{}^2k^2/2m`$, we find
$$d_3=\frac{\pi m}{16\mathrm{}^3}𝑑k_1k_1^3ze^{\beta \lambda }e^{\beta \mathrm{}^2k_1^2/2m}𝑑\mathrm{\Omega }_1𝑑\mathrm{\Omega }_2e^{\frac{\mathrm{}k_1^2}{m\omega _{\mathrm{dip}}}(1\mathrm{cos}\alpha )}$$
(91)
where $`\alpha `$ is the angle between the two vectors $`\stackrel{}{k}_1`$ and $`\stackrel{}{k}_2`$ corresponding to the two scattered non-condensed particles. Since the condensate in the dip is very localized, its typical linear dimension being $`l_{\mathrm{dip}}=\sqrt{\mathrm{}/m\omega _{\mathrm{dip}}}`$, then $`l_{\mathrm{dip}}l_{\mathrm{thermal}}`$, where $`l_{\mathrm{thermal}}=\sqrt{\mathrm{}^2/2mk_\mathrm{B}T}`$ is the thermal de Broglie wavelength. This means that in the above expression the first (thermal) exponential factor dominates, and we can approximate $`d_3`$ by
$$d_3\frac{\pi m(4\pi )^2}{16\mathrm{}^3}𝑑k_1k_1^3ze^{\beta \lambda }e^{\beta \mathrm{}^2k_1^2/2m}=\frac{\pi ^2m^2}{4\mathrm{}^4}\rho _\mathrm{E}v_T$$
(92)
where $`v_T=\sqrt{2k_\mathrm{B}T/m}`$ is the thermal velocity of non-condensed particles, and $`\rho _\mathrm{E}=N_\mathrm{E}/V=V^1_\stackrel{}{k}ze^{\beta \lambda }e^{\beta \mathrm{}^2k_1^2/2m}`$ is the density of non-condensed particles in the antisymmetric states.
Given the expression for $`d_3`$, we can give an estimate for the decoherence rate. In the $`|J,M`$ basis, with $`J=N/2`$ and $`J_z|J,M=(1/2)(N_AN_B)|J,M`$, the master equation reads
$$\dot{\rho }(M,M^{})16v^2d_3(MM^{})^2\rho (M,M^{})=16\pi ^3\gamma (MM^{})^2\rho (M,M^{})$$
(93)
For the quantum superposition state $`|N,0+|0,N`$, $`MM^{}=N`$, so our final estimate for the decoherence rate is
$$t_{\mathrm{dec}}^116\pi ^3\left(4\pi a^2\frac{N_\mathrm{E}}{V}v_T\right)N^2$$
(94)
The only $`O(z)`$ contribution to the amplitude decoherence can come from a term proportional to $`d_4`$. However, under closer inspection $`d_4`$ turns out to be $`O(z^2)`$. $`d_4`$ comes from a depletion/growth inelastic process such that a noncondensate particle in the initial state of $`s_1`$ collides with a condensate particle and as a result they both end in noncondensate modes $`s_2`$ and $`s_3`$. For this to happen $`s_1`$ must have a sufficiently high energy to overcome a gap between the condensate mode and the environmental modes. Thus in this case $`n_{s_1}^O=O(z^2)`$ and not $`O(z)`$.
## V Discussion
The aim of this paper was to discuss the “longevity” of Schrödinger cats in BEC’s. We have shown that while in the standard traps decoherence rates are significant enough to prevent long-lived macroscopic superpositions of internal states of the condensate, the strategy of trap engineering and symmetrization of the environment will be able to deal with that issue.
What remains to be considered is how one can generate such macroscopic quantum superposition, and how one can detect it. The issue of generation was already touched upon in Refs. . We have little to add to this. However, in the context of the Gordon and Savage proposal, it is fairly clear that the time needed to generate the cat state would have to be short compared to the decoherence time. If our estimates of Eq.(1) are correct, symmetrization procedure appears necessary for the success of such schemes.
Detection of Schrödinger cat states is perhaps a more challenging subject. In principle, states of the form $`(|N,0+|0,N)/\sqrt{2}`$ have a character of GHZ states, and one could envision performing measurements analogous to those suggested in and carried out in , where a 4-atom entangled state was studied. However, this sort of parity-check strategy, appropriate for $`N10`$, is likely to fail when $`N`$ is larger, or when (as would be the case for the “quasi-squeezed” states anticipated here ) $`N`$ is not even well defined.
A strong circumstantial evidence can be nevertheless obtained from two measurements. The first one would consist of a preparation of the cat state, and of a measurement of the internal states of the atoms. It is expected that in each instance all (to within the experimental error) would turn up to be in either A or B states. However, averaged over many runs, the number of either of these two alternatives would be approximately equal. Decoherence in which the environment also “monitors” the internal state of the atoms in the A versus B basis would not influence this prediction. We need to check separately whether the cat state was indeed coherent. To do this, one could evolve the system “backwards”. However, this is not really necessary. For, as Gordon and Savage point out, when, in their scheme, we let the system evolve unitarily for more or less twice the time needed for the generation of the cat state, it will approximately return to the initial configuration. Thus, we can acquire strong evidence of the coherence of the cat provided that this unitary return to the initial configuration can be experimentally confirmed.
These are admittedly rather vague ideas, which serve more as a “proof of principle” rather than as a blueprint for an experiment. Nevertheless, they may, we hope, encourage more concrete investigation of such issues with a specific experiment in mind.
## VI Acknowledgements
We are indebted to E.Cornell, R.Onofrio, E.Timmermans, and especially to J.Anglin for very useful comments. This research was supported in part by NSA. |
warning/0001/hep-ph0001198.html | ar5iv | text | # A HAMILTONIAN MODEL FOR MULTIPLE PRODUCTION IN HADRON-HADRON COLLISIONS
## 1 Description of the model
A Hamiltonian model for the description of the multiple production process in hadron-hadron collision is presented with the main aim of bringing together different observables within a unique frame. A particular attention is given to those features of the inelastic processes that can give informations on the proton structure.
The physical ingredients of this model are the following:
A sharp distinction is made between the soft dynamics, which provides the binding of the partons in the hadrons and also the final hadronization of the shaken-off partons and the hard dynamics that causes the parton scattering.
The hard collision gives a finite transverse momentum to the partons, which remains however small with respect to the typical longitudinal momentum. Hard rescattering is included, but not hard branching of the partons.
The discrete quantum numbers spin and colour are not taken into account.
The model describes the hadrons as sets of bound partons which, due to the interaction, may become finally unbound and show up as jets; the hadronization process is not described.
The only detailed kinematics is the transverse one, the whole treatment lies within the frame of the eikonal formalism. The sharp distinction between backward and forward degrees of freedom allows the following formulation: There are operators for the bound backward and forward partons $`a_b,a_f`$ and operators for the unbound partons $`c_b,c_f`$. Both are local in $`𝐛`$, the transverse impact parameter of the parton, and they have the standard commutation relations, where every backward operator commutes with every forward operator and every $`a`$ commutes with every $`c`$; so we can write a free Hamiltonian:
$$_o=\underset{v=f,b}{}\omega d^2b[a_v^{}(𝐛)a_v(𝐛)+c_v^{}(𝐛)c_v(𝐛)]$$
(1)
The interaction that we want to describe is the hard collision of two bound partons that give rise to two unbound partons is such a way however that they keep their property of being either backward or forward. Thus the interaction Hamiltonian is written as:
$$_I=\lambda d^2bh_b(𝐛)h_f(𝐛)h_v(𝐛)=c_v^{}(𝐛)a_v(𝐛)+a_v^{}(𝐛)c_v(𝐛).$$
(2)
With this choice it results $`[_o,_I]=0,`$ but the theory is not trivial even though it has been much simplified, the S-matrix can be written in the form $`𝒮=\mathrm{exp}[i\tau ]`$ where $`\tau `$ is an interaction time.
An alternative, and more realistic, form gives a finite size to the hard interaction and a discretization of the transverse plane. The size $`\mathrm{\Delta }`$ is related to the cut-off in the transverse momentum that must be put in order to be allowed to perform perturbative calculations, so the natural choice is $`\mathrm{\Delta }1/p_{}^2`$; this choice lead also to the interpretation of $`\tau 1/p_{}\sqrt{\mathrm{\Delta }}`$. The commutation relations become $`[A_{v,j},A_{u,i}^{}]=\delta _{i,j}\delta _{u,v},`$ and so on. In this way we get:
$$_o=\underset{v,j}{}\mathrm{\Omega }[A_{v,j}^{}A_{v,j}+C_{v,j}^{}C_{v,j}]$$
(3)
$$_I=(g/\sqrt{\mathrm{\Delta }})\underset{j}{}H_{b,j}H_{f,j}H_{v,j}=C_{v,j}^{}A_{v,j}+A_{v,j}^{}C_{v,j}.$$
(4)
The coupling constant $`g`$ is dimensionless, it is related to the previous coupling constant by $`g=\lambda /\sqrt{\mathrm{\Delta }}`$. Correspondingly the S-matrix is
$$𝒮=\underset{j}{}𝒮_j,𝒮_j=\mathrm{exp}[i(g/\sqrt{\mathrm{\Delta }})\tau H_{b,j}H_{f,j}],$$
(5)
with the previous interpretation of $`\tau `$, we get the simpler form:
$$𝒮_j=\mathrm{exp}[igH_{b,j}H_{f,j}].$$
(6)
In order to apply this model one must choose a definite initial state; it will be factorized in the same way as the S-matrix: as far as its structure in a site $`j`$ is concerned there are no strong indications. A possible choice, related to some theoretical ideas about the nonperturbative partonic structure of the hadrons, is a local coherent state, so we write
$$|I>=\underset{j}{}|I>_j,|I>_j=\mathrm{exp}[(|F_b|^2+|F_f|^2)/2]\mathrm{exp}[F_bA_b^{}+F_fA_f^{}]|>_j$$
(7)
It has to be noted that the weight $`F`$ of the coherent state may vary from site to site. For simplicity the index $`j`$ will be not written out, whenever possible.
It is easer to express $`|I>_j`$ and especially to perform the subsequent calculations in the basis generated by the auxiliary operators $`P`$ and $`Q`$.
$$P=(C+A)/\sqrt{2}Q=(CA)/\sqrt{2}.$$
(8)
In terms of them on gets
$$_o=\underset{v,j}{}\mathrm{\Omega }[P_{v,j}^{}P_{v,j}+Q_{v,j}^{}Q_{v,j}],H_{v,j}=P_{v,j}^{}P_{v,j}Q_{v,j}^{}Q_{v,j}$$
(9)
## 2 Some results of the model
### 2.1 Inelastic cross section
In the discrete formulation for the states and for the S-matrix the inelastic cross section is now calculated. In the basis generated by the operators $`P`$ and $`Q`$ the operator $`𝒮_j`$ is diagonal, so it is easy to calculate the matrix element $`S_j={}_{j}{}^{}<I|𝒮_j|I>_j`$, it has the expression:
$$S_j=N^2\underset{k_1\mathrm{}k_4}{}\frac{1}{k_1!k_2!k_3!k_4!}(|F_b|^2/2)^{k_1+k_2}(|F_f|^2/2)^{k_3+k_4}\mathrm{exp}[ig(k_1k_2)(k_3k_4)].$$
(10)
The indices $`k_1,k_2,k_3,k_4`$ refer to the quanta created by$`P_b^{},Q_b^{},P_f^{},Q_f^{}`$ respectively. The normalizing factor is $`N=\mathrm{exp}\left[(|F_b|^2+|F_f|^2)/2\right].`$ Using the representation:
$$\mathrm{exp}[ig(k_1k_2)(k_3k_4)]=(2\pi )^1𝑑u𝑑v\mathrm{exp}\left[iuv+i\alpha u(k_1k_2)+i\beta v(k_3k_4)\right],$$
(11)
with $`\alpha \beta =g`$, the multiple sum in the expression of $`S_j`$ can be transformed into an integral. In the final result we use the definitions $`T_v=|F_v|^2`$ and we have:
$$S_j=(2\pi )^1𝑑u𝑑v\mathrm{exp}[iuv]\mathrm{exp}\left[T_b(1\mathrm{cos}\alpha u)T_f(1\mathrm{cos}\beta v)\right].$$
(12)
When the distribution functions $`F_v`$ do not vary strongly from site to site one can devise a continuum limit. We consider a relation $`Ff\sqrt{\mathrm{\Delta }}`$; therefore, if $`f`$ is not singular, in the expression of $`S_j`$ the terms $`|F|^2`$ become small, the exponential in the integral representation can be expanded and integrated term by term with the result:
$$S_j1|F_b|^2|F_f|^2(1\mathrm{cos}g)+(1/2)|F_b|^2|F_f|^2(|F_b|^2+|F_f|^2)(1\mathrm{cos}g)^2+\mathrm{}$$
(13)
With the normalization we are using the inelastic cross section at fixed hadronic impact parameter $`𝐁`$ is given by
$$\sigma (𝐁)=2<I|(1\mathrm{}𝒮)|I>\left|<I|(1𝒮)|I>\right|^2.$$
(14)
The product of the matrix elements is expressed as the exponential of the sum of the logarithms and the sum $`\mathrm{\Delta }_j`$ is finally converted into the integration $`d^2b`$ with the final result:
$$\sigma (𝐁)=1\mathrm{exp}\left[\widehat{\sigma }d^2b|f_b(𝐛)|^2|f_f(𝐛𝐁)|^2+\widehat{\sigma }^2\mathrm{}\right]$$
(15)
the parameter $`\widehat{\sigma }=2\mathrm{\Delta }(1\mathrm{cos}g)`$ has the role of elementary partonic cross section.
The form of the inelastic cross section is quite usual,it contains however the explicit indication of the possible corrections due to rescattering, they will be discussed below, where a nonuniform model of the hadron will be explored. The cross section arises from the integration over the impact parameter, the result depends very much on the properties of the distribution functions $`f`$, which appear through their squared absolute values $`t_v(𝐛)=|f_v(𝐛)|^2`$, giving the transverse density of bound partons. In a first discussion the distribution is taken to be completely uniform in $`b`$ by setting:
$$t_b(𝐛)=\rho _b\vartheta (R|b|),t_f(𝐛)=\rho _f\vartheta (R|b|);$$
(16)
elementary geometrical considerations give $`|B|=2R\mathrm{cos}\gamma /2`$ with $`0\gamma \pi `$. The exponent in the integrand is given by the partial superposition of the two disks. Since the superposition area is $`W=R^2(\gamma \mathrm{sin}\gamma )=\pi R^2\xi `$ The expression for the inelastic cross section is:
$$\sigma _{in}=2\pi R^2_o^\pi 𝑑\gamma \mathrm{sin}\gamma \left[1\mathrm{exp}[\nu \xi ]\right],$$
(17)
with $`\nu =\widehat{\sigma }\rho _b\rho _f\pi R^2`$.
It is possible to give a simple analytical form for $`\sigma _{in}`$ in the two limiting situations of very small or very large $`\nu `$. In the first case one gets
$$\sigma _{in}=\pi R^2\nu .$$
(18)
In the second case we can start from the expression
$$\sigma _{in}=2\pi R^2[2D(\nu )]$$
(19)
where the real function $`D(\nu )`$ defined by eq(17) is monotonically decreasing, for large $`\nu `$ it results $`D(\nu )2(6\nu ^2/\pi ^2)^{1/3}\mathrm{\Gamma }(2/3)`$ so that the geometrical limit of black disks $`4\pi R^2`$ is approached.
### 2.2 Pair and double-pair production
The production of a pair can seen either as production of a backward parton or of a forward parton, the Hamiltonian being fully symmetric, so we can choose, arbitrarily to look at the forward particles; successively we shall investigate how much the rescattering processes may destroy the sharp correlation between backward and forward scattered partons. We start from the computation of the inclusive production from a single site. Straightforward, although lengthy calculations, easier in the basis generated by the operators $`P`$ and $`Q`$, yield the result
$$<X_j>={}_{j}{}^{}<I|𝒮_j^{}C_f^{}C_f𝒮_j|I>_j=(T_f/2)[1\mathrm{exp}[T_b(1\mathrm{cos}(2g)]].$$
(20)
We can now go to the continuum limit, always under the hypothesis of smooth distributions $`t_v(𝐛)`$; we find in eq(20) the function $`\mathrm{cos}(2g)`$ instead of $`\mathrm{cos}g`$, so we must define a quantity $`\kappa =(\mathrm{\Delta }/2)[(1\mathrm{cos}(2g)]`$, which is related to $`\widehat{\sigma }`$ in this way: $`\kappa =\widehat{\sigma }[1+\mathrm{cos}(g)]/2`$. The two constants coincides evidently for small $`g`$, when both are: $`\widehat{\sigma }=\kappa =g^2\mathrm{\Delta }`$, but the rescattering corrections, higher powers in $`g^2`$, are different.
In the smooth continuous limit we expand the exponential of eq. (4.3) and we get the usual expression:
$$X(𝐁)=\kappa d^2bt_f(𝐛)t_b(𝐛𝐁)$$
(21)
and the inclusive one-particle cross section is
$$D_1=X(𝐁)d^2B$$
(22)
Analogously the two-particle inclusive cross section is found to be
$$D_2=\kappa ^2d^2Bd^2bd^2b^{}t_f(𝐛)t_b(𝐛𝐁)t_f(𝐛^{})t_b(𝐛^{}𝐁);.$$
(23)
A ratio of the quantities that have been now calculated which is of phenomenological interest is $`\sigma _{\mathrm{eff}}=[D_1]^2/D_2`$, In the limit of rigid disk, using the geometrical considerations one gets
$$\sigma _{\mathrm{eff}}=\frac{\pi R^2}{116/(3\pi ^2)}2.2\pi R^2,$$
(24)
The previously calculated expression of $`\sigma _{in}`$ is really the hard part of the total inelastic cross section, where as ”hard” part we intend the contribution of all the events where at least one hard scattering happens. If we believe that, in going on with the total energy these events become more and more important we would like to have this term not too small with respect to the experimental $`\sigma _{in}`$, which in turn appears to be sizably larger, of about a factor 2, with respect to $`\sigma _{\mathrm{eff}}`$, so in this model we would like to approach, even though not reach, the black-hadron, limit which produces $`\sigma _{in}=4\pi R^2`$.
### 2.3 Multiplicity distribution and forward-backward correlations
The distribution of the multiplicities of the produced pairs is calculated by defining the projection operator over the number of free partons. Since we have always the sharp distinction between backward and forward particles we can choose one of the two particle to define the produced pair: to be definite we take the forward particle as signal of the pair production. For a fixed site $`j`$ the number projector may be expressed as:
$$𝒫_n=\frac{1}{n!}:[C^{}C]^ne^{C^{}C}:$$
(25)
The colon indicates the normal ordering of the $`C`$-operators. From this definition one can perform calculations whose qualitative results can be stated in this form: when the total production is not very copious the result may be written as the sum of two Poissonian distributions, one reflects the incoming coherent state the other the rescattering effect , if however the production in a single site is high so that the rescatterig is very important the resulting expression may be put into the form of a Poissonian distribution times another factor, but this further factor is not a small correction, it changes in essential way the shape of the distribution.
The same qualitative result is obtained by studying the forward-backward correlations; we can calculate both the variance $`\mathrm{\Sigma }_f=<X_f^2><X_f>^2`$ with $`<X_f^2>=<I|𝒮^{}C_f^{}C_fC_f^{}C_f𝒮|I>`$ and the covariance
$`\mathrm{\Sigma }_{f,b}=<V><X_f><X_b>`$ with: $`<V>=<I|𝒮^{}C_f^{}C_fC_b^{}C_b𝒮|I>`$
The actual calculations show , as expected, that the correlation
$$\kappa _{f,b}=\frac{\mathrm{\Sigma }_{f,b}}{[\mathrm{\Sigma }_f\mathrm{\Sigma }_b]^{1/2}}$$
(26)
for small values of $`T_v`$, and so for small production, goes to 1, but when $`T_v`$ becomes very large it goes to zero.
### 2.4 Non uniform hadrons
We wish now to explore the possibility that the hadron, and its projection over the transverse plane, shows strong inhomogeneities in the matter density. This is represented by assuming that there are black spots, that cover a limited amount of the transverse area, while a much fainter ”gray” background fills uniformly the rest of the hadron.
$$|I_b>_j=\left[x\mathrm{exp}[|F_b|^2/2]\mathrm{exp}[F_bA_b^{}]+y\mathrm{exp}[|G_b|^2/2]\mathrm{exp}[G_bA_b^{}]\right]|>$$
(27)
Since the two coherent states are not orthogonal the normalization condition is complicated, however if the two thickness are very different the cross term in normalization condition is very small and we are left with $`|x|^2+|y|^21.`$ The expression for the S-matrix may be given in the form
$$𝐒=𝐒_𝐨𝐒_𝐜,𝐒_𝐨=S_{o}^{}{}_{}{}^{w},w=W/\mathrm{\Delta }$$
(28)
the factor $`S_{o}^{}{}_{}{}^{w}`$ gives the contribution of the background scattering.
When one uses the assumption that $`y<<xi.e.`$ that the spots cover a small part of the transverse area it is possible to give a simple expression to $`𝐒_𝐜`$
$$𝐒_𝐜=1wy^2\mathrm{\Delta }[\rho _b(1\widehat{\sigma }\rho _f)+\rho _f(1\widehat{\sigma }\rho _b)]$$
(29)
It may be noticed that the factor, $`wy^2\mathrm{\Delta }`$, represents the part covered by black spots within the interacting area of the hadrons at given B.
The basic elements for calculating the pair and double pair production have already been given. The local production amplitude is the sum of 4 terms, it will be indicated as: $`<X_j>=[x^4X_{o,o}+x^2y^2(X_{s,o}+X_{o,s})+y^4X_{s,s}].`$ The first term represents the pure background interaction, the second and third terms represent the spot background interaction, the fourth term gives the spot-spot interaction. When one consider explicitly the superposition of the two disks it appears that the four terms have the same geometrical properties and we get for $`\sigma _{\mathrm{eff}}`$ the same expression as in the case of a uniform hadron. So, at first sight it seems that nothing is gained by introducing an inhomogeneity into the hadron, but in fact some new features are present. The expression of $`\sigma _{\mathrm{eff}}`$ is purely geometrical, it does not contain the parameters $`x,y`$. The expression of $`\sigma _{in}`$ can be obtained from the S-matrix (see eq.28, 29), so it depends on $`y`$, more in general both on $`x`$ and on $`y`$.
In order to give a clear, although unrealistic evidence of this fact we can consider the limit in which the background is so thin that it contribute negligibly in the inelastic cross section, then the S-matrix element, depends only on the spot-spot interaction In this situation the inelastic cross section goes to zero with $`y^4`$, the corresponding picture would be: few, wholly black spots distributed in a wide and very thin background. The conclusion of this analysis is therefore that in order to have $`\sigma _{\mathrm{eff}}<\sigma _{in}`$ the hadron should be compact, $`i.e.`$ without holes or transparent regions and also with quite sharp edges.
## 3 Further considerations and conclusions
The model allows further derivations that are here only mentioned: it is possible to introduce correlations in b in the case of nonuniform hadron; it is also possible to treat to some extent the longitudinal degrees of freedom provided the distinction between forward and backward particles is kept valid. Moreover, in some simple cases analytically and in more general situation in numerical way, it is possible to go beyond the two extreme cases that have been illustated here $`i.e.`$ the weakly interacting or the totally absorbing situation.
The model presented allows a systematization of different aspects of the hard processes in multiparticle production with a particular attention to the unitarity corrections and suggests also some interpretations in terms of hadron structure. The connection with QCD is not direct as it appears from the fact that the interaction term is quartic while the fundamental QCD interaction term is cubic, in other words the fundamental input is the parton hard scattering, not the branching process.
## Acknowledgments
This work has been partially supported by the Italian Ministery of University and of Scientific and Technological Research by means of the Fondi per la Ricerca scientifica - Università di Trieste .
## References |
warning/0001/cond-mat0001322.html | ar5iv | text | # 1 Introduction
## 1 Introduction
A number of important electromagnetic processes is conditioned by the matter: the Vavilov-Cherenkov radiation, the X-ray transition radiation, the radiation of channeled particles \- . In this connection it is of interest to study an influence of the matter on the radiation of the relativistic charge rotating along a circle in a permanent magnetic field (synchrotron radiation ).
The synchrotron radiation in an infinite uniform medium was studied in and further in . The radiation of a nonrelativistic particle rotating uniformly around a dielectric sphere, and the radiation of the particle rotating in close proximity to the ideally conducting sphere were considered in the . In the expressions were obtained for the spectral and spectral-angular distribution of the radiation intensity without restrictions on the orbit radius and velocity of a particle rotating around a sphere with an arbitrary dielectric permittivity.
In the present paper an analysis of the numerical calculations by the formulae obtained in is carried out. The peculiarities of the radiation conditioned by the matter of a sphere and by its size, are revealed.
## 2 Basic formulae
We present the basic formulae describing the radiation of a particle with the charge $`q`$ and velocity $`v=\omega _er_e`$ uniformly rotating around a sphere in its equatorial plane ($`r_e`$ is the radius of orbit). The magnetic permeability of the sphere we take equal to unity and consider its dielectric permittivity $`\epsilon _0`$ as an arbitrary real quantity (we do not take into account the effects connected with the radiation absorption), the sphere radius $`r_o<r_e`$. The radiation intensity at the frequency $`\omega =k\omega _e`$ (after an averaging over the rotation period $`2\pi /\omega _e`$) is determined by the expression
$$I_k=2\frac{q^2\omega _e^2}{c\sqrt{\epsilon _1}}\underset{s=0}{\overset{\mathrm{}}{}}(a_{kE}(s)^2+a_{kH}(s)^2),$$
(1)
where $`\epsilon _1`$ is the dielectric permittivity of a medium surrounded the sphere,
$`a_{kE}=kb_l(E)P_l^k(0)\sqrt{{\displaystyle \frac{(lk)!}{l(l+1)(2l+1)(l+k)!}}},l=k+2s,`$
$`a_{kH}=b_l(H)\sqrt{{\displaystyle \frac{(2l+1)(lk)!}{l(l+1)(l+k)!}}}{\displaystyle \frac{dP_l^k(y)}{dy}},y=0,l=k+2s+1`$ (2)
are the dimensionless amplitudes describing the contributions of multipole of the electric and magnetic kinds, respectively. In Eq.(2) $`P_l^k(y)`$ are the associated Legendre polynomials, and $`b_l`$ is a factor depending on $`k`$, $`x=r_0/r_e`$, $`\epsilon _0`$ and $`\epsilon _1`$:
$`b_l(H)=iu_1\left[j_l(u_1)h_l(u_1){\displaystyle \frac{\{j_l(\underset{¯}{x}u_0),j_l(\underset{¯}{x}u_1)\}}{j_l(xu_0)h_l(xu_1)}}\right],u_i=k\sqrt{\epsilon _i}{\displaystyle \frac{v}{c}},`$
$`b_l(E)=(l+1)b_{l1}(H)lb_{l+1}(H)+{\displaystyle \frac{1}{x^2}}({\displaystyle \frac{1}{\epsilon _0}}{\displaystyle \frac{1}{\epsilon _1}})\times `$
$`\times \left[j_{\underset{¯}{l1}}(xu_0)+j_{\underset{¯}{l+1}}(xu_0)\right]\left[h_{\underset{¯}{l1}}(u_1)+h_{\underset{¯}{l+1}}(u_1)\right]{\displaystyle \frac{l(l+1)u_0j_l(xu_0)}{lz_{l1}^l+(l+1)z_{l+1}^l}},`$ (3)
where $`h_l(y)=j_l(y)+in_l(y)`$; $`j_l`$ and $`n_l`$ are the spherical Bessel and Neumann functions, respectively. In Eq.(3) the following notations are introduced:
$`\{a(\underset{¯}{x}u_i),b(\underset{¯}{x}u_j)\}=a{\displaystyle \frac{b}{x}}{\displaystyle \frac{a}{x}}b,f_{\underset{¯}{l}}(y)={\displaystyle \frac{f_l(y)}{\{j_l(\underset{¯}{x}u_0),h_l(\underset{¯}{x}u_1)\}}},`$
$`z_\nu ^l={\displaystyle \frac{u_1j_\nu (xu_0)h_l(xu_1)/\epsilon _1u_0j_l(xu_0)h_\nu (xu_1)/\epsilon _0}{u_1j_\nu (xu_0)h_l(xu_1)u_0j_l(xu_0)h_\nu (xu_1)}}.`$ (4)
The derivation of Eq.(1) is given in .
In the case of homogeneous medium ($`\epsilon _0=\epsilon _1=\epsilon `$)
$`b_l(H)=iuj_l(u),u=k\sqrt{\epsilon }{\displaystyle \frac{v}{c}},`$
$`b_l(E)=iu(2l+1)\left[j_l^{^{}}(u)+{\displaystyle \frac{1}{u}}j_l(u)\right],`$ (5)
and therefore Eq.(1), naturally, does not depend on $`x`$. One can also be convinced that Eq.(1) is transformed into the known formula
$$I_k=kvq^2\frac{\omega _e^2}{c^2}\left[2J_{2k}^{^{}}(2k\beta \sqrt{\epsilon })+(1\frac{1}{\epsilon \beta ^2})\underset{0}{\overset{2k\beta \sqrt{\epsilon }}{}}J_{2k}(y)𝑑y\right],$$
(6)
where $`\beta =v/c`$, $`J_k(y)`$ is the integer-order Bessel function, and $`\phi ^{^{}}(y)=d\phi /dy`$.
## 3 Results of numerical calculations
In Fig.1 along the axis of ordinates we plotted an average number of electromagnetic field quanta
$$n_k=\frac{2\pi I_k}{k\mathrm{}\omega _e^2},$$
(7)
radiated per one period of rotation of electron with the energy 2 MeV (the logarithmic scale), and along the axis of abscissa an order of radiated harmonic in the range $`1k50`$ is plotted. The function $`n_k`$ is presented for the four values of $`x`$. The curves $`a`$, $`b`$, $`c`$, $`d`$ are the polygonal lines connecting the points with different $`k`$ and the same $`x_a`$, $`x_b`$, $`x_c`$ and $`x_d`$, respectively. The line $`a`$ describes a rotation in vacuum ($`x_a=0`$), and the line $`b`$ describes a rotation in the continuous medium ($`x_b=\mathrm{}`$) with the dielectric permittivity $`\epsilon =3`$ (the Cherenkov condition is satisfied). The calculations were carried out by the formula (6). For simplicity the dependence of $`\epsilon `$ on $`k`$ (the dispersion) is not taken into account. It followed from the plots that in a continuous media
$$n_k(\mathrm{})\frac{ve^2}{hc^2}\left(1\frac{1}{\epsilon \beta ^2}\right)<\frac{e^2}{hc}0.05$$
(8)
is larger than the analogous quantity $`n_k(0)`$ in the empty space. A difference between $`n_k(\mathrm{})`$ and $`n_k(0)`$ is conditioned by the contribution of the Cherenkov’s quanta. Along with this, the specific oscillations are revealed on the curve $`b`$. They results from the interference of waves in the conditions when the velocity of the electromagnetic waves propagation is lower than the velocity of the source motion $`c/\sqrt{\epsilon }<v`$.
A similar pattern should be observed also in the case when a medium has finite sizes. In the section 2 we considered the case of a sphere with the radius $`r_o`$, around of which electron rotates at the distance $`r_er_o`$. The polygonal lines $`c`$ and $`d`$ represent the results of calculations by the formula (1) for the two fixed values $`r_o/r_e=0.974733692=x_c`$ and $`0.980861592=x_d`$, respectively. The dielectric permittivity of the sphere $`\epsilon _0=3`$. Outside the sphere there is a vacuum ($`\epsilon _1=1`$). The electron energy $`E_e=2MeV`$. As it is seen, the specific oscillations are observed also in this case. However, there are also the peaks, and on the corresponding harmonics ($`k=26`$ for the case $`c`$ and $`k=40`$ for the case $`d`$) the radiation is abnormally intensive:
$`n_{26}(x_c)=4300\text{for the curve }c,`$
$`n_{40}(x_d)=94\text{for the curve }d.`$ (9)
At the same time on the neighbouring harmonics $`n_k(x)`$ is of the order $`n_k(\mathrm{})`$.
In the empty space the radiation intensity $`I_k`$ reaches a maximum on the harmonic with $`k_{max}=26`$: <sup>2</sup><sup>2</sup>2This result is obtained also from the formula $`k_{max}=0.44(E_e/m_ec^2)^3`$ which is valid for ultrarelativistic electron . $`I_{26}(0)=0.96e^2\omega _e^2/c`$. On this harmonic an influence of the sphere with the radius $`r_o=0.974733692r_e`$ is the most intensive: $`I_{26}(x_c)/I_{26}(0)2.5310^6`$ (just this value of $`r_o`$ is chosen in the case of the curve $`c`$). An analogous situation is possible also on other harmonics. For example, on the harmonic with $`k=40`$ an influence of the sphere is maximal at $`r_o=0.980861592r_e`$ (the curve $`d`$). In this case $`I_{40}(x_d)/I_{26}(0)55700`$.
Figs.2 and 3 show the dependence of $`n_k(x)`$ on $`x`$ for the harmonics with $`k=26`$ and $`k=40`$, respectively. In this plots also $`\epsilon _0=3`$, $`\epsilon _1=1`$ and $`E_e=2MeV`$. Against a background of the oscillations of the function $`n_k(x)`$, the extremely narrow and very high peaks are observed (on the right-hand part the function $`n_k(x)`$ is shown in the vicinity of the maximal peak). Already at a small deviation (along the axis of abscissa) from the centre of any of these peaks $`n_k`$ rapidly decreases. Therefore the value $`x=r_o/r_e`$ must be fixed with a high accuracy (for example, by an external electric field sustaining a uniform rotation of a particle). The energy radiated per one period of the electron rotation, is equal to
$$\frac{2\pi }{\omega _e}I_k=k\mathrm{}\omega _en_k.$$
(10)
The radiative losses are negligible if the cyclic frequency
$$\omega _e\frac{E_e}{k\mathrm{}n_k}10^{13}\frac{E_e}{MeV}\frac{10^8}{kn_k}Hz.$$
(11)
An analogous pattern takes place for other $`1<\epsilon _05`$ and $`E_e5MeV`$, when the Cherenkov condition is satisfied (see Table1). Moreover, in certain cases (see the 2-4th rows of Table 1) one can observe a superintensive radiation with
$$n_k>\frac{2\pi r_e}{\lambda _k}=k\frac{v}{c}.$$
(12)
Table 1: The average number $`n_k`$ of electromagnetic field quanta emitted per revolution
of electron.
| | | Rotation in a continuos medium | Rotation around a sphere in a vacuum | |
| --- | --- | --- | --- | --- |
| $`k`$ | $`E_e`$ | $`\epsilon =1`$ $`\epsilon =3`$ $`\epsilon =5`$ | $`\epsilon =3`$ | $`\epsilon =5`$ |
| | MeV | $`n_k`$ $`n_k`$ $`n_k`$ | $`\mu `$ $`n_k(\mu )`$ | $`\mu `$ $`n_k(\mu )`$ |
| | $`1`$ | $`3.0710^4`$$`2.3710^2`$$`3.1810^2`$ | $`6.6433228`$ $`4.13`$ | $`5.2992`$ $`1.76`$ |
| $`20`$ | $`3`$ | $`2.7210^3`$$`3.3210^2`$$`3.3310^2`$ | $`0.5432354`$ $`201`$ | $`3.482`$ $`0.34`$ |
| | $`5`$ | $`3.0010^3`$$`3.4210^2`$$`3.6310^2`$ | $`1.480803`$ $`133`$ | $`2.596109`$ $`133`$ |
| | $`1`$ | $`2.3910^5`$$`1.9310^2`$$`3.1110^2`$ | $`0.82132`$ $`9.64`$ | $`1.13910742`$$`2260`$ |
| $`40`$ | $`3`$ | $`1.5710^3`$$`2.9010^2`$$`3.7710^2`$ | $`1.2224`$ $`0.65`$ | $`0.9986`$ $`0.65`$ |
| | $`5`$ | $`1.8510^3`$$`3.2210^2`$$`3.4710^2`$ | $`4.801`$ $`0.16`$ | $`1.50036`$ $`1.45`$ |
Note: $`\epsilon `$ is the dielectric permittivity of the matter. In the case of a sphere for every three values of $`k,E_e`$ and $`\epsilon `$ we chosed and presented one value of the ratio of the sphere radius to the radius of the electron orbit $`r_o/r_e=10.01\mu `$, for which $`n_k(\mu )`$ is considerably larger than $`e^2/hc`$.
The formulae (3) are not valid for electron rotating inside a spherical cavity in an infinite medium, and therefore we did not carry out the corresponding calculations.
The numerical calculations were duplicated by two independent programs. One of them, a more simple, was made with the help of the Mathematica, and an another, more fast-acting, on the Pascal language.
## 4 Conclusions
We calculated the intensity of radiation for electron with an energy of several MeV uniformly rotating around a sphere in its equatorial plane. The matter of the sphere is regarded as transparent, and its dielectric permittivity $`1<\epsilon 5`$. It is obtained that on the average the $`n>k`$ quanta of the electromagnetic field may be radiated per revolution of electron, where $`k`$ is the number of the radiated harmonic ($`k50`$). In the absence of a sphere or at the rotation of electron in an infinite medium with the same $`\epsilon `$, the analogous quantity $`n_k<0.05e^2/hc`$. Such an intense radiation takes place when the sphere surface is at a specific distance from the electron orbit and when the Cherenkov condition for electron and the matter of the sphere is satisfied.
The authors are grateful to A. R. Mkrtchyan for the continued interest to this work and support.
Figure captions:
Fig.1: Average number $`n_k(x)`$ of electromagnetic field quanta emitted per revolution of electron, as a function of the radiated harmonic’s number $`k`$. The polygonal lines $`a,b,c`$ and $`d`$ differ by the value of $`x`$ (the ratio of the sphere radius to the radius of the electron orbit): $`x_a=0`$ (vacuum), $`x_b=\mathrm{}`$ (infinite medium), $`x_c0.9747337`$, $`x_d0.9808616`$. The dielectric permittivity of the matter $`\epsilon =3`$, the electron energy $`E_e=2MeV`$.
Fig.2: The same quantity, as in Fig.1, depending on $`x`$. A number of the radiated harmonic is fixed: $`k=26`$. Here also $`\epsilon =3`$ and $`E_e=2MeV`$. On the right-hand side the function $`n_k(x)`$ is plotted in the vicinity of the maximal peak.
Fig.3: The same dependence, as in Fig.2, in the case $`k=40`$. |
warning/0001/physics0001072.html | ar5iv | text | # Monodisperse approximation in the metastable phase decay
## 1 Homogeneous nucleation
The condensation kinetics equation for the number $`G`$ of the molecules in the liquid phase can be written in a well known form ,
$$G(z)=f_0^z𝑑x(zx)^3\mathrm{exp}(\mathrm{\Gamma }G(x))$$
(1)
where parameter $`f`$ is the amplitude value of the droplets sizes distribution $`\mathrm{\Gamma }`$ is some fixed positive parameter.
One can analyse behavior of subintegral function $`g`$ defined by
$$G(z)=_0^zg(z,x)𝑑x$$
(2)
as a function of a size $`\rho =zx`$, $`z`$. This function has the sense of the distribution of the number of molecules in droplets over their sizes $`\rho `$.
In some ”moment” $`z`$ ( or $`t(z)`$) it can be presented in the following form
* When $`\rho >z`$ it is equal to zero (there are no droplets with such a big size)
* When $`\rho <0`$ it is also equal to zero (there aren’t droplets with a negative size)
* At the intermediate $`\rho `$ it grows rather quickly with a growth of $`\rho `$. It is easy to note that it grows faster than $`\rho ^3`$ grows.
Really, if one takes into account that supersaturation decreases in time then we get $`g\rho ^3`$. But supersaturation falls in time and there aren’t so many droplets of the small size as of the big size.
As the result one can see that the function $`g`$ as the function of $`\rho `$ has the sharp peak near $`\rho z`$. This property takes place under the arbitrary $`z`$ (or $`t(z)`$).
The sharp peak of $`g`$ allows to use for $`g`$ the monodisperse approximation - a representation in the $`\delta `$-like form with a coordinate corresponding to a position of the peak of function $`g`$, i.e.
$$g\delta (\rho =z)$$
As the result one can state that the monodisperse approximation is based now. But it is necessary to determine the number of droplets in this approximation.
It would be wrong to include the total number of already appeared droplets in this peak. Really, in the spectrum of sizes there are many droplets with small sizes. One can not describe these droplets as containing the same substance as the droplets of a big size. It would be more correct to exclude them from the substance balance. So, it is necessary to cut off the small droplets. It can be done according to two recipes.
The first recipe is the differential one. One can note that during all times which don’t exceed the time of nucleation essentially the function $`g`$ near maximum is close to
$$g_{appr}=f\rho ^3$$
This approximation corresponds to the constant value of supersaturation.
One can cut off this approximation at a half of amplitude value (i.e. at a level $`fz^3/2`$). Then one can get for the width $`\mathrm{\Delta }_{diff}z`$ the following expression
$$\mathrm{\Delta }_{diff}z=(12^{1/3})z$$
This cut off means that all droplets $`\rho <z\mathrm{\Delta }_{diff}z`$ are excluded from consideration and all droplets with $`\rho >z\mathrm{\Delta }_{diff}z`$ are taken into account in a $`\delta `$-like peak.
The second recipe is the integral one. One can integrate $`g_{appr}`$ and require that
$$_0^zg_{appr}(z,x)𝑑x=Nz^3$$
An integration gives
$$_0^zg_{appr}(z,x)𝑑x=f_0^z(zx)^3𝑑x=f\frac{z^4}{4}$$
The width of spectrum is defined from condition that the number of droplets has to be equal to the amplitude multiplied by the width of spectrum $`\mathrm{\Delta }_{int}z`$:
$$N=f\mathrm{\Delta }_{int}z$$
This gives the following expression
$$\mathrm{\Delta }_{int}z=z/4$$
One can see that $`\mathrm{\Delta }_{diff}z`$ and $`\mathrm{\Delta }_{int}z`$ practically coincide. This shows the high selfconsistency of this approximation. The second recipe will be more convenient for concrete calculations.
In fig.1 one can see the application of the monodisperse approximation in the homogeneous case.
As the result one can say that all parameters of approximation are defined. Now it will be used to solve (1).
Instead of (1) one can get
$$G(z)=N(z/4)z^3$$
where
$$N(z/4)=fz/4$$
is the number of droplets formed until $`t(z/4)`$. This leads to
$$G(z)=fz^4/4$$
which coincides with the resulting iteration in the iteration method , . It is known (see , ) that this expression is very accurate which shows the effectiveness of the monodisperse approximation. Here the cut off of the tail of the sizes spectrum compensates the unsymmetry of the initial spectrum.
The main result of the nucleation process is the total number of the droplets which can be found as
$$N_{tot}=f_0^{\mathrm{}}𝑑x\mathrm{exp}(\mathrm{\Gamma }G(x))𝑑x$$
or
$$N_{tot}=f_0^{\mathrm{}}𝑑x\mathrm{exp}(f\mathrm{\Gamma }z^4/4)=f^{3/4}\mathrm{\Gamma }^{1/4}D$$
where
$$D=_0^{\mathrm{}}\mathrm{exp}(x^4/4)𝑑x=1.28$$
The error of this expression is less than two percents (it is the same as in the iteration method).
## 2 Heterogeneous condensation on similar centers
The condensation equations system can be written in the following form
$$G(z)=f_0^z𝑑x(zx)^3\mathrm{exp}(\mathrm{\Gamma }G(x))\theta (x)$$
$$\theta (z)=\mathrm{exp}(b_0^z\mathrm{exp}(\mathrm{\Gamma }G(x))𝑑x)$$
with positive parameters $`f`$, $`b`$, $`\mathrm{\Gamma }`$. An appearance of a new function $`\theta `$ which is a relative number of free heterogeneous centers requires the second equation.
The first equation of the system is rather analogous to the homogeneous case. The subintegral function here is also sharp. A function $`\theta `$ is a decreasing function of time according to the second equation of the system. Then the function $`g`$ which is again determined by (2) is more sharp than in the homogeneous case. As far as the supersaturation has to fall one can see that $`g`$ is more sharp than $`g_{appr}`$. It allows to use here the monodisperse approximation for all $`z`$ or $`t(z)`$.
As the result the monodisperse approximation is based for heterogeneous condensation. One needs here only the sharp peak of $`g(\rho )`$ which can be easily seen.
The successive application of the monodisperse approximation in the homogeneous case shows that all droplets necessary for a metastable phase consumption at $`t(z)`$ were formed until $`t(z/4)`$. In the heterogeneous case the exhaustion of heterogeneous centers increases in time. So, all essential at $`t(z)`$ droplets were formed before $`t(z/4)`$.
At the same time the presence of a long tail in the situation of a weak exhaustion of heterogeneous centers requires to cut off the spectrum for the monodisperse approximation. As far as the long tail is essential in the situation of a weak exhaustion one has to cut off the spectrum by the same recipe as in the situation of the homogeneous condensation: one has to exclude all droplets formed after $`z/4`$ which have the sizes $`\rho <z\mathrm{\Delta }_{int}z=zz/4`$.
One can see in fig. 2 the monodisperse approximation in the situation of the heterogeneous condensation on similar centers. The form of the spectrum in this situation is illustrated in fig. 3. So, the way to construct approximation is known. Now one can turn to concrete calculations.
The number of the droplets formed until $`t(z/4)`$ has to be calculated as
$$N(z/4)=\frac{f}{b}(1\theta (z/4))$$
An approximation for $`G`$ has the form
$$G(z)=\frac{f}{b}(1\theta (z/4))z^3$$
The total number of droplets can be determined as
$$N_{tot}=\frac{f}{b}(1\theta (\mathrm{}))$$
or
$$N_{tot}=\frac{f}{b}(1\mathrm{exp}(b_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }G(x))𝑑x))$$
or
$$N_{tot}=\frac{f}{b}(1\mathrm{exp}(b_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\frac{f}{b}(1\theta (z/4))z^3)𝑑z))$$
or
$$N_{tot}=\frac{f}{b}(1\mathrm{exp}(b_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\frac{f}{b}(1\mathrm{exp}(b_0^{z/4}\mathrm{exp}(\mathrm{\Gamma }G(x))𝑑x))z^3)𝑑z))$$
The last expression has a rather complicate form. It contains several iterations in a hidden form which ensures the high accuracy.
The last expression can be simplified. One of the possible recipes is the following. One can note that an expression for $`G`$ is necessary at $`\mathrm{\Gamma }G1`$. Then $`z`$ attains some values $`\mathrm{\Delta }_\zeta z`$. But until $`\mathrm{\Delta }_\zeta z/4`$ the value $`\mathrm{\Gamma }G`$ is small and $`\mathrm{exp}(\mathrm{\Gamma }G(z))`$ is close to unity. This leads to simplification of last expression which can be written in the following form
$$N_{tot}=\frac{f}{b}(1\mathrm{exp}(b_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\frac{f}{b}(1\mathrm{exp}(bz/4))z^3)𝑑z))$$
Then one can fulfil calculation according to the last formula. The relative error is less than two percents. Here it is a little bit greater than in the homogeneous case because the form of initial spectrum is changed and there is no full compensation of the unsymmetry of spectrum and an exclusion of the tail. The relative error in the situation of heterogeneous condensation on similar centers is drawn in fig. 4.
Now one can turn to explicit calculation of the integral in the last expression. After the appropriate renormalization the subintegral function is more sharp than $`\mathrm{exp}(x^3)`$ and more smooth than $`\mathrm{exp}(x^4)`$. Both these functions have a sharp back front of spectrum. It allows to introduce the characteristic scale $`\mathrm{\Delta }z`$ by equation
$$\frac{f}{b}\mathrm{\Gamma }(1\mathrm{exp}(b(\mathrm{\Delta }z)/4))(\mathrm{\Delta }z)^31$$
Then
$$_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\frac{f}{b}(1\mathrm{exp}(bz/4))z^3)𝑑z=\mathrm{\Delta }z\frac{A+B}{2}$$
where
$$A=_0^{\mathrm{}}\mathrm{exp}(x^3)𝑑x=0.89$$
$$B=_0^{\mathrm{}}\mathrm{exp}(x^4)𝑑x=0.90$$
Now the calculation is reduced to some algebraic manipulations. The error of the last approximation is less than one percent.
As the result
$$N_{tot}=\frac{f}{b}(1\mathrm{exp}(b\mathrm{\Delta }z\frac{A+B}{2}))$$
One can note that it is possible to formulate the recipe already in terms of $`\mathrm{\Delta }z`$. The long way is adopted here to give the most clear picture for the monodisperse approximation.
## 3 Nucleation on several types of heterogeneous centers
The main advantage of monodisperse approximation is the possibility to use it for the condensation on the several types of centers. The iteration procedure can not be applied in this case successfully. The result of calculations according to shows this fact explicitly. The reason is the existence of the cross influence of the different types of centers through vapor consumption.
In the condensation on similar heterogeneous centers in the situation of exhaustion the influence of this phenomena on the vapor consumption isn’t important because in the situation of consumption the converging force of the heterogeneous centers exhaustion is extremely high. But in the situation with two types of heterogeneous centers the exhaustion of the first type centers can have a certain influence on a vapor consumption but the exhaustion of the second type centers is weak and there is no converging force due to the weak exhaustion of the second type centers.
This effect in very thin and it can not be taken into account in the second iteration. But one can not calculate the third iteration analytically and this stops an application of iterations. Really, this phenomena isn’t evident from the first point of view but it exits and leads to the error of the second iteration in many times.
The application of the monodisperse approximation is based on the sharpness of function $`g`$. This property takes place already in this situation. So, there are no objections to apply the monodisperse approximation here.
Here we shall reproduce the same formulas but with the lower indexes which determine the sort of heterogeneous centers.
The system of condensation equations can be written in the following form
$$G_i(z)=f_i_0^z𝑑x(zx)^3\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}G_j(x))\theta _i(x)$$
$$\theta _i(z)=\mathrm{exp}(b_i_0^z\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}G_j(x))𝑑x)$$
where the lower indexes denote the sorts of centers. This system can be seen by the direct generalization of the one type case.
The subintegral function in the substance balance equations is also sharp. As far as all $`\theta _i`$ are the decreasing functions of arguments then the function $`g`$ defined by (2) (with proper indexes) is sharper than without the exhaustion of heterogeneous centers. So, due to the supersaturationdecreasing $`g`$ is more sharp than $`g_{appr}`$. It allows here to use the monodisperse approximation for all $`z`$ or $`t(z)`$.
As the result one can see that the monodisperse approximation in this case is justified on the base of the sharpness of $`g(\rho )`$.
The same properties as in the previous case can be also seen here. One has to cut off the spectrum at $`z/4`$. Here all justifications are absolutely same as in the previous section. The characteristic situation for the nucleation on two types of heterogeneous centers is drawn in fig.5. As the result the way to construct the monodisperse approximation is known. Now one can present calculations.
The number of the droplets formed until $`t(z/4)`$ on the centers of sort $`i`$ has to be calculated as
$$N_i(z/4)=\frac{f_i}{b_i}(1\theta _i(z/4))$$
An approximation for $`G_i`$ can be now presented as
$$G_i(z)=\frac{f_i}{b_i}(1\theta _i(z/4))z^3$$
The total number of droplets is defined as
$$N_{itot}=\frac{f_i}{b_i}(1\theta _i(\mathrm{}))$$
or
$$N_{itot}=\frac{f_i}{b_i}(1\mathrm{exp}(b_i_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}G_j(x))𝑑x))$$
or
$$N_{itot}=\frac{f_i}{b_i}(1\mathrm{exp}(b_i_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}\frac{f_j}{b_j}(1\theta _j(z/4))z^3)𝑑z))$$
or
$`N_{itot}={\displaystyle \frac{f_i}{b_i}}(1\mathrm{exp}(b_i{\displaystyle _0^{\mathrm{}}}\mathrm{exp}(\mathrm{\Gamma }`$
$`{\displaystyle \underset{j}{}}{\displaystyle \frac{f_j}{b_j}}(1\mathrm{exp}(b_j{\displaystyle _0^{z/4}}\mathrm{exp}(\mathrm{\Gamma }{\displaystyle \underset{k}{}}G_k(x))dx))z^3)dz))`$
Now one can simplify the last expression by the same way as in the one type case.
Expressions for $`G_i`$ are essential at $`\mathrm{\Gamma }_jG_j1`$. Then $`z`$ is near $`\mathrm{\Delta }_\zeta z`$. Until $`\mathrm{\Delta }_\zeta z/4`$ the value $`\mathrm{\Gamma }_jG_j`$ is small and $`\mathrm{exp}(\mathrm{\Gamma }_jG_j(z))`$ is near unity. It leads to
$$N_{itot}=\frac{f_i}{b_i}(1\mathrm{exp}(b_i_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}\frac{f_j}{b_j}(1\mathrm{exp}(b_jz/4))z^3)𝑑z))$$
Now one can fulfil the calculations according the explicit formula. The relative error of the last expression is less than five percents (here it increases slightly due to the complex form of the spectrums on different sorts. The relative error in the number of droplets is drawn in fig. 6. The calculation of the last integral is absolutely analogous to the previous section. The subintegral function after renormalization lies between $`\mathrm{exp}(x^3)`$ and $`\mathrm{exp}(x^4)`$. It allows to get the characteristic size $`\mathrm{\Delta }z`$ from
$$\mathrm{\Gamma }\underset{j}{}\frac{f_j}{b_j}(1\mathrm{exp}(b_j(\mathrm{\Delta }z)/4))(\mathrm{\Delta }z)^31$$
Then
$$_0^{\mathrm{}}\mathrm{exp}(\mathrm{\Gamma }\underset{j}{}\frac{f_j}{b_j}(1\mathrm{exp}(b_jz/4))z^3)𝑑z=\mathrm{\Delta }z\frac{A+B}{2}$$
The relative error of the last expression is less than one percent.
As the result
$$N_{itot}=\frac{f_i}{b_i}(1\mathrm{exp}(b_i\mathrm{\Delta }z\frac{A+B}{2}))$$
The formula is similar to the final expression in the previous section. But parameters in the last formula have to be determined in another manner.
The physical sense of the last expression is the separate exhaustion of heterogeneous centers. One sort of centers can influence on the other sort only through a vapor consumption. This fact can be seen also in the initial precise system of the condensation equations.
Fig.1
Monodisperse approximation in homogeneous condensation. Here one can see four pictures for different periods of time (or for different values of $`z`$. One can introduce $`\mathrm{\Delta }z`$ according to $`\mathrm{\Gamma }G(\mathrm{\Delta }z)=1`$ and it will be the characteristic scale of the supersaturation fall. In part ”A” $`z=\mathrm{\Delta }z/2`$, in part ”B” $`z=\mathrm{\Delta }z`$, in part ”C” $`z=3\mathrm{\Delta }z/2`$, in part ”D” $`z=2\mathrm{\Delta }z`$. One can see that the spectrums in part ”A” and part ”B” are practically the same. It corresponds to the property of the similarity of spectrums until the end of the nucleation period.
Fig.2
Monodisperse approximation in condensation on the similar centers. The value $`\mathrm{\Delta }z`$ is the same as in the previous figure (i.e. it is determined without the exhaustion of centers). Now $`\mathrm{\Gamma }G(\mathrm{\Delta }z)<1`$ and $`\mathrm{\Delta }z`$ will be the characteristic scale of the supersaturation fall in the situation without exhaustion. In part ”A” $`z=\mathrm{\Delta }z/2`$, in part ”B” $`z=\mathrm{\Delta }z`$, in part ”C” $`z=3\mathrm{\Delta }z/2`$, in part ”D” $`z=2\mathrm{\Delta }z`$. One can see that the spectrums in part ”A” and part ”B” aren’t similar. Now all spectrums are more sharp than in the homogeneous case.
Fig.3
One can see two curves which are going from $`\rho =z`$ to the small sizes $`\rho `$. The lower curve corresponds to the real spectrum calculated with account of heterogeneous centers exhaustion. The upper curve corresponds to the condensation without exhaustion of heterogeneous centers which is the worst situation where there is converging force due to the centers exhaustion.
Concrete situation drawn here corresponds to $`b=2`$ after renormalization (the values of parameters $`f`$ and $`\mathrm{\Gamma }`$ can be canceled by appropriate renormalization). The value of $`z`$ here equals to $`3\mathrm{\Delta }z/2`$.
The solid lines correspond to the precise numerical solution. The dashed lines correspond to application of monodisperse approximation. One can not separate the numerical solutions from the approximate ones except the slight deviation in the region of small $`\rho `$. As far as all (precise and approximate) solutions will go to zero there will be no deviations for $`z\mathrm{\Delta }z`$ (i.e. we stop at the worst moment).
Fig.4
The relative error of approximate solution for the nucleation on the similar heterogeneous centers. Here the values $`f`$ and $`\mathrm{\Gamma }`$ can be canceled after renormalization and there remains only one parameter $`b`$ which is the argument of the function drawn here. It is clear that this function is small. All asymptotes can be checked analytically (see ).
Fig.5
Characteristic behavior of size spectrums for nucleation on two types of heterogeneous centers. One can cancel $`f_1`$, $`\mathrm{\Gamma }`$ by renormalization. One can put $`f_2<1`$ due to the choice of centers. Here $`b_1=2`$, $`f_2=1/2`$, $`b_2=1/2`$. The value $`z`$ is taken as $`2\mathrm{\Delta }z`$ (see fig.1). The are three curves here. The lower one corresponds to the spectrum of droplets formed on the first type centers, the intermediate one corresponds to the droplets size spectrum for the second type centers, the upper one corresponds to the spectrum calculated without exhaustion of heterogeneous centers (the reasons are the same as in fig.3). The solid lines are the numerical solutions, the dashed lines are the approximate solutions. One can not see the difference for the lower curve. For the upper and intermediate curve one can see only very slight difference. All spectrums are renormalized to have one and the same amplitude (which is marked by $`f`$).
Fig.6
Relative error for the nucleation on two types of centers. For two types of centers there exists five parameters (two of them can be canceled by renormalization). We have already adopt that $`\mathrm{\Gamma }`$ is one and the same for different types of centers (for the reasons see , for numerical results see , for analytical estimates see , for recipes of calculation in this situation see ). We cancel here $`f_1`$, $`\mathrm{\Gamma }`$. We consider $`f_2<1`$ and the first component is the leading component in the metastable phase consumption. So, the worst situation for the error in the droplets number formed on the second type centers (which is drawn here) will be when $`b_2=0`$ and the is no converging force of the centers exhaustion. So, there remain two parameters $`f_2`$ (it is marked by $`f`$) and $`b_1`$ (which is marked by $`b`$). One can see that the error is small. The calculations for the variations of all parameters give the same value of error but it is hard to reproduce these numerical results. |
warning/0001/hep-th0001195.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The BRST quantization procedure for a system of the first class constraints is straightforward. By the definition, the first class constraints form a closed algebra with respect to the commutators (the Poisson brackets). For simplicity we consider only linear algebras – Lie algebras of constraints.
More general systems include the second class constraints as well, whose commutators contain terms which are nonzero on mass shell (on the subspace where all constraints vanish). In the simplest cases these terms are a numbers or central charges, but sometimes, they are operators which act nontrivially on the space of the physical states. Moreover, the commutators between these operators and the constraints can be nontrivial. In some cases the total system of the constraints and the operators mentioned above form a Lie algebra.
So, in such cases we have a system of operators which form a Lie algebra, but the physical meaning of different operators is different. Some of them play the role of constraints and annihilate the physical states, others are nonzero and simply transform the physical states into other ones. It means, that in the BRST approach for the description of the corresponding physical system we can not use the standard BRST charge for the given Lie algebra. Instead, we have to construct the nilpotent BRST charge in a manner, that some of the operators play the role of the first class constraints, others are second class constraints and the others do not imply any conditions on the physical space of the system.
In this letter we demonstrate the possibility of a different BRST constructions for the given system of generators, which form a given Lie algebra and have different physical meaning.
In the Section 2 we discuss the general method of the BRST quantization, when some of the Cartan generators are excluded from the total system of constraints. The method is based on the introduction of an auxiliary representations of the algebra under consideration, in the way that these representations effectively lead to the desired properties of initial generators: some of them are of the first class, some - of the second class and rest of the generators (only the Cartan ones) do not imply any equations on the space of physical states. The auxiliary representations of the algebra are constructed by means of some additional degrees of freedom, which are usually exploited for the quantization of the systems with the second class constraints .
In the Section 3 we describe the construction of auxiliary representations of the algebra by means of the method of induced representations. The resulting operators act in a new Fock space, generated by a set of additional creation and annihilation operators. The construction automatically destroys the hermiticity properties of the generators. The consequence of this fact is that the BRST charge $`𝒬`$ becomes non – hermitian: $`𝒬^+`$ is not equal to $`𝒬`$. As an example of this general construction the case of the $`so(3,2)`$ algebra is considered in details.
In the Section 4 we show, how these hermiticity properties are restored by introduction of a new scalar product with some kernel $`K`$ in this auxiliary Fock space. As a consequence, the new operator $`K𝒬`$ becomes hermitian and can be used for the construction of the lagrangians, which are gauge invariant due to the nilpotency of (non – hermitian) BRST charge $`𝒬`$.
## 2 The general method
In this section, we describe the method of the BRST construction, which leads to the desirable division of the generators of a given Lie algebra into the first and second class constraints. Let $`\widehat{H}^i,(i=1,\mathrm{},k)`$ and $`\widehat{E}^\alpha `$ be the Cartan generators and root vectors of the algebra with the following commutation relations
$`[\widehat{H}^i,\widehat{E}^\alpha ]=\alpha (i)\widehat{E}^\alpha ,`$ (2.1)
$`[\widehat{E}^\alpha ,\widehat{E}^\alpha ]=\alpha ^i\widehat{H}^i,`$ (2.2)
$`[\widehat{E}^\alpha ,\widehat{E}^\beta ]=N^{\alpha \beta }\widehat{E}^{\alpha +\beta }.`$ (2.3)
Roots $`\alpha (i)`$ and parameters $`\alpha ^i,N^{\alpha \beta }`$ are structure constants of the algebra in the Cartan – Weyl basis. Our goal is to construct nilpotent BRST charge, which after quantization leads to the following conditions: all positive root vectors $`\widehat{E}^\alpha (\alpha >0)`$ of the algebra annihilate the physical states. Contrary, the operators $`\widehat{H}^i`$ which form the Cartan subalgebra may or may not be constraints, depending on the physical nature of these operators.
The simplest case, when all Cartan generators annihilate the physical states, is well known. We introduce the set of anticommuting variables $`\eta _i,\eta _\alpha ,`$ $`\eta _\alpha =\eta _\alpha ^+`$, having ghost number one and corresponding momenta $`𝒫_i,𝒫_\alpha =𝒫_\alpha ^+,𝒫_\alpha `$, with the commutation relations:
$$\{\eta _i,𝒫_k\}=\delta _{ik},\{\eta _\alpha ,𝒫_\beta \}=\{\eta _\alpha ,𝒫_\beta \}=\delta _{\alpha \beta }$$
(2.4)
we define the “ghost vacuum” as
$$\eta _\alpha |0=𝒫_\alpha |0=𝒫_i|0=0$$
(2.5)
for positive roots $`\alpha `$. The BRST charge for the Cartan – Weyl decomposition of the algebra has a standard form
$`Q`$ $`=`$ $`{\displaystyle \underset{i}{}}\eta _i\widehat{H}^i+{\displaystyle \underset{\alpha >0}{}}\left(\eta _\alpha \widehat{E}^\alpha +\eta _\alpha \widehat{E}^\alpha \right){\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha \beta }{}}N^{\alpha \beta }\eta _\alpha \eta _\beta 𝒫_{\alpha +\beta }+`$ (2.6)
$`{\displaystyle \underset{\alpha >0,i}{}}\{\alpha (i)\left(\eta _i\eta _\alpha 𝒫_\alpha \eta _i\eta _\alpha 𝒫_\alpha \right)+\alpha ^i\eta _\alpha \eta _\alpha 𝒫_i\}`$
The physical states are then the cohomology classes of the BRST operator.
The quantization in this case is similar to the quantization $`\stackrel{`}{a}`$ la Gupta – Bleuler, because physical states satisfy equations
$$(\widehat{H}^i+\underset{\alpha >0}{}\alpha (i))|Phys=0,\widehat{E}^\alpha |Phys=0$$
only for positive values of $`\alpha `$. The appearance of the $`\alpha (i)`$ in the quantization conditions does not cause problems since these terms can be absorbed after the redefinition of $`(\widehat{H}^i`$ as we shall see below.
The situation becomes different if some of the Cartan operators $`\widehat{H}^i`$, say $`\widehat{H}^{i_l},l=1,2,\mathrm{}N`$ are nonvanishing from the physical reasons. In this case the following method can be used.
First of all we construct some auxiliary representation for the generators $`\widehat{H}^i,\widehat{E}^\alpha `$ of the algebra in terms of additional creation and annihilation operators. The only condition for this representation is that it depends on some parameters $`h^n`$. The total number of these parameters is equal to the number of the Cartan generators, which are nonzero in the physical sector. In what follows, we consider the realizations of the algebra with a linear dependence of the Cartan generators on these parameters: $`H^m(h)=\stackrel{~}{H}^m+c_n^mh^n`$, where $`c_n^m`$ are some constants. The $`h^n`$ dependence of other generators can be arbitrary. In the next section we describe the general method of construction of such representations. Here we simply assume that they exist.
The next step is to consider the realization of the algebra as a sum of ”old” and ”new” generators
$$^i=\widehat{H}^i+H^i(h),^\alpha =\widehat{E}^\alpha +E^\alpha (h).$$
(2.7)
The BRST charge for the total system has the same form as (2.6), with modified generators:
$`𝒬`$ $`=`$ $`{\displaystyle \underset{i}{}}\eta _i^i+{\displaystyle \underset{\alpha >0}{}}\left(\eta _\alpha ^\alpha +\eta _\alpha ^\alpha \right){\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha \beta }{}}N^{\alpha \beta }\eta _\alpha \eta _\beta 𝒫_{\alpha +\beta }+`$ (2.8)
$`{\displaystyle \underset{\alpha >0,i}{}}\{\alpha (i)\left(\eta _i\eta _\alpha 𝒫_\alpha \eta _i\eta _\alpha 𝒫_\alpha \right)+\alpha ^i\eta _\alpha \eta _\alpha 𝒫_i\}`$
The ghost variables $`\eta _{i_l}`$, correspond to the set of nonvanishing generators $`\widehat{H}^{i_l}`$ and therefore one needs to remove the $`\eta _{i_l}`$ dependence
$$𝒬_{i_l}=\eta _{i_l}\{\widehat{H}^{i_l}+\stackrel{~}{H}^{i_l}+c_n^{i_l}h^n+\underset{\beta >0}{}\beta (i_l)\left(\eta _\beta 𝒫_\beta \eta _\beta 𝒫_\beta \right)\}.$$
(2.9)
from the BRST charge. For this purpose consider an auxiliary $`N`$ \- dimensional space with coordinates $`x_{i_l}`$ and conjugated momenta $`p^{i_l}`$, where $`c_n^{i_l}h^n=p^{i_l}`$:
$$[x_{i_l},p^{i_n}]=i\delta _{i_l}^{i_n}.$$
(2.10)
After the similarity transformation, which corresponds to the dimensional reduction
$$\stackrel{~}{𝒬}=e^{i\pi ^{i_l}x_{i_l}}𝒬e^{i\pi ^{i_l}x_{i_l}},$$
(2.11)
where
$$\pi ^{i_l}=\widehat{H}^{i_l}+\stackrel{~}{H}^{i_l}+\underset{\beta >0}{}\beta (i_l)\left(\eta _\beta 𝒫_\beta \eta _\beta 𝒫_\beta \right)$$
(2.12)
the transformed BRST charge $`\stackrel{~}{𝒬}`$ does not depend on the ghost variables $`\eta _{i_l}`$. All parameters $`p^{i_l}`$ in the BRST charge are replaced by the corresponding operators $`\pi ^{i_l}`$. The transformation (2.11) does not change the nilpotency property of the BRST charge. It means that the $`𝒫_{i_l}`$ independent part $`\stackrel{~}{𝒬}_0`$ of the total charge $`\stackrel{~}{𝒬}`$ is nilpotent as well. Moreover, as a consequence of the nilpotency of $`\stackrel{~}{𝒬}`$ all coefficients at the corresponding antighost operators $`𝒫_{i_l}`$ commute with $`\stackrel{~}{𝒬}_0`$. One can show that the quantization with the help of the BRST operator $`\stackrel{~}{𝒬}_0`$ will lead to the desirable reduced system of constraints on the physical states.
## 3 Construction of auxiliary representations of the algebra
Consider the highest weight representation of the algebra under consideration with the highest weight vector $`|\mathrm{\Phi }_V`$, annihilated by the positive roots
$$E^\alpha |\mathrm{\Phi }_V=0$$
(3.1)
and being the proper vector of the Cartan generators
$$H^i|\mathrm{\Phi }_V=h^i|\mathrm{\Phi }_V.$$
(3.2)
As it was shown in the representations of this algebra can be (in principle) constructed by means of the so called Gelfand – Tsetlin schemes . However difficulties in such construction arise, if one considers algebras, different from the simplest ones of rank $`1`$.
In this section we describe another method, based on the construction given in . The representation which is given by (3.1) and (3.2) in the mathematical literature is called the Verma module . Following the Poincare – Birkhoff – Witt theorem, the basis space of this representation is given by vectors
$$|n_1,n_2,\mathrm{},n_r_V=(E^{\alpha _1})^{n_1}(E^{\alpha _2})^{n_2}\mathrm{}(E^{\alpha _r})^{n_r}|\mathrm{\Phi }_V$$
(3.3)
where $`\alpha _1,\alpha _2,\mathrm{}\alpha _r`$ is some ordering of positive roots and $`n_iN`$.
Using the commutation relations of the algebra and the formula
$$AB^n=\underset{k1}{\overset{n}{}}\left(\genfrac{}{}{0pt}{}{n}{k}\right)B^{nk}[[\mathrm{}[A,B],B]\mathrm{}]$$
one can calculate the explicit form of the Verma module. In it was shown that, making use of the map
$$|n_1,n_2,\mathrm{},n_r_V|n_1,n_2,\mathrm{},n_r$$
(3.4)
where $`|n_1,n_2,\mathrm{},n_r`$ are base vectors of the Fock space
$$|n_1,n_2,\mathrm{},n_r=(b_1^+)^{n_1}(b_2^+)^{n_2}\mathrm{}(b_r^+)^{n_r}|0.$$
(3.5)
generated by creation and annihilation operators $`b_i^+,b_ii=1,2,\mathrm{},r`$ with the standard commutation relations
$$[b_i,b_j^+]=\delta _{ij},$$
(3.6)
the Verma module can be rewritten as polynomials in creation operators on the Fock space.
As an explicit example of the construction given above let us consider the representations of $`so(3,2)`$ algebra, which can be used for the description of the higher spin fields. In this case commutation relations between the corresponding generators $`L_1^+,L_{12}^+,L_2^+,T^+,L_1,L_{12},L_2,T,H_1,H_2`$ are given by
$`[L_1,L_1^+]=H_1,[L_2,L_2^+]=H_1,[L_{12},L_{12}^+]=H_1+H_2,`$
$`[T^+,L_1^+]=L_{12}^+,[T^+,L_{12}^+]=2L_2^+,[T,L_1^+]=0,`$
$`[T,L_1]=L_{12},[T,L_{12}]=2L_2,[T,T^+]=H_1H_2,`$
$`[L_2,L_{12}^+]=T,[L_2,T^+]=L_{12},[T,L_2^+]=L_{12}^+,`$ (3.7)
$`[L_1,L_{12}^+]=T^+,[L_{12},L_1^+]=T,[L_{12},L_2^+]=T^+,`$
$`[T,L_{12}]=2L_1^+,[T^+,L_{12}]=2L_1,`$
In this case we take the representation space the following space of vectors
$$|n_V=|n_1,n_2,n_3,n_4_V=(L_1^+)^{n_1}(L_{12}^+)^{n_2}(L_2^+)^{n_3}(T^+)^{n_4}|\mathrm{\Phi }_V$$
and after the simple calculations one obtains
$`L_1^+|n_V=|n_1+1,n_2,n_3,n_4_V,L_{12}^+|n_V=|n_1,n_2+1,n_3,n_4_V,`$
$`L_2^+|n_V=|n_1,n_2,n_3+1,n_4_V,`$
$`T^+|n_V=|n_1,n_2,n_3,n_4+1_Vn_1|n_11,n_2+1,n_3,n_4_V`$
$`2n_2|n_1,n_21,n_3+1,n_4_V,`$
$`H_1|n_V=(2n_1+n_2n_4+h_1)|n_V,`$
$`H_2|n_V=(n_2+2n_3+n_4+h_2)|n_V,`$ (3.8)
$`L_1|n_V=(n_1+n_2n_4+h_11)n_1|n_11,n_2,n_3,n_4_V`$
$`n_2|n_1,n_21,n_3,n_4+1_V+n_2(n_21)|n_1,n_22,n_3+1,n_4_V,`$
$`L_{12}|n_V=(2n_1+n_2+2n_3+h_1+h_21)n_2|n_1,n_21,n_3,n_4_V`$
$`n_3|n_1,n_2,n_31,n_4+1_V+n_3n_1|n_11,n_2+1,n_31,n_4_V`$
$`+n_4(n_41)n_1|n_11,n_2,n_3,n_41_V`$
$`+(h_2h_1)n_4n_1|n_11,n_2,n_3,n_41_V,`$
$`L_2|n_V=(n_2+n_3+n_4+h_21)n_3|n_1,n_2,n_31,n_4_V`$
$`+n_2(n_21)|n_1+1,n_22n_3,n_4_V`$
$`+(h_2h_1+n_41)n_4n_2|n_1,n_21,n_3,n_41_V,`$
$`T|n_V=(h_1h_2n_4+1)n_4|n_1,n_2,n_3,n_41_V`$
$`2n_2|n_1+1,n_21,n_3,n_4_Vn_3|n_1,n_2+1,n_31,n_4_V.`$
It can be seen that the action of the following operators in the Fock space
$`L_1^+=b_1^+,L_{12}^+=b_2^+,L_2^+=b_3^+,`$
$`T^+=b_4^+b_2^+b_12b_3^+b_2,`$
$`T=(h_1h_2b_4^+b_4)b_42b_1^+b_2b_2^+b_3,`$
$`H_1=2b_1^+b_1+b_2^+b_2b_4^+b_4+h_1,`$
$`H_2=b_2^+b_2+2b_3^+b_3+b_4^+b_4+h_2,`$ (3.9)
$`L_1=(b_1^+b_1+b_2^+b_2b_4^+b_4+h_1)b_1b_4^+b_2+b_3^+b_2b_2,`$
$`L_{12}=(2b_1^+b_1+b_2^+b_2+2b_3^+b_3+h_1+h_2)b_2+b_4^+b_4b_4b_1`$
$`+b_2^+b_3b_1b_4^+b_3+(h_2h_1)b_4b_1,`$
$`L_2=(b_2^+b_2+b_3^+b_3+b_4^+b_4+h_2)b_3`$
$`+(h_2h_1)b_4b_2+b_1^+b_2b_2+b_4^+b_4b_4b_2.`$
is identical to the expressions (3) for the Verma module.
As it can be concluded from the realization of the $`so(3,2)`$ algebra, the generators, corresponding to opposite roots are not hermitian conjugated to each other. Indeed, following the usual rules of hermitian conjugation for creation and annihilation operators
$$(b_i)^+=b_i^+,(b_i^+)^+=b_i$$
(3.10)
the operator conjugated to $`L_1^+`$ is not equal to the operator $`L_1`$ since $`(L_1^+)^+=(b_1^+)^+=b_1L_1`$. The same statement is true for all other pairs of root generators. It means that the BRST charge $`𝒬`$, constructed with the help of these operators, though being nilpotent, is not hermitian. This causes the serious problems, because the BRST gauge invariance has been lost and consequently the lagrangians of the form $`L\mathrm{\Psi }|𝒬|\mathrm{\Psi }`$ , are no longer gauge invariant.
## 4 The restoration of the hermiticity properties
The situation, when the generators corresponding to the opposite roots are not mutually hermitian conjugated holds for any algebra under consideration and is a consequence of the method used for the construction of the corresponding representations.
The reason is that if we consider the usual scalar product in the Fock space with basis (3.5), we find that these vectors form the orthogonal (not orthonormal) basis. At the same time the corresponding vectors (3.3) in Verma module are not orthogonal. For example, the scalar product of two vectors
$$|\mathrm{\Phi }_1_V=L_1^+T^+|\mathrm{\Phi }_V,|\mathrm{\Phi }_2_V=L_{12}^+|\mathrm{\Phi }_V$$
(4.1)
is different from zero
$${}_{V}{}^{}\mathrm{\Phi }_1|\mathrm{\Phi }_2_{V}^{}={}_{V}{}^{}\mathrm{\Phi }|TL_1L_{12}^+|\mathrm{\Phi }_{V}^{}=(h_2h_1),$$
(4.2)
where we assumed that $`{}_{V}{}^{}\mathrm{\Phi }|\mathrm{\Phi }_{V}^{}=1`$. Therefore the correspondence between these two spaces is not complete because of the difference in the scalar products of pairs of corresponding vectors.
The idea how to improve the situation lies on the modification of the scalar product in the auxiliary Fock space. The standard scalar product of two vectors of type (3.5), namely $`|\mathrm{\Phi }_1`$ and $`|\mathrm{\Phi }_2`$ is defined as
$$(\mathrm{\Phi }_1|\mathrm{\Phi }_2)_{st}=\mathrm{\Phi }_1|\mathrm{\Phi }_2$$
(4.3)
and is calculated by transition of the annihilation operators $`b_i`$ to the right by means of the commutation relations (3.6) with the subsequent use of the property $`b_i|0=0`$.
Let us introduce the new scalar product in the Fock space
$$(\mathrm{\Phi }_1|\mathrm{\Phi }_2)_{new}=\mathrm{\Phi }_1|K|\mathrm{\Phi }_2,$$
(4.4)
with a kernel $`K`$, which depends on the creation and annihilation operators $`b_i`$ and $`b_i^+`$. The only condition on this new scalar product in the Fock space is that it has to coincide with the scalar product in the Verma module. Therefore taking two arbitrary vectors in the Verma module $`|\mathrm{\Phi }_1_V,|\mathrm{\Phi }_2_V`$ and corresponding vectors in the Fock space $`|\mathrm{\Phi }_1,|\mathrm{\Phi }_2`$ one has the following defining relation:
$$\mathrm{\Phi }_1|K|\mathrm{\Phi }_2={}_{V}{}^{}\mathrm{\Phi }_1|\mathrm{\Phi }_2_{V}^{}.$$
(4.5)
According to this relation the hermiticity properties of the root generators are restored in the following sense. Let us consider the scalar product of the states $`|\mathrm{\Phi }_1_V`$ and $`E^\alpha |\mathrm{\Phi }_2_V`$: $`{}_{V}{}^{}\mathrm{\Phi }_1|E^\alpha |\mathrm{\Phi }_2_{V}^{}`$. Due to hermitian properties of the root generators in the Verma module it coincides with the scalar product of the states $`E^\alpha |\mathrm{\Phi }_1_V`$ and $`|\mathrm{\Phi }_2_V`$ since
$$(E^\alpha |\mathrm{\Phi }_1_V)^+=_V\mathrm{\Phi }_1|E^\alpha .$$
(4.6)
In the Fock space the relation (4.6) looks as
$$(E^\alpha |\mathrm{\Phi }_1_V)^+=_V\mathrm{\Phi }_1|(E^\alpha )^+.$$
(4.7)
So, taking the new scalar product of the pairs of corresponding vectors in the Fock space $`|\mathrm{\Phi }_1,E^\alpha |\mathrm{\Phi }_2`$ and $`E^\alpha |\mathrm{\Phi }_1,|\mathrm{\Phi }_2`$ one has the following relations
$$\mathrm{\Phi }_1|KE^\alpha |\mathrm{\Phi }_2=\mathrm{\Phi }_1|(E^\alpha )^+K|\mathrm{\Phi }_2.$$
(4.8)
Therefore all of the root generators of the algebra under consideration satisfy the relations
$$KE^\alpha =(E^\alpha )^+K,$$
(4.9)
which play the role of hermiticity relations.
Now, consider the part of the BRST charge in the Fock space dependent on the root generators
$$𝒬_{nonh}=\underset{\alpha >0}{}\left(\eta _\alpha E^\alpha +\eta _\alpha E^\alpha \right).$$
(4.10)
being the only non – hermitian part of the BRST charge
$$𝒬_{nonh}^+=\underset{\alpha >0}{}(\eta _\alpha (E^\alpha )^++\eta _\alpha (E^{\alpha )^+})𝒬_{nonh}.$$
(4.11)
It can be easily shown, that the following relations take place
$$𝒬_{nonh}^+K=K𝒬_{nonh},𝒬_{nonh}K=K𝒬_{nonh}^+$$
So one can conclude, that the total BRST charge of the form (2.8) constructed with the help of the generators (2.7) satisfies the modified hermiticity relation
$$𝒬^+K=K𝒬.$$
(4.12)
This gives the possibility to construct the lagrangians of the form $`L\mathrm{\Psi }|K𝒬|\mathrm{\Psi }`$, which are gauge invariant with the following transformation rules for the field $`|\mathrm{\Psi }`$
$$\delta |\mathrm{\Psi }=𝒬|\mathrm{\Psi },\delta \mathrm{\Psi }|=\mathrm{\Psi }|𝒬^+.$$
(4.13)
Obviously the gauge invariance is guaranteed by the nilpotency of $`𝒬`$ and $`𝒬^+`$ and by the relation (4.12).
Bellow we prove the existence of the hermitian kernel $`K`$ and show, how it can be constructed. The central role in this construction will play the matrix of scalar products of basic elements of the Verma module
$$C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}{}_{V}{}^{}n_1,\mathrm{},n_r|m_1,\mathrm{},m_r_{V}^{}.$$
(4.14)
Let us introduce the notion of ancestor for the pair of multiindeces $`\{n_1,\mathrm{},n_r|m_1,\mathrm{},m_r\}`$. It is defined in the following way. Consider a pair of indices $`\{n_k|m_k\}`$ standing on the $`k`$ -th place of the given pair of multiindices. The ancestor is the pair of representations, which has on the $`k`$ -th place the following pair:
$`\{n_km_k|0\}`$ if $`n_k>m_k`$;
$`\{0|m_kn_k\}`$ if $`n_k<m_k`$;
$`\{0|0\}`$ if $`n_k=m_k`$;
It means that we reduce the pair on the maximal common number. We can illustrate graphically this procedure for the case of $`SO(3,2)`$ algebra, which has the rank $`2`$. Its root diagram is shown on the following picture.
Here are drown two different vectors of the Verma module. Each vector corresponds to the line, which begins at the origin and ends at the point $`A`$. Different segments of these lines correspond to negative roots, which are present in the definition of the vector $`|n_1,\mathrm{},n_r_V`$. The first (lower) line corresponds to the vector $`|4,1,1,1_V`$, while the second one corresponds to the vector $`|4,2,0,2_V`$. The lines described above are not unique for the given vectors, since all lines with the same numbers of each negative roots represent the same vector. However this fact does not affect the result obtained with the help of this picture. The vectors with corresponding lines ended at the same point, say at the point $`A`$, are the only ones which can have nonzero scalar product. The pair of representations for the vectors drawn on the picture is $`\{4,1,1,1|4,2,0,2\}`$, while the corresponding ancestor is $`\{0,0,1,0|0,1,0,1\}`$. In general all pairs of representations are divided into the equivalence classes by their ancestors.
The following expression solves the problem of finding the kernel $`K`$.
$$K=\underset{anc}{}(b_1^+)^{n_1}\mathrm{}(b_r^+)^{n_r}C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}(b_1^+b_1,b_2^+b_2,\mathrm{},b_r^+b_r)(b_1)^{m_1}\mathrm{}(b_r)^{m_r},$$
(4.15)
where the summation goes over all possible ancestors and
$`C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}(b_1^+b_1,b_2^+b_2,\mathrm{},b_r^+b_r)`$ are the functions of the number operators $`b_1^+b_1,b_2^+b_2,\mathrm{},b_r^+b_r`$ with the following properties:
$$C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}(0,0,\mathrm{},0)=C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r},$$
(4.16)
$$C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}(l_1,\mathrm{},l_r)=C_{n_1+l_1,\mathrm{},n_r+l_r}^{m_1+l_1,\mathrm{},m_r+l_r},$$
(4.17)
The hermiticity of the kernel $`K`$ is a consequence of the relations
$$C_{n_1,\mathrm{},n_r}^{m_1,\mathrm{},m_r}(l_1,\mathrm{},l_r)=C_{m_1,\mathrm{},m_r}^{n_1,\mathrm{},n_r}(l_1,\mathrm{},l_r)$$
(4.18)
Therefore, the expression (4.15) solves the problem of finding the kernel for the scalar product for an arbitrary Lie algebra.
## 5 Conclusions
It might be interesting to apply the method described in the paper to construct the gauge invariant lagrangians for the particles with the higher spins . Namely, the case of $`so(3,2)`$ algebra which we have considered in details corresponds to the subset of constraints obtained after the quantization of the three – particle bound system (three – point discrete string) . The description of the various irreducible representations of the Poincare group with the corresponding Young tableau having two rows can be achieved after elimination of the Cartan generators of $`so(3,2)`$ algebra from the total set of constraints. The application of the BRST approach given in this letter to the description of the above mentioned system will be given elsewhere.
Acknowledgments. This investigation has been supported in part by the Russian Foundation of Fundamental Research, grant 99-02-18417, joint grant RFFR-DFG 99-02-04022, grant INTAS 96-0308 and grant of the Committee for collaboration Czech Republic with JINR. One of us (M.T.) is grateful to the Abdus Salam International Centre for Theoretical Physics, Trieste, where the part of the present work was done. |
warning/0001/astro-ph0001228.html | ar5iv | text | # Quantitative analysis of WC stars: Constraints on neon abundances from ISO/SWS spectroscopy
## 1 Introduction
Understanding the physics of massive ($`M_{\mathrm{init}}`$$``$25 $`M_{}`$) stars, their atmospheres, radiation, and evolution is important for many aspects of astrophysics since their powerful winds affect the energy and momentum balance of the interstellar medium (ISM). In particular, Wolf-Rayet (WR) stars provide crucial tests of nuclear reaction chains.
However, quantitative analysis of such stars, represents a formidable challenge. The assumptions of plane-parallel geometry and local thermodynamic equilibrium (LTE), which are often adopted for lower luminosity stars, are inadequate. Nevertheless, the properties of a large sample of carbon sequence (WC-type) WR stars have now been quantitatively derived using detailed models, accounting for non-LTE effects, spherical geometry and an expanding atmosphere (Howarth & Schmutz 1992; Koesterke & Hamann 1995). However, recent studies have demonstrated that clumping (Moffat et al. 1988; Hillier 1991, 1996; Schmutz 1997) and line blanketing (Schmutz 1997; Hillier & Miller 1998) may have a significant effect on the derived physical properties of WR stars.
Overall, evolutionary predictions for massive stars (e.g. Meynet et al. 1994) are in good agreement with the observed properties of Wolf-Rayet stars. van der Hucht & Olnon (1985) derived a Ne/He ratio for $`\gamma ^2`$ Vel (WR11, HD 68273, WC8+O7.5) from IRAS space-based observations, which was found to be in good agreement with expectations. However, Barlow et al. (1988, hereafter BRA88) identified a numerical flaw in these calculations and added new ground-based observations to reveal Ne/He=1.0$`\pm `$0.35$`\times `$10<sup>-3</sup>, a factor of six lower than predicted. Is this discrepancy due to failure of evolutionary models, peculiarities of the $`\gamma ^2`$ Vel binary system, or incorrect assumptions for the stellar properties of the WC8 star? The combination of Short Wavelength Spectrometer (SWS), Infrared Space Observatory (ISO) observations of a larger sample of WC stars, plus recent progress in quantitative modelling of WR stars, should provide a definitive answer to this question.
Recent space-based spectroscopy of WR11, obtained with the Infrared Space Observatory (ISO), was presented by van der Hucht et al. (1996). who quoted excellent agreement of the fine-structure neon line fluxes with observations used by BRA88. Morris et al. (1998) subsequently re-estimated Ne/He using contemporary information on the stellar distance (van der Hucht et al. 1997; Schaerer et al. 1997) and mass-loss rate (Stevens et al. 1996) of WR11, which revealed a considerable neon enrichment. Willis et al. (1997, hereafter Paper I) also used ISO/SWS to observe WR146, another WC+O binary, again revealing significantly enriched neon. Recently, Morris et al. (2000) analysed ISO/SWS observations of WR147 (WN8+OB), revealing neon, sulphur and calcium abundances in good agreement with cosmic values, as expected by evolutionary models.
In this paper, we supplement results from Paper I with other WC stars for which ISO-SWS observations are available, HD 156385 (WR 90, WC7) and HD 192103 (WR135, WC8). We also re-analyse WR146 (WC5+O8) using more sophisticated analysis techniques and in the light of other observational evidence (Niemela et al. 1998; Dougherty et al. 2000). A neon abundance is also re-derived for WR11, following recent spectroscopic results from De Marco et al. (2000).
The outline of the present work is as follows. New UV, optical and IR observations of our programme stars are presented in Sect. 2, with basic properties discussed in Sect. 3. The spectroscopic technique is introduced in Sect. 4, and applied in Sect. 5. Our spectroscopic results are discussed in Sect. 6, with neon and sulphur abundances derived in Sect. 7. Finally, conclusions are reached in Sect. 8.
## 2 Observations
The programme Galactic WC stars are listed in Table 1 where we give the various catalogue names and our adopted spectral types, following Smith et al. (1990) and Crowther et al. (1998). We will refer to our programme stars by their WR catalogue number (van der Hucht et al. 1981). Wind velocities are taken from UV resonance line measurements (Prinja et al. 1990; St Louis et al. 1993), except for WR146 for which a terminal velocity was obtained in Paper I from ISO/SWS spectroscopy. In this section, we discuss the ground-based (AAT, INT, ESO) and space-based (ISO, IUE) observations, the journal of which is presented in Table 2.
### 2.1 Ultraviolet spectroscopy
Our UV dataset was obtained solely from the International Ultraviolet Explorer (IUE) archive. Available high dispersion (HIRES), large aperture, short-(SWP) and long-wavelength (LWP, LWR) observations of WR90 and WR135 were combined to provide high S/N datasets (St Louis 1990). Absolute flux calibration was achieved following the calibration curve obtained by Howarth & Phillips (1986). Agreement between our final HIRES datasets and archival low resolution (LORES) observations was found to be excellent.
Willis et al. (1986) have discussed the UV spectral morphology of WC stars, including WR90 and WR135. The principal spectral features for both stars are Si iv $`\lambda \lambda `$1393-1407, C iv $`\lambda \lambda `$1548–51, He ii $`\lambda `$1640, and C iii $`\lambda `$2297. A forest of Fe v-vi lines are present in both stars, with Fe iv also prominent in WR135.
### 2.2 Optical spectroscopy
Previously unpublished spectroscopic observations of WR90 and WR135 were obtained at the 2.5m Isaac Newton Telescope (INT) and 3.9m Anglo Australian Telescope (AAT). The INT dataset was obtained with the IDS, 500mm camera, and GEC CCD during 1991 September, while the AAT dataset was obtained with the RGO spectrograph, 25cm camera, and a MIT/LL CCD during 1998 March. Each dataset was obtained using CCDs and reduced in a standard manner using software available on STARLINK. Spectra were debiased, flat-fielded, optimally extracted, wavelength calibrated, and flux calibrated in the case of the INT dataset. Flux calibration was achieved for the AAT observations via scaling to the level of the blue CTIO spectrophotometry taken from Torres-Dodgen & Massey (1988).
Arc spectra provided a measure of the instrumental resolution: 2–3Å (INT) and 4Å (AAT). Complete spectral coverage between 3800–7300Å (INT) and 5000–10300Å (AAT) was achieved with, respectively, six and one grating settings. The atmospheric absorption bands removed using suitable comparison stars.
The optical spectral appearance of WR146 has previously been discussed in Paper I, while Dougherty et al. (2000) present a new high quality blue optical observation of that star. The morphologies of WR90 and WR135 are relatively similar, dominated by He i-ii and C iii-iv emission features, except that the emission lines of WR90 are substantially broader. C iii $`\lambda \lambda `$4747–51 and C iv $`\lambda \lambda `$5801–12 are the two strongest features in each case. Several O iii-v features are present around $`\lambda \lambda `$2950–3150, and $`\lambda `$5592 with C ii present at $`\lambda `$4267 for WR135.
### 2.3 Near-IR spectroscopy
Our principal near-IR dataset was obtained at the 3.8m U.K. Infrared Telescope (UKIRT) with the cooled grating spectrograph CGS4, the 300mm camera, a 75 l/mm grating and a 62$`\times `$58 InSb array in 1994 August. Observations of WR135 and WR146 were bias-corrected, flat-fielded, extracted and sky-subtracted using cgs4dr (Daly & Beard 1992). Subsequent reductions and analysis were carried out using figaro (Shortridge et al. 1999) and dipso (Howarth et al. 1998). In order to remove atmospheric features, the observations were divided by an appropriate standard star (whose spectral features were artificially removed) observed at around the same time and similar air mass. In regions of low atmospheric transmission at UKIRT the reliability of line shape and strength must be treated with caution (e.g. the P$`\alpha `$ region). The near-IR morphology of WR135 and WR146 have been discussed previously by Eenens, Williams & Wade (1991).
We also utilise intermediate dispersion CCD spectra of WR135 covering 0.97–1.03$`\mu `$m, published by Howarth & Schmutz (1992), and obtained at the INT with the Intermediate Dispersion Spectrograph (IDS) and a GEC CCD in 1990 October. Although Howarth & Schmutz (1992) obtained observations of the He i $`\lambda `$10830+P$`\gamma `$ line, we prefer to use our new, lower resolution UKIRT observations of this feature. (The sharp fall in efficiency of the GEC CCD longward of P$`\gamma `$ leads to an ill-defined continuum.). We attempt an approximate flux calibration for these data, by setting the local continuum via interpolation of the observed 0.7$`\mu `$m (INT) and 1.03$`\mu `$m (UKIRT) flux levels.
### 2.4 Mid-IR spectroscopy
New mid-IR data of WR90 and WR146 were obtained as part of Guest Observer programme, AJWWOLF (P.I. Willis), with the Short Wavelength Spectrograph (SWS; de Graauw et al. 1996) onboard the ESA Infrared Space Observatory (Kessler et al. 1996). Observations of HD 117297 (WR53), HD 164270 (WR103) and HD 165763 (WR111) from the AJWWOLF programme were of insufficient quality to provide mid-IR line flux measurements, and so these were excluded from the present study. Consequently, we have included SWS observations of WR11 and WR135 from the Guaranteed Time programme WRSTARS (P.I. van der Hucht). In all cases the SWS AOT6 observing mode was used to achieve full grating resolution, $`\lambda /\mathrm{\Delta }\lambda \mathrm{\hspace{0.33em}1300}2500`$. The continuous wavelength coverage was 2.60–19.6 $`\mu `$m for datasets from the AJWWOLF programme, and 2.38–45.0$`\mu `$m for WR135 and WR11.
Data of detector “bands” 1 and 2 respectively cover wavelengths of 2.38–4.08 $`\mu `$m using 12 In:Sb detectors and 4.00–12.05 $`\mu `$m with 12 Si:Ga detectors, employing entrance slits that give an effective aperture area of 14 $`\times `$ 20 arcsec on the sky. Band 3A to 3D data cover 12.0–27.6 $`\mu `$m using 12 Si:As detectors, with sky coverage of 14 $`\times `$ 27 arcsec. finally, band 3E and 4 data cover 27.5–45.0$`\mu `$m using 12 Ge:Be detectors, with sky coverage of 20$`\times `$33 arcsec.
The total integration time was set to allow one complete scan over the wavelengths selected within each “AOT band”, defined by the permissible combinations of detector band, aperture, and spectral order (cf. de Graauw et al. 1996). The observation time includes dark current measurements, and a monitor of photometric drift for the detectors of bands 2–3. A drift measurement is not normally made for the relatively stable In:Sb detectors.
The data processing of the SWS data for WR146 was discussed in Paper I, with WR11, WR90 and WR135 reduced in a similar manner. The stellar spectrum of WR11 will be discussed elsewhere (Morris et al. 2000, in prep), with solely mid-IR fine structure line fluxes measured here.
## 3 Interstellar reddening and distances
In this section, interstellar reddenings and distances to the programme stars are discussed. Pre-empting results from Sect. 4, reddenings are directly obtained from comparing theoretical synthetic spectra with de-reddened observations. Distances either follow from cluster/association membership, or assumed absolute magnitudes. Table 3 provides a summary of the derived reddenings and distances for our sample of stars. A comparison with reddenings from the recent literature shows good agreement. The distances to WR135 (Cyg OB3 member) and WR11 (hipparcos) are known with confidence, while those adopted here for WR90 and WR146 deserve comment.
### 3.1 WR 90
WR90 is not a member of an association or cluster. We therefore estimated its distance based on the mean absolute visual magnitude of other Galactic WC7 stars. Unfortunately, all five WC7 stars that are members of associations/clusters are within binary systems (Lundstrom & Stenholm 1984). We restricted the sample to those for which reliable reddenings were known (HD97152=WR42, HD152270=WR79 and HD192641=WR 137) from Morris et al. (1993). An absolute visual magnitude of $`M_\mathrm{v}=`$4.7 mag($`\sigma `$=0.8) was obtained for the WC7 components by comparing their emission line strengths with WR90 (specifically C iii-iv $`\lambda `$ 4650, C iv$`\lambda `$ 5804). Our estimate is in good agreement with van der Hucht et al. (1988) and Smith et al. (1990) who derived $`M_\mathrm{v}=`$4.8 mag. Pre-empting results from Sect. 5 we derive $`E_{BV}`$=0.38 mag, so that the uncertainty in the calibration and reddening ($`\pm `$ 0.02 mag) implies $`d`$=1.55 kpc$`{}_{0.4}{}^{}{}_{}{}^{+0.5}`$.
### 3.2 WR 146
Dougherty et al. (1996) used IR and mm photometry to estimate a distance of 1.2$`\pm `$0.3 kpc towards WR146. Radio observations revealed two separate components, namely the (thermal) WC emission plus the (non-thermal) bow shock emission between the two winds.
Meanwhile, a lower distance of 0.75$`\pm `$0.15 kpc was derived in Paper I, using the mean $`M_v`$ for a WCE star (Smith et al. 1990), plus an assumed spectral type of O8.5 V for the companion. Subsequently, Niemela et al. (1998) used WFPC2 aboard the Hubble Space Telescope (HST) to measure the individual UBV magnitudes of the WC and OB components of WR146. They derived V=13.64 mag for the WC star, 0.24 mag brighter than the OB component. In contrast, a difference of 0.8 mag was assumed in Paper I.
The combined HST and radio datasets indicate that the OB companion possesses a powerful stellar wind, with a giant or supergiant luminosity class. Recent optical spectroscopy supports a supergiant O8 classification (Dougherty et al. 2000). As discussed by Dougherty et al., a large distance to this system would result if the absolute magnitude of the companion was typical of late O supergiants (Conti & Alschuler 1971). Instead, we adopt the distance of 1.4kpc, as derived by Dougherty et al. At this distance, WR146 would be a foreground object to Cyg OB2 (Lundstrom & Stenholm 1984; Torres-Dodgen et al. 1991).
We derived $`E_{\mathrm{B}\mathrm{V}}`$=2.87 mag and R=2.9 using the optical-IR reddening law of Steenman & Thé (1989, 1991). This provided a superior comparison between theoretical predictions and de-reddened observations than with either the Howarth (1983) or Cardelli et al. (1989) laws. Therefore, $`M_v=`$5.3 mag for the WCE component, by far in excess of ‘typical’ WCE stars (Smith et al. 1990).
## 4 Modelling technique
Before discussing the fine analysis of each programme star, we introduce the spectroscopic technique followed here. We use the non-LTE code of Hillier & Miller (1998) which iteratively solves the transfer equation in the co-moving frame subject to statistical and radiative equilibria in an expanding, spherically symmetric and steady-state atmosphere. Relative to earlier versions of this code (Hillier 1987, 1990), two major enhancements have been incorporated, of particular relevance to WC-type stars, namely (i) line blanketing, (ii) clumping. Specific details of the techniques used are provided by Hillier & Miller (1998, 1999), with only a brief overview given here.
As discussed by Hillier & Miller (1998, 1999), extremely complex atomic models are necessary for the quantitative analysis of WC stars, a computationally demanding requirement. Consequently, we make use of the technique of ‘super-levels’, in which several atomic levels of similar energies and properties are combined into a single one, a super level, with the populations of the super level calculated in the rate equations. Populations of individual atomic levels are then calculated by assuming that it has the same departure coefficient as the corresponding super level to which it belongs. In this way, extremely complex atoms of helium, carbon, oxygen and iron can be considered. In Table 4 a total of 1299 full ($`N_\mathrm{F}`$) atomic levels are combined to produce just 440 super levels ($`N_\mathrm{s}`$). The number of iron transitions is limited by $`gf10^4`$, so that a total of 20,007 transitions are included. In addition to the specific models discussed here, we have also considered cases with expanded atomic datasets and ionization stages (e.g. Fe iii), for which the differences in emergent synthetic spectrum were negligible.
Oscillator strengths, collision and photoionization cross-sections are taken from a wide variety of sources (Table 4). The OPACITY project (Seaton 1987, 1995) formed the basis of most radiative rates, supplemented by calculations for CNO by Nussbaumer & Storey (1983, 1984) and Storey (unpublished), and for iron by Becker & Butler (1992, 1995ab) and Butler (unpublished).
The stellar radius ($`R_{}`$) is defined as the inner boundary of the model atmosphere and is located at Rosseland optical depth of 10 with the stellar temperature ($`T_{}`$) defined by the usual Stefan-Boltzmann relation. Similarly, $`T_{2/3}`$ relates to the radius ($`R_{2/3}`$) at which the Rosseland optical depth equals 2/3.
Although the majority of spectral analyses of WR stars have adopted a standard $`\beta `$=1 velocity law, there is both observational and theoretical evidence for a more slowly accelerating outflow (e.g. Schmutz 1997; Lepine & Moffat 1999). Consequently, we adopt a form for the velocity law (Eqn 8 from Hillier & Miller 1999) such that two exponents are considered with the result that acceleration is modest at small radii, but continues to large distances, i.e.
$$v(r)=\frac{v_0+(v_{\mathrm{}}v_{\mathrm{ext}}v_0)(1R_{}/r)^{\beta _1}+v_{\mathrm{ext}}(1R_{}/r)^{\beta _2}}{1+(v_0/v_{\mathrm{core}})(\mathrm{exp}((R_{}r)/h_{\mathrm{eff}}))}$$
Here $`v_{ext}`$ is an intermediate terminal velocity, $`v_{\mathrm{core}}`$ the core velocity (typically a few km s<sup>-1</sup>), $`v_0`$ is the photospheric velocity (typically 100.0 km s<sup>-1</sup>), $`v_{\mathrm{}}`$ is the final wind velocity and $`h_{\mathrm{eff}}`$ the scale height ($``$0.01$`R_{}`$). For all models, we take $`\beta _1`$=1 and $`\beta _2`$=50 as in Hillier & Miller (1999).
There is now overwhelming evidence for the clumped nature of WR stars (e.g., Moffat et al. 1988; Moffat 1999), so we have adopted a simple filling factor approach. We assume that the wind is clumped with a volume filling factor, $`f`$, and that there is no inter clump material. Since radiation instabilities are not expected to be important in the inner wind we parameterise the filling factor so that it approaches unity at small velocities. Clumped and non-clumped spectra are very similar, except that line profiles are slightly narrower with weaker electron scattering wings in the former. Although non-clumped models can be easily rejected, because of the severe line blending in WC winds $`\dot{M}/\sqrt{f}`$ is derived by our spectroscopic analysis, rather than $`\dot{M}`$ and $`f`$. Also, this formulation introduces a revision of the BRA88 analytical formula, which will be addressed in Sect 5.
## 5 Spectroscopic analysis of individual stars
In this section, we discuss the analysis of our programme WC stars. Stellar parameters ($`T_{}`$, log $`L/L_{}`$, $`\dot{M}/\sqrt{f}`$, C/He, O/He) were adjusted until the observed ionization balance, line strengths and de-reddened optical continuum flux distribution were reproduced. Because of the substantial effect that differing mass-loss rates, temperatures and elemental abundances have on the emergent spectrum, this was an iterative process, in which initial model parameters were adopted from the literature if available (e.g. Koesterke & Hamann 1995; Morris et al. 1993). Terminal wind velocities are tabulated in Table 1.
The wind ionization balance was ideally selected on the basis of isolated optical lines from adjacent ionization stages of carbon and/or helium. In practice this was difficult to achieve because of the severe blending in WC winds. The low wind velocity of WR135 allowed He i $`\lambda `$5876 and He ii $`\lambda `$5412 to be used as ionization balance constraints. In other cases, the wind ionization balance was selected on the basis of carbon diagnostics. Unfortunately, from all the possible WC diagnostics, the usual classification lines C iii $`\lambda `$5696/C iv $`\lambda `$5804 are difficult to match, being so sensitive to minor stellar parameter changes. Consequently alternative diagnostics were sought. For WR90, our primary diagnostics were C iii $`\lambda `$8500/C iv $`\lambda `$7736 since these suffered from negligible contamination and are predicted by the model to vary smoothly across the temperature space. In the case of WR146 this spectral region, as well as lines of low ionization, were not available. We selected He ii $`\lambda `$5412/He i $`\lambda `$10830, and C iii $`\lambda `$6740/C iv 1.74$`\mu `$m as our primary ionization diagnostics.
Our experience with a variety of stellar models led us to select the strong ultraviolet P Cygni profiles of carbon (C iii $`\lambda `$1909, 2297) as our principal mass-loss diagnostics. A major limitation with C iv $`\lambda `$1550 was that different mass-loss rates also affected nearby Fe lines, which strongly modulated the predicted strength of the emergent P Cygni profile. Since ultraviolet observations of WR146 are unavailable, we relied on He i $`\lambda `$10830 as the principal mass-loss diagnostic in that case.
As in other recent spectroscopic studies of WC stars, He ii $`\lambda `$5412/C iv $`\lambda `$5471 were selected as the diagnostics for C/He determinations since the relative strength of these features are insensitive to differences of temperature or mass-loss rate. As discussed by Hillier & Miller (1998), this spectral region also contains a number of additional weak features, common to all stars. More problematic is that misleading C/He ratios would be obtained if (i) a limited number of C iv atomic levels were included; (ii) homogeneous models were adopted in which electron scattering wings were incorrectly predicted.
Test calculations were also performed using recombination theory. Comparisons of C/He with our more sophisticated results for WCE stars was found to be reasonable – we obtain C/He=0.08 by number for WR146 in this study, in accord with the value obtained by Eenens & Williams (1992) based on IR recombination lines. WCL stars are more problematic since recombination coefficients for C iii are not available. In addition, the fraction of recombined helium is not straightforward to assess in such studies, so that helium abundances could be underestimated.
Oxygen abundances were more difficult to constrain, as already discussed by Hillier & Miller (1999), with the principal diagnostic region spanning $`\lambda \lambda `$2900–3500. Since this spectral region was absent for WR146, we adopted C/O=4 by number, as predicted by current stellar evolutionary models. Note that our neon abundance determination in Sect. 7 is relatively insensitive to the precise oxygen content. Regarding silicon and iron, we adopt solar abundances since all distances are $``$2 kpc.
We now proceed to discuss individual stars in detail.
### 5.1 HD 156385 (WR 90)
The principal datasets for WR90 comprised HIRES IUE ultraviolet spectroscopy and AAT/RGO spectrograph optical observations. Secondary datasets were the blue CTIO spectroscopy from Torres-Dodgen & Massey (1988) and the 2.6-5$`\mu `$m ISO spectroscopy (longer wavelength data were of insufficient S/N to be used as stellar diagnostics). The combined, UV-optical-IR flux calibrated dataset for WR90 allowed a well constrained reddening of $`E_{\mathrm{B}\mathrm{V}}`$=0.38 mag, in agreement with the determination by Morris et al. (1993) (see Table 3). A distance of 1.55 kpc is implied from our assumed WC7 $`M_v`$-calibration.
The difficulty in identifying the stellar continuum in the rich emission line spectrum of WC stars makes rectification imprecise, as emphasised by Hillier & Miller (1998, 1999). Fig. 1 demonstrates the excellent agreement between the line and continuum distribution of the model spectrum (dotted lines) and observations (solid lines), and includes the true theoretical continuum distribution (dashed lines). The number of line features that are poorly reproduced is small, and includes C iii $`\lambda `$5696 (too weak), $`\lambda `$9710 (too strong), plus C iv $`\lambda `$1548–51, Si iv $`\lambda `$1393–1402 (both too weak because of Fe-absorption) and O vi $`\lambda \lambda `$3811–34 (too weak). The use of absolute fluxes appears to give poor results in the line strengths around C iii $`\lambda `$2297 since the de-reddened and theoretical continuum do not exactly match.
Our analysis reveals stellar parameters of $`T_{}`$=71,000K, log($`L/L_{}`$)=5.5 and $`\dot{M}/\sqrt{f}`$=8$`\times `$10<sup>-5</sup> M yr<sup>-1</sup>. Adopting a volume filling factor of $`f`$=0.1 indicates log $`\dot{M}/(M_{}`$ yr$`{}_{}{}^{1})=`$4.6 and a wind performance number of $``$8. Use of C iv $`\lambda `$5471/He ii $`\lambda `$5412 revealed C/He=0.25$`\pm `$0.05 by number. An oxygen abundance of O/He=0.03$`\pm `$0.01 by number was obtained by matching O iv $`\lambda \lambda `$2916–26, $`\lambda \lambda `$3063–72, and $`\lambda `$3560–63.
### 5.2 HD 192103 (WR 135)
A substantial observational dataset was available for the WR135 analysis, particularly high quality UV (HIRES IUE), optical (INT), near-IR (INT, UKIRT) and mid-IR (ISO) spectrophotometry. An interstellar reddening of $`E_{\mathrm{B}\mathrm{V}}`$=0.37 was obtained for WR135, in accord with Smith et al. (1990) (see Table 3), but 0.18 mag lower than the more recent determination of Morris et al. (1993). Membership of Cyg OB3 implied an absolute visual magnitude of $`M_v=`$4.3 mag.
Dereddened spectroscopy of WR135 (solid lines) is compared to our final synthetic model (dotted lines) in Figure 3. Agreement is overall excellent, even for the forest of iron lines in the UV. The relatively low wind velocity of WR135 permits a greater number of individual diagnostics to be selected, including He ii $`\lambda `$4686, He i $`\lambda `$5876, so that the derived properties can be treated with confidence. Indeed, C iii $`\lambda `$5696 and C iv $`\lambda `$5804 are fairly well matched in this case. The strong C iii spectral features at $`\lambda `$2297 and $`\lambda `$9710 are predicted to be 20–30% too strong. Other notable model deficiencies include underestimating the strength of C ii emission at $`\lambda `$4267, $`\lambda `$9900, and O vi $`\lambda \lambda `$3811–34, with O iii $`\lambda `$3130 too strong. As with WR90, the use of absolute fluxes, appears to give poor results in the line strengths around He ii $`\lambda `$2530 since the de-reddened and theoretical continuum do not exactly match. This discrepancy simply reflects the inadequacy of the adopted reddening law, rather than a fundamental flaw with current models.
We obtain the following stellar parameters: T=63 kK, log($`L/L_{}`$)=5.2 and $`\dot{M}/\sqrt{f}`$=3.8$`\times `$10<sup>-5</sup>M yr<sup>-1</sup>. Adopting a volume filling factor of $`f`$=0.1 indicates log $`\dot{M}/(M_{}`$ yr$`{}_{}{}^{1})=`$4.9 and a wind performance number of $``$8. Use of C iv $`\lambda `$5471/He ii $`\lambda `$5412 revealed C/He=0.13$`\pm `$0.03 by number. Once again, the oxygen abundance is poorly constrained, although O/He=0.03$`\pm `$0.01 provides a reasonable match to O iv $`\lambda \lambda `$2916–26 and $`\lambda \lambda `$3063–72, with O iii $`\lambda `$3127 too strong.
In Fig. 2, we present the wind structure of the final WR135 synthetic model, including line formation regions, ionization balance etc. Similar relations are shown for WR111 (WC5) in Hillier & Miller (1999).
### 5.3 WR 146
Our principal observational datasets for WR146 are identical to those used in Paper I, namely INT (optical), UKIRT (near-IR) and ISO (2.6-5$`\mu `$m because of the low S/N at longer wavelengths). The difference in our approach is to derive stellar and chemical properties solely from spectral synthesis, rather than recombination theory and independent modelling of the continuum. As discussed in Sect. 3.2, we follow the distance estimate of 1.4kpc from Dougherty et al. (2000), in the light of new observations from Niemela et al. (1998).
We have included the spectral energy distribution of a O8 supergiant in our synthesis, using that for HD 151804 (O8 If) derived by Crowther & Bohannan (1997). The IR free-free excess of this model is somewhat greater than equivalent temperature Kurucz (1991) models for which the lowest gravity available is log $`g`$=4. Because of the contamination from the late O supergiant, the line spectrum of the WC component of WR146 is relatively weak, as shown in Fig. 4. The flux level of the O companion, the contribution of which declines with increasing wavelength, is illustrated as a dashed line in Fig. 4. Comparison between de-reddened observations and our synthetic spectrum is excellent, with the notable exception of C iv $`\lambda `$5804.
Our final stellar parameters for the WC5 star were, T=$`57kK`$, log $`L/L_{}=5.7`$, and $`\dot{M}/\sqrt{f}`$=1$`\times `$10<sup>-4</sup>M yr<sup>-1</sup>. A filling factor of $`f`$=0.1 reproduced the red wing of C iii-iv $`\lambda `$4650–He ii $`\lambda `$4686. The high wind velocity of WR146 meant that C iv $`\lambda `$5471/He ii $`\lambda `$5412 were blended. Nevertheless, we were able to constrain the carbon content, deriving C/He=0.08$`\pm `$0.02 by number, a factor of two times lower than that estimated in Paper I from recombination line analysis, but now in accord with Eenens & Williams (1992). Oxygen is extremely difficult to measure in WR146 since the usual near-UV diagnostics are unavailable, so that O/He=0.02$`\pm `$0.01 is adopted.
Finally, should the absolute magnitude of WR146 be more typical of other WCE stars, namely $`M_v`$=$``$3.7 mag (Smith et al. 1990), what would be the effect on the derived stellar parameters? In this case, our reddening towards WR146 would imply a distance of 660 pc, such that the companion star would be extremely faint $`M_v`$=$``$3.5 mag, typical of an early B dwarf. The stellar properties of WR146 would be unchanged, except that log $`L/L_{}=5.0`$, and $`\dot{M}/\sqrt{f}`$=3.3$`\times `$10<sup>-5</sup>M yr<sup>-1</sup>. In this case it would be difficult to reconcile these properties with the HST/radio observations of WR146 (Dougherty et al. 1996; Niemela et al. 1998).
## 6 Summary of spectroscopic results
In this section we discuss the results of our quantitative analyses, and make comparisons with previous studies.
### 6.1 Stellar parameters of WC stars
In Table 5, we provide a summary of the derived properties of our programme WC stars, including results for WR11 and WR111 from De Marco et al. (2000) and Hillier & Miller (1999), obtained using identical techniques. Despite our sample spanning WC5 to WC8, there is no obvious trend between spectral type and stellar temperature. Indeed, the WC5 component of WR146 is found to have the lowest stellar temperature, although this is certainly a most unusual WCE star! The subtle behaviour of C iii 5696Å and C iv 5808Å from our modelling indicates that, unfortunately, we are unable to use these as probes of wind ionization for WC5–8 stars (while we cannot obtain a simultaneous fit to both lines, a relatively small change in parameters can lead to a fit of either line).
Similarly, although there is considerable spread in carbon abundances, WR146 exhibits the lowest carbon mass fraction, in contrast to the predicted increase in C/He ratio at earlier spectral type. Koesterke & Hamann (1995) derived a broad range of C/He ratios at each spectral type for WC5–8 stars.
How do the present line blanketed, clumped results compare with previous studies? Unfortunately, WR135 is the sole programme star that has been the subject of quantitative studies in the past. Eenens & Williams (1992) derived elemental abundances from IR recombination lines, estimating C/He=0.12 by number, in excellent agreement with our determination of 0.13. Howarth & Schmutz (1992) used a pure helium non-LTE model analysis to investigate WR135 based solely on 1$`\mu `$m spectroscopy. Since pure helium models are expected to be inadequate for WC analyses, Koesterke & Hamann (1995) considered both carbon and helium in their study of WC stars, including WR135.
In Table 6 we compare results from our present analysis of WR135 with these previous studies, scaling their results to our absolute visual magnitude. Wind velocities and $`\dot{M}/\sqrt{f}`$ are in relatively good agreement, while the effect of including carbon and blanketing has a dramatic effect on the stellar temperatures and luminosities, such that the luminosity derived here is a factor of two times higher than Howarth & Schmutz (1992), who adopted a stellar temperature of 35,000K.
Stellar parameters are in much better agreement with Koesterke & Hamann (1995). They compared specific line strengths with model grids at fixed C/He ratio, and chose simple model atoms of He i-ii and C ii-iv. Although spectral comparisons between model predictions and observations were not presented by Koesterke & Hamann (1995), the fit quality was judged to be poor, with large discrepancies for the C iii $`\lambda `$5696 and $`\lambda `$6740 lines. Therefore, we have greater confidence in our results since detailed UV, optical and IR synthetic spectrophotometry compare favourably with observations.
From Table 6, blanketing and clumping conspire to revise the wind performance number, $`\dot{M}/v_{\mathrm{}}/(L/c)`$, from 30 in the study of Koesterke & Hamann (1995) to just 5 for WR135! For our entire sample, performance numbers are $``$10, with previous studies indicating values of up to 100 (Howarth & Schmutz 1992; Koesterke & Hamann 1995).
### 6.2 Evolutionary status
The wide range in stellar luminosity of our sample of WC stars, $`L/L_{}`$=10<sup>5</sup> to 10<sup>5.7</sup>, implies a considerable range in current stellar masses. Following the mass-luminosity relation for hydrogen-free WR stars of Schaerer & Maeder (1992), present masses of 9$`M_{}`$ (WR135), 13$`M_{}`$ (WR90) and 19$`M_{}`$ (WR146) are implied. In contrast, initial mass estimates are much more dependent on specific evolutionary tracks.
All models predict stars to pass through the WC phase at ages of between 2.7–4.5Myr. For initial 40 and 60$`M_{}`$ models, stellar masses during the early WC phase are predicted to be $``$14$`M_{}`$ and $``$24$`M_{}`$, respectively. At higher initial mass, the situation is extremely model dependent. For example, an evolutionary model with initial mass 120$`M_{}`$ will produce a 60$`M_{}`$ WC star using standard de Jager et al. (1988) mass-loss rates. Alternatively, a 10$`M_{}`$ WC star will result based on (2$`\times `$) enhanced mass-loss for the post main-sequence and WNL phases (Schaller et al. 1992), or only a 4$`M_{}`$ WC star for (2$`\times `$) enhanced mass-loss during the entire stellar evolution (Meynet et al. 1994). Therefore, the present masses and ages of our programme WC stars are well constrained, but initial masses are not.
In order to attempt estimates of initial masses, we compare our derived (C+O)/He ratios versus stellar luminosities with evolutionary predictions in Fig. 5. Predictions for 40–120$`M_{}`$ initial models assume (2$`\times `$) enhanced mass-loss relative to de Jager et al. (1988) during the entire stellar evolution (Meynet et al. 1994). These models favour 60$`M_{}`$ for WR146 and WR90 and 40$`M_{}`$ or 85$`M_{}`$ for WR135. Consequently, multiple initial mass estimates may result for individual stars. Nevertheless, we currently favour 40$`M_{}`$ for WR135 and 60$`M_{}`$ for WR146 and WR90.
Specific comparisons between the stellar properties of WR90, WR135 and WR146 and Meynet et al. (1994) evolutionary predictions are made in Table 7. Both the initial 60$`M_{}`$ and 85$`M_{}`$ models adopt mass-loss rates that exceed observations by large factors during early WC stages. This is because evolutionary models for hydrogen-free WR stars adopt mass-loss rates that are solely functions of mass (Langer 1989). Clearly, future evolutionary models should take allowance for more appropriate WR mass-loss rates.
### 6.3 Radio fluxes
We now compare predicted radio fluxes from our spectroscopic mass-loss rates. Hillier & Miller (1999) provide a formulation based on Eqn 9 in Wright & Barlow (1975), allowing for the filling factor according to Abbott et al. (1981).
In all cases, the dominant ionization at the outer boundary of our models, $`N_e10^8`$ cm<sup>-3</sup> or 200$`R_{}`$ is He<sup>+</sup>, C<sup>2+</sup> and O<sup>2+</sup> (see Fig. 2). Since the radio emitting region lies at lower densities, we also consider cases in which C<sup>+</sup> and/or O<sup>+</sup> are dominant. Outer wind electron temperatures are typically 8000K (Fig. 2), except for WR146 for which the lower cooling produced by less metals implies 9000K.
We compare predicted radio fluxes with observed values taken from the literature in Table 8, in which a uniform UV/optical filling factor of $`f`$=0.1 has been adopted throughout. Note that the quoted radio flux for WR146 refers solely to the WC component (the southern mean emitted flux S<sub>5</sub> in Dougherty et al. 2000). We find that consistency is excellent for WR135 and WR146 for doubly ionized carbon and oxygen, while singly ionized carbon and oxygen are favoured for WR90. However, C<sup>+</sup> is not predicted to be the dominant ionization stage in the outer wind of these stars (Fig. 2), so that the predicted radio flux for this star appears to be too high.
It is possible that the filling factor in the radio and optical forming regions differs significantly. The predicted radio flux of WR90 would be in excellent agreement with the observed value if the radio filling factor was $``$0.2, while that of WR11 requires $``$0.05. Hillier & Miller (1999) found a similar discrepancy in their study of WR111. Note also that there is observational evidence that WR90 may not be single, since it is a non–thermal emitter (Leitherer et al 1997; Chapman et al 1999).
We also include calculations for WR11 in Table 8, since we attempt to re-derive its neon abundance determination in Section 7. As for the other stars in our sample, possible variations in filling factor between the inner and outer wind are important.
## 7 Neon and sulphur abundances in WC stars
We are now in a position to determine neon abundances in our programme WC stars based on ISO spectroscopy, supplemented by sulphur determinations for WR11. In this section we first provide a revised formulation for the determination of neon in a clumped medium, following BRA88, and subsequently provide measurements for each star.
### 7.1 Ionic abundances from fine-structure lines in an inhomogeneous wind
We have re-derived the numerical and analytical forms for the determination of ionic abundances in clumped winds from BRA88. We consider a fine-structure line from ion $`i`$, with transition energy $`h\nu _{ul}`$. If $`D`$ is the distance to the star and $`I_{ul}`$ is the observed line flux, then
$$4\pi D^2I_{ul}=_0^{\mathrm{}}n_uA_{ul}h\nu _{ul}4\pi r^2f𝑑r\mathrm{erg}\mathrm{s}^1$$
(1)
where $`A_{ul}`$ is the line transition probability and we have introduced the filling factor $`f`$ into their formulation. $`n_u`$ represents the density of ions in the upper level, and can be written as,
$$n_u=f_un_i\mathrm{cm}^3$$
(2)
where $`n_i`$ is the species ion density and $`f_u`$ is the fractional population of the upper level. Upper level populations, $`f_u`$, were determined for each ion by solving the equations of statistical equilibrium using equib (Adams & Howarth, priv. comm) for $``$30 electron densities in the range 10<sup>0</sup> to 10<sup>12</sup> cm<sup>-3</sup>, and 9 electron temperatures covering 5k to 14kK. Thus,
$$n_u=\frac{f_u\gamma _iA}{fr^2}\mathrm{cm}^3$$
(3)
where $`\gamma _i`$ is the fraction of all ions represented by ion species $`i`$
$$\gamma _i=\frac{n_i}{_jn_j}$$
(4)
and $`A`$ is the mass loss parameter (Eqn 8 from BRA88). Combining equations 13, the filling factor term cancels out, leaving
$$I_{ul}=\frac{\gamma _i}{D^2}A_{ul}h\nu _{ul}A_0^{\mathrm{}}f_u(r,f,T)𝑑r\mathrm{erg}\mathrm{cm}^2\mathrm{s}^1$$
(5)
We deviate from BRA88 by carrying out the integral in density, rather than radial space. Since $`r(N_e`$) is a bijection, we can modify this integral so that the dependency of $`f_u`$ on $`A`$ and $`f`$ (and hence $`\dot{M}`$) is removed from the integral term. Equation 5 can then be modified to:
$$_0^{\mathrm{}}f_u(r,f,T)𝑑r=\sqrt{\frac{\gamma _eA}{4f}}_0^{\mathrm{}}\frac{f_u(N_e,T)}{N_e^{1.5}}𝑑N_e$$
(6)
For a more straightforward numerical computation, Equation 6 is slightly adjusted:
$$_0^{\mathrm{}}\frac{f_u(N_e,T)}{N_e^{1.5}}𝑑N_e=ln(10)_0^{\mathrm{}}\frac{f_u(N_e,T)}{\sqrt{N_e}}𝑑log(N_e)$$
(7)
The effect of this modification leads to a significant improvement over BRA88, who used an average value for each region $`\mathrm{\Delta }`$r. since the line formation region is a very sensitive function of f<sub>u</sub>. Also, the dependence of the integral on $`\dot{\mathrm{M}}`$ is removed. Consequently, BRA88 overestimated the integral term, producing a lower elemental abundance. Our final numerical expression for the ion number fraction $`\gamma _i`$, is (cgs units):
$$\gamma _i=\frac{(4\pi \mu m_Hv_{\mathrm{}})^{1.5}}{ln(10)f^{0.25}}\left(\frac{\sqrt{f}}{\dot{\mathrm{M}}}\right)^{1.5}\frac{1}{F_u(T)}\frac{2D^2I_{ul}}{\sqrt{\gamma _e}A_{ul}h\nu _{ul}}$$
(8)
with
$$F_u(T)=_0^{\mathrm{}}\frac{f_u(N_e,T)}{\sqrt{N_e}}𝑑log(N_e).$$
(9)
Provided radio and recombination processes are used to derive $`\dot{\mathrm{M}}`$/$`\sqrt{f}`$, elemental abundances are only weakly dependent on the distance ($``$ D<sup>-0.25</sup>) since mass-loss rates depend on $`D^{1.5}`$. Therefore elemental abundances obtained with filling factors of $`f`$=0.1 or 1.0 differ by a factor of 1.8.
In order to allow for the possibility of a clumped medium, the analytical expression for the ion number fraction given by BRA88 (their equation A13) also needs to be multiplied by a factor of $`\sqrt{f}`$. We include determinations of $`\gamma _i`$ using both the integral and analytical expressions in our subsequent calculations. We find that the analytic expression is reliable to within $``$20%.
### 7.2 Neon abundances for WC stars
We are now in a position to evaluate neon abundances in our programme WC stars, using Equation 8. Figure 6 shows ISO/SWS observations of \[Ne iii\] 15.55$`\mu `$m emission for each of our programme stars, together with \[Ne ii\] 12.8$`\mu `$m in WR11. Emission line fluxes are listed in Table 9.
Table 9 also includes atomic quantities used to derive ion fractions. Transition probabilities are taken from the NIST Atomic Spectra Database, while collision strengths are obtained from papers in the IRON Project series (see Table 9 for references). ISO flux measurements for WR11 compare closely with previous measurements from IRAS and ground-based observations (van der Hucht & Olnon (1985); BRA88), as indicated in Table 9.
In the following comparison, we include abundances derived from ISO mid-IR neon lines for WR11, using stellar parameters derived by De Marco et al. (2000).
Neon abundance determinations are sensitive to the ionization balance in the line forming region for the fine-structure lines, and is comparable to that of the radio forming continuum, $`N_e`$$``$10<sup>5</sup> cm<sup>-3</sup>. Ne<sup>+</sup>, with an ionization potential that is higher than C<sup>+</sup> and O<sup>+</sup>, is observed solely in WR11 (WC8). Therefore, we have allowed for the possibility that the carbon and oxygen ionization balance are singly ionized for WC8 stars, with He<sup>+</sup> and doubly ionized carbon and oxygen otherwise. Table 10 shows predicted ion abundances of Ne<sup>2+</sup> for each case, plus measured Ne<sup>+</sup> for WR11, with upper limits otherwise. In all cases Ne/He$``$3–4$`\times `$10<sup>-3</sup> by number, significantly greater than the expected cosmic value of Ne/He$``$5$`\times 10^4`$ in the C and O-enriched WC environment.
It is possible that neon exists in (unseen) higher ionization stages, specifically Ne<sup>3+</sup>. However, the ionization potential for Ne<sup>3+</sup>, 97eV, is significantly higher than C<sup>3+</sup> (64eV) and O<sup>3+</sup> (77eV) which are not expected to be present in the outer winds of WC stars (Fig. 2).
In summary, measurement of fine-structure lines of neon from ISO/SWS observations reveal Ne/He=0.003–0.004 (Ne/C$``$0.02), a factor of 6–8 times higher than cosmic abundances of Ne/He=0.0005 for the carbon rich winds of WC stars, supported also by sulphur abundance determinations. However, what additional sources of uncertainty are there in our Ne/He determinations, and why do our results for WR11 differ from BRA88, who obtained Ne/He=0.001?
Although the ISO-SWS neon line fluxes are in good agreement with the ground-based and IRAS neon line fluxes from BRA88 (Morris et al. 1998), we derive an elemental abundance that is three times higher. The source of this discrepancy is due to a different distance to WR11, and the use of a non-clumped mass-loss rate by BRA88. Morris et al. (2000) have recently emphasised the need for a reliable mass-loss rate estimate in the determination of neon abundances. In addition, our derived neon abundance is in good agreement with Morris et al. (1998) who combined ISO/SWS neon fluxes with the hipparcos distance to WR11, and the X-ray derived mass-loss rate of Stevens et al. (1996).
The other major factor affecting neon abundances is clumping. Our UV/optical analyses use filling factors of $`f`$=0.1, yet we cannot observationally constrain the filling factor to better than 0.05$`f`$0.25, indicating a further 20% uncertainty in Ne/He. Of greater importance, we assume identical filling factors for the optical line forming region ($``$10<sup>11</sup> cm<sup>-3</sup>) and the neon emitting region ($``$10<sup>5</sup> cm<sup>-3</sup>). Neon abundances would be increased by 40$`\%`$ for WR90, decreased by 30$`\%`$ for WR11, with WR146 and WR135 unchanged, assuming that (i) UV/optical mass-loss rates are fully consistent with radio fluxes (recall Sect. 6.3), and that (ii) volume filling factors in the neon emitting region are identical to those in the radio region.
### 7.3 Sulphur abundances for WR11
In order to assess the reliability of our derived abundances, we have also calculated the sulphur abundance for WR11 based on ISO observations of fine-structure \[S iv\] 10.5$`\mu `$m and \[S iii\] 18.7$`\mu `$m lines (Table 9). Since sulphur is not enhanced by nucleosynthesis, abundances should correspond to the cosmic value. The line forming region for the sulphur fine-structure lines peaks at $`N_e`$$``$10<sup>4</sup> cm<sup>-3</sup>, somewhat lower than the neon lines.
ISO spectroscopy of \[S iv\] 10.5$`\mu `$m confirms the line flux measured from ground-based spectroscopy by BRA88, while stellar modelling anticipates weak contamination from lines of He i (12–8) 10.52$`\mu `$m and C iii (20–17) 10.54$`\mu `$m. The stellar analysis of WR11 by De Marco et al. (2000) predicts a 15% contribution from these lines to the observed flux, which has been corrected accordingly. For the first time in a Wolf-Rayet star, ISO reveals the presence of \[S iii\] 18.7$`\mu `$m, blended with the stellar He i (14–10) 18.62$`\mu `$m feature, resulting in a 30% decrease in \[S iii\] flux. Use of an inappropriate mass-loss rate and distance for WR11 led BRA88 to suggest S<sup>3+</sup>/He=2.5$`\times 10^5`$, such that they were obliged to predict S$`{}_{}{}^{2+}>`$S<sup>3+</sup>, with an expected high \[S iii\] 18.7$`\mu `$m line flux that is not confirmed by ISO observations.
From Table 10, we find $`\gamma _{\mathrm{S}^2+}=1.9\times 10^5`$ and $`\gamma _{\mathrm{S}^3+}=3.1\times 10^5`$, which imply S/He=6$`\times 10^5`$ by number. This is in good agreement with the cosmic value of 7.5$`\times 10^5`$ for the C and He enriched environment of WR11. Therefore, our determinations imply Ne/S=50 for WR11, a factor of eight times greater than the cosmic value of Ne/S$``$7. Since these lines are formed in similar regions of the stellar wind, Ne/S abundances are essentially independent of clumping, and reveal a degree of neon enrichment relative to sulphur that is equivalent to that derived earlier for He.
### 7.4 Comparison of abundances with theoretical predictions
Regarding theoretical expectations, the level of neon enrichment is expected to be strongly correlated with the carbon content (Schaller et al. 1992; Meynet et al. 1994). In Fig.7 we compare observed Ne/He and Ne/C versus C/He ratios with theoretical expectations from the Schaller et al. (1992) and Meynet et al. (1994) evolutionary models at solar metallicity for an initial mass of 60$`M_{}`$. We find that observed neon abundances are a factor of two below expectations. Specifically, we derive Ne/C=0.02$`\pm `$0.01 in all cases, while Ne/C$``$0.03 is predicted during this phase of the WC evolution. Error bars shown in the figure account for uncertainties in distance and (uniform) filling factors, but do not allow for the possibility of a varying volume filling factor between the neon and UV/optical line forming regions, which could be responsible for the discrepant cases.
## 8 Summary
We have performed quantitative analyses of a small sample of WC5–8 stars, using models that account for line blanketing and clumping. Comparisons between synthetic spectra and de-reddened UV to mid-IR observations are excellent, with few modelling deficiencies identified. Stellar parameters support previous determinations (e.g. Koesterke & Hamann 1995), except that the incorporation of blanketing yields higher stellar luminosities, while clumping indicates lower wind performance numbers, supporting the conclusions of Hillier & Miller (1999) for WR111 (WC5). Future studies will derive properties of WC-type stars, at both earlier (WO) and later (WC9) spectral type, and investigate whether predicted ionizing properties are consistent with nebular observations (e.g. Crowther et al. 1999).
ISO/SWS spectroscopy reveals the presence of neon fine structure transitions, allowing abundance determinations. Using a revised formulation of the BRA88 technique to account for wind clumping, we derive neon abundances of Ne/He=3–4$`\times `$10<sup>-3</sup> by number, seven times higher than the cosmic value adjusted for the H-depleted WC environment, supported by Ne/S=50 for WR11 from sulphur fine structure lines. The Ne enrichment is a factor of $``$2 times lower than predictions of current theoretical models. However, differences in volume filling factors between the (high density) UV/ optical line formation regions and (low density) mid-IR fine-structure forming regions represent the greatest source of uncertainty in current Ne/He abundance determinations. Nevertheless, Ne/S provides an independent confirmation of the neon enrichment since it is independent of outer wind filling factors. Future large ground and space-based telescopes that are optimised for the IR, will allow neon and sulphur line flux measurements and abundance determinations for more distant Wolf-Rayet stars. Of particularly interest are abundances in carbon and oxygen-rich WO stars.
## Acknowledgments
This work is based on observations with ISO, an ESA project with instruments funded by ESA Member States (especially the PI countries: France, Germany, the Netherlands and the United Kingdom) with the participation of ISAS and NASA. Theoretical predictions presented here were possible only as a result of the Opacity Project, led by Prof Michael Seaton. Thanks to Dr Tim Harries for observing WR90 on our behalf at the AAT. LD would like to acknowledge financial support from the UCL Perren Fund. PAC acknowledges financial support from a Royal Society University Research Fellowship. DJH gratefully acknowledges support by NASA through grant number NAG5–8211. The Anglo-Australian Telescope, Isaac Newton Telescope and U.K. Infrared Telescope are operated by the Anglo-Australian Observatory, Isaac Newton Group and Joint Astronomy Centre, respectively, on behalf of the Particle Physics and Astronomy Research Council. |
warning/0001/hep-ph0001314.html | ar5iv | text | # CP violation in 𝜓(2𝑆)→𝐽/𝜓𝜋𝜋 processes
## Abstract
We propose to search for CP-violating effects in the decay $`\psi (2S)J/\psi \pi \pi `$. The scheme has the advantage that one does not need to track two or more CP-conjugate processes. Model independent amplitudes are derived for this purpose. The fact that leading CP violating terms are $`𝒪(k)`$ under low energy expansion and the processes are flavor disconnected make the measurement of these CP breaking parameters practical. Our results can be extended to the case of $`\mathrm{{\rm Y}}(2S)\mathrm{{\rm Y}}(1S)\pi \pi `$ and $`\mathrm{{\rm Y}}(3S)\mathrm{{\rm Y}}(2S)\pi \pi `$ straightforwardly.
PACS numbers: 11.30.Er, 13.25.Gv, 11.80.Cr
CP violation is a subject attracting much interest. In the Standard Model, it arises as a phase entering the CKM matrix. It is believed that, with the CP violation presented in the Standard Model, it is not possible to generate the observed size of matter-antimatter asymmetry of the Universe . However, CP violation has only been observed in neutral-kaon systems till now. The evidence come from the measurements of $`\eta _+`$, $`\eta _{00}`$, and the semileptonic decay charge asymmetry for $`K_L`$. Currently experimental efforts are concentrated on neutral systems such as $`K^0`$-$`\overline{K}^0`$, $`B^0`$-$`\overline{B}^0`$, and $`D^0`$-$`\overline{D}^0`$ complex. Other searches, as summarized by Wolfenstein and Trippe, devide into two classes: (a) Those that involve weak interactions or parity violation. The most sensitive are the searches for an electric dipole moment of the elementary particles such as neutron and electron. (b) Those that involve processes otherwise allowed by the strong or electromagnetic interactions. This includes the search for C violation in $`\eta `$ decay and searches for T violation in a number of nuclear and electromagnetic reactions .
In this paper we suggest a new means to search for CP violation. We propose to measure CP asymmetries in the decay $`\psi (2S)J/\psi \pi ^+\pi ^{}`$ or $`\psi (2S)J/\psi \pi ^0\pi ^0`$. The process $`\psi (2S)J/\psi \pi ^+\pi ^{}`$ has been discussed based on an effective lagrangian , the multipole expansion hypothesis , or current algebra . These discussions are model dependent, and C, P, T invariance are presumed. We derive model independent amplitudes of the two processes. CP invariance requires particular parameters in the decay amplitude vanish. Any non-zero value of those particular parameters implies CP violation. One will not need to track two or more channels at the same time. And we need not to compare the phases or decay rate of two or more CP-conjugate processes to get CP violation observables, thus avoid precision loss when we subtracting two very close numbers or information loss because of using total transition rates instead of differential cross sections.
$`J/\psi \pi ^+\pi ^{}`$ and $`J/\psi \pi ^0\pi ^0`$ are the two largest decay modes of $`\psi (2S)`$, with branching ratios $`(30.2\pm 1.9)\%`$ and $`(17.9\pm 1.8)\%`$ . The masses for $`\psi (2S)`$ and $`J/\psi `$ are $`3686.00\pm 0.09`$ MeV and $`3096.88\pm 0.04`$ MeV, with a difference of 589 MeV. The energy available for the pions are small and as we will shown, most part of the amplitude are expected to be the contribution of contact interactions. This validates the low energy expansion. The leading CP-violating terms in the model independent amplitudes are of the first order of the soft pion momentum($`𝒪(k)`$). BES has enough data to see D-wave contributions , which is $`𝒪(k^2)`$. Because the processes are flavor disconnected, the strong interactions are suppressed, and we have chances to see CP-violating effects beyond (or even within) the Standard Model in these decays. So it is practical to measure these CP-violating parameters. It will not take long for BEPC to accumulate $`10^8`$ $`\psi (2S)`$ events. Now BEPC is scheduled for another upgrade in the near future. After the upgrade it will have the ability to take more than $`10^9`$ $`\psi (2S)`$ events and give a high precision test of CP invariance in $`\psi (2S)J/\psi \pi \pi `$ decays.
The Feynman amplitude for the process $`\psi (2S)J/\psi \pi ^+\psi ^{}`$ reads
$$_{\lambda ,\sigma }(p,q,k_1,k_2)=e_\psi ^{}^\mu (\stackrel{}{p},\lambda )e_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{\mu \nu }.$$
(1)
Here $`p,q,k_1,k_2`$ are four-momenta of the particles $`\psi (2S),J/\psi ,\pi ^+,\pi ^{}`$. The helicities of $`\psi (2S)`$ and $`J/\psi `$ are $`\lambda `$ and $`\sigma `$, while $`e_\psi ^{}^\mu (\stackrel{}{p},\lambda )`$ and $`e_{J/\psi }^\nu (\stackrel{}{q},\sigma )`$ are the corresponding polarization vectors for the two particles. $`\mathrm{\Gamma }_{\mu \nu }`$ is the effective vertex for the process, which contains all details of the decay.
The effective vertex $`\mathrm{\Gamma }^{\mu \nu }`$ should be constructed from the four-momenta $`p^\alpha `$, $`q^\alpha `$, $`k_1^\alpha `$, $`k_2^\alpha `$ and isotropic tensors $`g^{\alpha \beta }`$, $`\epsilon ^{\alpha \beta \gamma \delta }`$. The effective vertex contains only the zeroth and first order of the antisymmetric tensor because of the identity
$$\epsilon _{\alpha \beta \gamma \delta }\epsilon _{\alpha ^{^{}}\beta ^{^{}}\gamma ^{^{}}\delta ^{^{}}}=det\left(\begin{array}{cccc}g_{\alpha \alpha ^{^{}}}& g_{\beta \alpha ^{^{}}}& g_{\gamma \alpha ^{^{}}}& g_{\delta \alpha ^{^{}}}\\ g_{\alpha \beta ^{^{}}}& g_{\beta \beta ^{^{}}}& g_{\gamma \beta ^{^{}}}& g_{\delta \beta ^{^{}}}\\ g_{\alpha \gamma ^{^{}}}& g_{\beta \gamma ^{^{}}}& g_{\gamma \gamma ^{^{}}}& g_{\delta \gamma ^{^{}}}\\ g_{\alpha \delta ^{^{}}}& g_{\beta \delta ^{^{}}}& g_{\gamma \delta ^{^{}}}& g_{\delta \delta ^{^{}}}\end{array}\right).$$
(2)
Considering energy-momentum conservation
$`p^\alpha =q^\alpha +k_1^\alpha +k_2^\alpha `$ (3)
and the Lorentz conditions for polarization vectors
$`p_\alpha e_\psi ^{}^\alpha (\stackrel{}{p},\lambda )`$ $`=`$ $`0,`$ (4)
$`q_\alpha e_{J/\psi }^\alpha (\stackrel{}{q},\sigma )`$ $`=`$ $`0,`$ (5)
we find the general form of the effective vertex
$`\mathrm{\Gamma }^{\mu \nu }(p,q,k_1,k_2,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (6)
$`=`$ $`c_1g^{\mu \nu }+c_2k_1^\mu k_1^\nu +c_3k_2^\mu k_2^\nu +c_4k_1^\mu k_2^\nu +c_5k_2^\mu k_1^\nu +c_6A_1^{\mu \nu }+c_7A_2^{\mu \nu }`$ (8)
$`+c_8A_3^{\mu \nu }+c_9Q^\mu k_1^\nu +c_{10}Q^\mu k_2^\nu +c_{11}k_1^\mu Q^\nu +c_{12}k_2^\mu Q^\nu ,`$
where we define
$`A_1^{\alpha \beta }`$ $`=`$ $`p_\gamma k_{1\delta }\epsilon ^{\alpha \beta \gamma \delta },`$ (9)
$`A_2^{\alpha \beta }`$ $`=`$ $`p_\gamma k_{2\delta }\epsilon ^{\alpha \beta \gamma \delta },`$ (10)
$`A_3^{\alpha \beta }`$ $`=`$ $`k_{1\gamma }k_{2\delta }\epsilon ^{\alpha \beta \gamma \delta },`$ (11)
$`Q^\alpha `$ $`=`$ $`p_\beta k_{1\gamma }k_{2\delta }\epsilon ^{\alpha \beta \gamma \delta }.`$ (12)
The twelve form factors in $`\mathrm{\Gamma }^{\mu \nu }`$ are not independent since
$`Q_\mu k_{1\nu }e_\psi ^{}^\mu e_{J/\psi }^\nu \left(k_{1\mu }Q_\nu +(k_1k_2)A_{1\mu \nu }k_1^2A_{2\mu \nu }+(pk_1)A_{3\mu \nu }\right)e_\psi ^{}^\mu e_{J/\psi }^\nu ,`$ (13)
$`Q_\mu k_{2\nu }e_\psi ^{}^\mu e_{J/\psi }^\nu \left(k_{2\mu }Q_\nu +k_2^2A_{1\mu \nu }(k_1k_2)A_{2\mu \nu }+(pk_2)A_{3\mu \nu }\right)e_\psi ^{}^\mu e_{J/\psi }^\nu ,`$ (14)
$`\left(Q_\mu k_{1\nu }+Q_\mu k_{2\nu }\right)e_\psi ^{}^\mu e_{J/\psi }^\nu \left((pk_2)A_{1\mu \nu }(pk_1)A_{2\mu \nu }+p^2A_{3\mu \nu }\right)e_\psi ^{}^\mu e_{J/\psi }^\nu .`$ (15)
We can eliminate three terms in the effective vertex without introducing kinematic singularities. After a redefinition of form factors we have nine independent terms,
$`\mathrm{\Gamma }^{\mu \nu }(p,q,k_1,k_2,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (16)
$`=`$ $`c_1g^{\mu \nu }+c_2\left(k_1^\mu k_1^\nu +k_2^\mu k_2^\nu \right)+c_3\left(k_1^\mu k_1^\nu k_2^\mu k_2^\nu \right)+c_4\left(k_1^\mu k_2^\nu +k_2^\mu k_1^\nu \right)`$ (19)
$`+c_5\left(k_1^\mu k_2^\nu k_2^\mu k_1^\nu \right)+c_6\left(A_1^{\mu \nu }+A_2^{\mu \nu }\right)+c_7\left(A_1^{\mu \nu }A_2^{\mu \nu }\right)`$
$`+c_8A_3^{\mu \nu }+c_9Q^\mu \left(k_1^\nu k_2^\nu \right),`$
with
$$c_i=c_i((k_1k_2),[p(k_1k_2)])$$
(20)
for $`i=1,2,\mathrm{},9`$.
Expand the form factors in Taylor series of $`(k_1k_2)`$ and $`[p(k_1k_2)]`$, and keep those terms up to $`𝒪(k^2)`$,
$`\mathrm{\Gamma }^{\mu \nu }(p,q,k_1,k_2,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (21)
$`=`$ $`\{f_0+{\displaystyle \frac{[p(k_1k_2)]}{M_\psi ^{}^2}}f_1+{\displaystyle \frac{(k_1k_2)}{M_\psi ^{}^2}}f_2+{\displaystyle \frac{[p(k_1k_2)]^2}{M_\psi ^{}^4}}f_3\}g^{\mu \nu }`$ (27)
$`+{\displaystyle \frac{(k_1^\mu k_1^\nu +k_2^\mu k_2^\nu )}{M_\psi ^{}^2}}f_4+{\displaystyle \frac{(k_1^\mu k_1^\nu k_2^\mu k_2^\nu )}{M_\psi ^{}^2}}f_5`$
$`+{\displaystyle \frac{(k_1^\mu k_2^\nu +k_2^\mu k_1^\nu )}{M_\psi ^{}^2}}f_6+{\displaystyle \frac{(k_1^\mu k_2^\nu k_2^\mu k_1^\nu )}{M_\psi ^{}^2}}f_7`$
$`+i{\displaystyle \frac{(A_1^{\mu \nu }+A_2^{\mu \nu })}{M_\psi ^{}^2}}f_8+i{\displaystyle \frac{(A_1^{\mu \nu }+A_2^{\mu \nu })[p(k_1k_2)]}{M_\psi ^{}^4}}f_9`$
$`+i{\displaystyle \frac{(A_1^{\mu \nu }A_2^{\mu \nu })}{M_\psi ^{}^2}}f_{10}+i{\displaystyle \frac{(A_1^{\mu \nu }A_2^{\mu \nu })[p(k_1k_2)]}{M_\psi ^{}^4}}f_{11}`$
$`+i{\displaystyle \frac{A_3^{\mu \nu }}{M_\psi ^{}^2}}f_{12},`$
where $`M_\psi ^{}`$ is the mass of $`\psi (2S)`$ and $`f_i`$ are dimensionless complex constants. The reason why the energy scale should be $`M_\psi ^{}`$, and why we extract a factor “$`i`$” from $`f_8,f_9,\mathrm{},f_{12}`$ will be explained later.
Now let’s see what CP invariance can say about the form factors. If CP parity is conserved, then
$`\psi ^{}|𝐒|J/\psi \pi ^+\pi ^{}`$ (28)
$`=`$ $`\psi ^{}|(\mathrm{𝐂𝐏})^{}(\mathrm{𝐂𝐏})𝐒(\mathrm{𝐂𝐏})^\mathrm{𝟏}(\mathrm{𝐂𝐏})|J/\psi \pi ^+\pi ^{}`$ (29)
$`=`$ $`\psi ^{}|(\mathrm{𝐂𝐏})^{}𝐒(\mathrm{𝐂𝐏})|J/\psi \pi ^+\pi ^{}.`$ (30)
The spin-parity $`J^{PC}`$ for $`\psi (2S)`$, $`J/\psi `$ and pions are $`1^{}`$, $`1^{}`$ and $`0^+`$ ,
$`\mathrm{𝐂𝐏}|\stackrel{}{p},\lambda `$ $`=`$ $`|\stackrel{}{p},\lambda ,`$ (31)
$`\mathrm{𝐂𝐏}|\stackrel{}{q},\sigma ,\stackrel{}{k}_1,\stackrel{}{k}_2`$ $`=`$ $`|\stackrel{}{q},\sigma ,\stackrel{}{k}_2,\stackrel{}{k}_1.`$ (32)
So conservation of CP require
$$_{\lambda ,\sigma }(p,q,k_1,k_2)_{\lambda ,\sigma }(\overline{p},\overline{q},\overline{k}_2,\overline{k}_1).$$
(33)
Where a vector with a bar indicate its space reflected value, e.g.,
$$\overline{q}^\alpha 𝒫_\beta ^\alpha q^\beta ,$$
(34)
and $`𝒫`$ is the space reflection matrix, $`(𝒫_\beta ^\alpha )=\mathrm{diag}\{1,1,1,1\}`$. That is
$`e_\psi ^{}^\mu (\stackrel{}{p},\lambda )e_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{\mu \nu }(p,q,k_1,k_2,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (35)
$`=`$ $`e_\psi ^{}^\mu (\stackrel{}{p},\lambda )e_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{\mu \nu }(\overline{p},\overline{q},\overline{k}_2,\overline{k}_1,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (36)
$`=`$ $`\overline{e}_\psi ^{}^\mu (\stackrel{}{p},\lambda )\overline{e}_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{\mu \nu }(\overline{p},\overline{q},\overline{k}_2,\overline{k}_1,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta })`$ (37)
$`=`$ $`e_\psi ^{}^\mu (\stackrel{}{p},\lambda )e_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{\mu \nu }(p,q,k_2,k_1,g^{\alpha \beta },\epsilon ^{\alpha \beta \gamma \delta }).`$ (38)
Since all form factors in Eq. (21) are independent, it is easy to see that any non-vanishing value of $`f_1`$,$`f_5`$,$`f_7`$,$`f_8`$ or $`f_{11}`$ implies CP violation. $`f_1/f_0`$, $`f_5/f_0`$, $`f_7/f_0`$, $`f_8/f_0`$ and $`f_{11}/f_0`$ can be taken as CP breaking parameters. Among them, the $`f_1`$ and $`f_8`$ terms are $`𝒪(k)`$. In fact, the parameters in Eq. (21) can be classified into four types(as shown in Tab. I): (1)$`f_0`$, $`f_2`$, $`f_3`$, $`f_4`$ and $`f_6`$ terms keep both C and P symmetry. (2)$`f_1`$, $`f_5`$ and $`f_7`$ terms are C nonconserving but keep P symmetry. (3)$`f_8`$ and $`f_{11}`$ terms keep C symmetry but break P. (4)$`f_9`$, $`f_{10}`$ and $`f_{12}`$ terms break both C and P, but keep CP invariance.
If we suppose the amplitude is completely the first order contribution of an effective lagrangian, and the lagrangian is time reversal invariant, the form factors $`f_0,f_2,\mathrm{},f_{12}`$ will be relatively real. That’s why we extract a factor ”$`i`$” in the abnormal terms. Non-zero relative phases may come from T violation in the effective lagrangian or high order terms(loops and resonance). However, since there is no strong resonance in the process, we can expect the Feynman amplitude to be dominated by contact interaction terms in the effective lagrangian, so that their relative phases will be small. At least this will be true for those terms keeping both CP and isospin symmetry, i.e., the relative phases between $`f_0,f_2,f_3,f_4,f_6`$ will be very small(but one should not mistake a nonzero relative phase as the signal of CP violation).
The $`\psi (2S)J/\psi \pi ^+\pi ^{}`$ process is flavor disconnected. In an effective theory, the tree level diagram with a spin-0 resonance(e.g. $`f_0`$) of $`\pi \pi `$ is illustrated in Fig. 2. The gluon-scalar vertex $`\mathrm{\Gamma }_2^\beta `$ has to be proportional to $`(k_1^\beta +k_2^\beta )`$. When contracted with the gluon propagator and the $`\psi (2S)J/\psi `$-gluon vertex $`\mathrm{\Gamma }_2^{\mu \nu \alpha }`$, gauge invariance will ensure the contribution vanish. For spin-$`n`$($`n2`$) meson, the gluon-meson vertex contracted with the spin-$`n`$ meson propagator will contain a factor $`(k_1+k_2)^2m^2`$, which gives a zero on the mass shell instead of a pole. So we come to the conclusion:there is no $`\pi `$-$`\pi `$ resonance other than those with spin-1.
Because the isospins of $`\psi (2S)`$ and $`J/\psi `$ are all zero, isospin symmetry will require the two pions to form an isospin singlet, which is symmetric under the interchange of the two pions’ four-momenta. Since the amplitude for an odd spin meson’s decay into two pions will be anti-symmetric when interchanging the pions’ momenta, we see that any spin-1 resonance here is suppressed by isospin symmetry.
Nor can we find any strong resonance in $`J/\psi \pi ^\pm `$ channels. $`\psi (2S)`$ decays mainly through contact interactions. Therefor, $`M_\psi ^{}`$(or heavy quark mass) can be taken as the typical energy scale of the process.
Not all terms in Eq. (21) should be kept when fitting data. CP violating $`𝒪(k^2)`$ terms, $`f_5`$, $`f_7`$ and $`f_{11}`$, are strongly suppressed and we can not to see them when we have not enough data. $`f_9`$ and $`f_{12}`$ terms should also be dropped because they are $`𝒪(k^2)`$ and break isospin symmetry. Eight terms are kept:
$`\mathrm{\Gamma }^{\mu \nu }`$ $`=`$ $`\left\{f_0+{\displaystyle \frac{[p(k_1k_2)]}{M_\psi ^{}^2}}f_1+{\displaystyle \frac{(k_1k_2)}{M_\psi ^{}^2}}f_2+{\displaystyle \frac{[p(k_1k_2)]^2}{M_\psi ^{}^4}}f_3\right\}g^{\mu \nu }`$ (40)
$`+{\displaystyle \frac{(k_1^\mu k_1^\nu +k_2^\mu k_2^\nu )}{M_\psi ^{}^2}}f_4+{\displaystyle \frac{(k_1^\mu k_2^\nu +k_2^\mu k_1^\nu )}{M_\psi ^{}^2}}f_6+i{\displaystyle \frac{(A_1^{\mu \nu }+A_2^{\mu \nu })}{M_\psi ^{}^2}}f_8+i{\displaystyle \frac{(A_1^{\mu \nu }A_2^{\mu \nu })}{M_\psi ^{}^2}}f_{10}.`$
The $`f_1`$ and $`f_8`$ term in the amplitude are CP-violating.
For the process $`\psi (2S)J/\psi \pi ^0\pi ^0`$, boson symmetry of the two pions demands the amplitude invariant when the four-momenta $`k_1`$ and $`k_2`$ are interchanged. Similar analysis leads to an effective vertex of the form
$`\mathrm{\Gamma }^{\mu \nu }`$ $`=`$ $`\left\{g_0+{\displaystyle \frac{(k_1k_2)}{M_\psi ^{}^2}}g_1+{\displaystyle \frac{[p(k_1k_2)]^2}{M_\psi ^{}^4}}g_2\right\}g^{\mu \nu }+{\displaystyle \frac{(k_1^\mu k_1^\nu +k_2^\mu k_2^\nu )}{M_\psi ^{}^2}}g_3`$ (42)
$`+{\displaystyle \frac{(k_1^\mu k_2^\nu +k_2^\mu k_1^\nu )}{M_\psi ^{}^2}}g_4+i{\displaystyle \frac{(A_1^{\mu \nu }+A_2^{\mu \nu })}{M_\psi ^{}^2}}g_5+i{\displaystyle \frac{(A_1^{\mu \nu }A_2^{\mu \nu })[p(k_1k_2)]}{M_\psi ^{}^4}}g_6,`$
and the Feynman amplitude is
$$_{}^{}{}_{\lambda ,\sigma }{}^{}(p,q,k_1,k_2)=e_\psi ^{}^\mu (\stackrel{}{p},\lambda )e_{J/\psi }^\nu (\stackrel{}{q},\sigma )\mathrm{\Gamma }_{}^{}{}_{\mu \nu }{}^{}.$$
(43)
The $`g_0`$, $`g_1`$, $`g_2`$, $`g_3`$ and $`g_4`$ terms are CP conserving, while $`g_5`$, $`g_6`$ terms are CP violating. Argument for why we take $`M_\psi ^{}`$ as the energy scale is similar to the above, and we only need to point out the fact that any odd spin particle’s decay into two identical spin-0 particles is strictly forbidden by boson symmetry. Relative phases between $`g_0`$, $`g_1`$, $`g_2`$, $`g_3`$ and $`g_4`$ are small. The CP breaking $`𝒪(k^2)`$ term $`g_6`$ can be set to zero when one fits data since it is strongly suppressed.
The cross section of such a three-body decay depends on five variables: $`E_1`$ and $`E_2`$, the energies of the two pions; and $`(\alpha ,\beta ,\gamma )`$, the Euler angles that specify the orientation of the final system relative to the initial particle . The cross section of $`\psi (2S)J/\psi \pi \pi `$ does not depend on the variable $`\alpha `$ provided that $`\psi (2S)`$ is produced by $`e^+e^{}`$ collision, and the $`\alpha `$ represents the rotation angle around the beam line. We have
$`k_1k_2=(E_1+E_2)M_\psi ^{}{\displaystyle \frac{1}{2}}(M_\psi ^{}^2M_{J/\psi }^2+2m_\pi ^2),`$ (44)
$`p(k_1k_2)=(E_1E_2)M_\psi ^{}`$ (45)
and the nine independent terms in Eq. (16) represent different angular distributions. Those terms proportional to $`|f_1|^2`$, $`|f_8|^2`$, $`f_1f_8^{}`$ and $`f_1^{}f_8`$ in the cross section can be ignored since it is relatively small. The coherent part that proportional $`f_8`$ (or $`f_8^{}`$) has a significantly different distribution comparing with backgrounds. The part proportional to $`f_1`$ (or $`f_1^{}`$) is an odd distribution when we exchange $`E_1`$ with $`E_2`$ in the Dalitz plot, while the backgrounds are even distributions. These two facts will help us to distinguish the CP-violating parts from background contributions, and will remarkably improve the precision of the measurement of the CP-violating parameters $`f_1`$ and $`f_8`$. Unlike the case in the decays of $`K^0\overline{K}^0`$ complex, CP violating terms here are not mixed with backgrounds. Fitting experimental data with such differential cross sections can give more information on CP violation than the partial width, e.g., if we integrate out the variable $`p(k_1k_2)`$, the part proportional to $`f_1`$(or $`f_1^{}`$) will vanish.
We would like to point out that to sum over the spins of $`J/\psi `$ or averaging over the spins of $`\psi (2S)`$ in the cross section when comparing with data(as some references) is incorrect. $`\psi (2S)`$ are polarized. The density matrix of $`\psi (2S)`$ is determined by a vector coupling with $`e^+e^{}`$ at $`10^4`$ precision. And because $`J/\psi `$ is re-constructed through $`e^+e^{}`$ and $`\mu ^+\mu ^{}`$ channels, $`J/\psi `$ with helicity-0 has a much less chance to be selected than those with helicity-$`\pm `$1.
One might doubt if the CP-violating effects in the processes can be measured, given that they are mediated by the strong interactions. The Standard Model prediction of the CP-violating effects in the processes are difficult to evaluate and it is out of the scope of our present paper. However, we can give a rough estimation. The ratio of the weak interactions and the strong interactions at $`c\overline{c}`$-meson energy scale should be taken as $`G_Fm_{J/\psi }^210^4`$. There is a unique source for CP violation in the Standard Model. Any CP-violating effects in the Standard Model is proportional to the rephasing-invariant $`Im\mathrm{\Delta }^{(4)}`$ . The magnitude of CP violation comparing with the weak interactions can be represented by the $`\eta `$-parameter measured in $`K^0`$ decays, $`\eta 2\times 10^3`$ . So we come to the conclusion that CP-violating effect in the Standard Model, when compared with the strong interactions, is of the order $`10^7`$.
But it is not the case when a process is flavor disconnected and the strong interactions in it are much smaller. As we have discussed above, there is no resonance in $`\psi ^{}J/\psi \pi \pi `$ processes. Now suppose the decay is through single gluon exchange, as shown in Fig. 3. Gauge invariance demands that
$$k^\alpha \mathrm{\Gamma }_{1\mu \nu \alpha }e_\psi ^{}^\mu e_{J/\psi }^\nu =0,$$
(46)
so the only possible form of the gluon-$`\pi `$-$`\pi `$ vertex is
$$\mathrm{\Gamma }_2^\beta =c(k_1k_2)^\beta ,$$
(47)
with $`c`$ a constant. The two pions must form a isospin singlet provided the isospin symmetry conserved, and this require $`\mathrm{\Gamma }_2^\beta `$ symmetric when exchanging $`k_1`$ and $`k_2`$. So the single gluon transition in this decay, comparing with flavor connected processes, is suppressed by $`\alpha _S`$ and isospin symmetry. Two-gluon transitions keeping isospin symmetry are suppressed by $`\alpha _S^2`$. Be advised that $`\alpha _S`$ is not the coupling constant in QCD which is very large at low energy. Here it is the effective coupling of gluon and should be very small. The precise value of such transitions are very difficult to calculate. However, we can evaluate such a suppression by comparing the decay $`\varphi \pi ^+\pi ^{}`$ with $`\omega \pi ^+\pi ^{}`$. The difference between two processes is that the former one is flavor disconnected(both processes violate isospin symmetry). The amplitude of the two processes are all of the form
$$A_\lambda =c_{\varphi ,\omega }e^\mu (p,\lambda )(p_{\pi ^+}p_\pi ^{})_\mu ,$$
(48)
here $`c_\varphi `$ and $`c_\omega `$ are coupling constants. The ratio of the partial width of the two processes is
$$\frac{(14m_\pi ^2/m_\varphi ^2)^{3/2}|c_\varphi |^2m_\varphi }{(14m_\pi ^2/m_\omega ^2)^{3/2}|c_\omega |^2m_\omega }1.4\left|\frac{c_\varphi }{c_\omega }\right|^2$$
(49)
The experimental value of the partial widths for the decay $`\varphi \pi ^+\pi ^{}`$ is $`4.43\times (8\times 10^5)3.5\times 10^4`$ MeV, and $`8.41\times 2.21\%0.19`$ MeV for $`\omega \pi ^+\pi ^{}`$. One can find that the coupling constant of a flavor disconnected vertex is suppressed by a factor of $`10^2`$. Now we can conclude that the CP-violating effect predicted by the Standard Model in the process $`\psi (2S)J/\psi \pi \pi `$ is of the order $`10^5`$, comparing with the background amplitudes. For the process $`\mathrm{{\rm Y}}(2S)\mathrm{{\rm Y}}(1S)`$, it is estimated to be of $`10^4`$ order.
It is believed that the Standard Model does not provide the complete description of CP violation in nature. Almost any extension of the Standard Model has additional sources of CP violating effects. In addition, theories that explain the observed baryon asymmetry of the universe must include new sources of CP violation . The Standard Model can not generate a large enough matter-antimatter imbalance to produce the baryon number to entropy ratio observed in the universe today . If the CP violating effect beyond the Standard Model is about ten times larger, it will be of the order $`10^4`$ comparing with backgrounds.
BES has accumulated $`3.8\times 10^6`$ $`\psi (2S)`$ events . These events even make it possible to see the $`\pi ^+\pi ^{}`$ D-wave ($`𝒪(k^2)`$) in the $`J/\psi \pi ^+\pi ^{}`$ decay mode , although the D-wave is highly suppressed . Noticing that the leading CP violating terms ($`f_1`$,$`f_8`$ and $`g_5`$) are $`𝒪(k)`$, it is significative to determine whether they are zero using current data. With BEPC has the ability to accumulate $`10^8`$ events within a few months, and it is scheduled for a upgrade. The upgraded BEPC will be able to accumulate $`10^910^{10}`$ events and can measure the CP violating parameters at a high accuracy. Provided that the CP violating effects beyond the Standard Model are of $`10^4`$ order, they can be detected in the near future.
We have given the model independent amplitudes for the decay $`\psi (2S)J/\psi \pi \pi `$ and suggest to search for CP violation in these processes. It is practical to measure the CP violation parameters. And it has the advantage that CP violation observables can be directly measured, so that one does not need to track two or more CP-conjugate processes. The extension of our results to the case of $`\mathrm{{\rm Y}}(2S)\mathrm{{\rm Y}}(1S)\pi \pi `$ and $`\mathrm{{\rm Y}}(3S)\mathrm{{\rm Y}}(2S)\pi \pi `$ is straightforward. $`\mathrm{{\rm Y}}(2S)`$ and $`\mathrm{{\rm Y}}(3S)`$ can be produced at B-factories.
Acknowledgement: JJZ would like to thank Dr. Xiao-Jun Wang and Ting-Liang Zhuang for discussions. The work is partly supported by funds from IHEP and NSF of China through C N Yang. |
warning/0001/hep-ex0001060.html | ar5iv | text | # Search for 𝐷⁰-𝐷̄⁰ Mixing
## Abstract
We report on a search for $`D^0\overline{D}^0`$ mixing made by studying the ‘wrong-sign’ process $`D^0K^+\pi ^{}`$. The data come from an integrated luminosity of 9.0 fb<sup>-1</sup> of $`e^+e^{}`$ collisions at $`\sqrt{s}10`$GeV recorded with the CLEO II.V detector. We measure the time integrated rate of the ‘wrong-sign’ process $`D^0K^+\pi ^{}`$ relative to that of the Cabibbo-favored process $`\overline{D}^0K^+\pi ^{}`$ to be $`R=(0.332_{0.065}^{+0.063}\pm 0.040)\%`$. We study $`D^0K^+\pi ^{}`$ as a function of decay time to distinguish direct doubly Cabibbo-suppressed decay from $`D^0\overline{D}^0`$ mixing. The amplitudes that describe $`D^0\overline{D}^0`$ mixing, $`x^{}`$ and $`y^{}`$, are consistent with zero. At the 95% C.L. and without assumptions concerning charge-parity (CP) violating parameters, we find $`(1/2)x^2<0.041\%`$ and $`5.8\%<y^{}<1.0\%`$.
preprint: CLNS 99/1659 CLEO 99-23
R. Godang,<sup>1</sup> K. Kinoshita,<sup>1,</sup><sup>*</sup><sup>*</sup>*Permanent address: University of Cincinnati, Cincinnati OH 45221 I. C. Lai,<sup>1</sup> S. Schrenk,<sup>1</sup> G. Bonvicini,<sup>2</sup> D. Cinabro,<sup>2</sup> L. P. Perera,<sup>2</sup> G. J. Zhou,<sup>2</sup> G. Eigen,<sup>3</sup> E. Lipeles,<sup>3</sup> M. Schmidtler,<sup>3</sup> A. Shapiro,<sup>3</sup> W. M. Sun,<sup>3</sup> A. J. Weinstein,<sup>3</sup> F. Würthwein,<sup>3,</sup>Permanent address: Massachusetts Institute of Technology, Cambridge, MA 02139. D. E. Jaffe,<sup>4</sup> G. Masek,<sup>4</sup> H. P. Paar,<sup>4</sup> E. M. Potter,<sup>4</sup> S. Prell,<sup>4</sup> V. Sharma,<sup>4</sup> D. M. Asner,<sup>5</sup> A. Eppich,<sup>5</sup> J. Gronberg,<sup>5</sup> T. S. Hill,<sup>5</sup> C. M. Korte,<sup>5</sup> R. Kutschke,<sup>5</sup> D. J. Lange,<sup>5</sup> R. J. Morrison,<sup>5</sup> H. N. Nelson,<sup>5</sup> C. Qiao,<sup>5</sup> A. Ryd,<sup>5</sup> H. Tajima,<sup>5</sup> M. S. Witherell,<sup>5</sup> R. A. Briere,<sup>6</sup> B. H. Behrens,<sup>7</sup> W. T. Ford,<sup>7</sup> A. Gritsan,<sup>7</sup> J. Roy,<sup>7</sup> J. G. Smith,<sup>7</sup> J. P. Alexander,<sup>8</sup> R. Baker,<sup>8</sup> C. Bebek,<sup>8</sup> B. E. Berger,<sup>8</sup> K. Berkelman,<sup>8</sup> F. Blanc,<sup>8</sup> V. Boisvert,<sup>8</sup> D. G. Cassel,<sup>8</sup> M. Dickson,<sup>8</sup> P. S. Drell,<sup>8</sup> K. M. Ecklund,<sup>8</sup> R. Ehrlich,<sup>8</sup> A. D. Foland,<sup>8</sup> P. Gaidarev,<sup>8</sup> L. Gibbons,<sup>8</sup> B. Gittelman,<sup>8</sup> S. W. Gray,<sup>8</sup> D. L. Hartill,<sup>8</sup> B. K. Heltsley,<sup>8</sup> P. I. Hopman,<sup>8</sup> C. D. Jones,<sup>8</sup> N. Katayama,<sup>8</sup> D. L. Kreinick,<sup>8</sup> M. Lohner,<sup>8</sup> A. Magerkurth,<sup>8</sup> T. O. Meyer,<sup>8</sup> N. B. Mistry,<sup>8</sup> C. R. Ng,<sup>8</sup> E. Nordberg,<sup>8</sup> J. R. Patterson,<sup>8</sup> D. Peterson,<sup>8</sup> D. Riley,<sup>8</sup> J. G. Thayer,<sup>8</sup> P. G. Thies,<sup>8</sup> B. Valant-Spaight,<sup>8</sup> A. Warburton,<sup>8</sup> P. Avery,<sup>9</sup> C. Prescott,<sup>9</sup> A. I. Rubiera,<sup>9</sup> J. Yelton,<sup>9</sup> J. Zheng,<sup>9</sup> G. Brandenburg,<sup>10</sup> A. Ershov,<sup>10</sup> Y. S. Gao,<sup>10</sup> D. Y.-J. Kim,<sup>10</sup> R. Wilson,<sup>10</sup> T. E. Browder,<sup>11</sup> Y. Li,<sup>11</sup> J. L. Rodriguez,<sup>11</sup> H. Yamamoto,<sup>11</sup> T. Bergfeld,<sup>12</sup> B. I. Eisenstein,<sup>12</sup> J. Ernst,<sup>12</sup> G. E. Gladding,<sup>12</sup> G. D. Gollin,<sup>12</sup> R. M. Hans,<sup>12</sup> E. Johnson,<sup>12</sup> I. Karliner,<sup>12</sup> M. A. Marsh,<sup>12</sup> M. Palmer,<sup>12</sup> C. Plager,<sup>12</sup> C. Sedlack,<sup>12</sup> M. Selen,<sup>12</sup> J. J. Thaler,<sup>12</sup> J. Williams,<sup>12</sup> K. W. Edwards,<sup>13</sup> R. Janicek,<sup>14</sup> P. M. Patel,<sup>14</sup> A. J. Sadoff,<sup>15</sup> R. Ammar,<sup>16</sup> A. Bean,<sup>16</sup> D. Besson,<sup>16</sup> R. Davis,<sup>16</sup> I. Kravchenko,<sup>16</sup> N. Kwak,<sup>16</sup> X. Zhao,<sup>16</sup> S. Anderson,<sup>17</sup> V. V. Frolov,<sup>17</sup> Y. Kubota,<sup>17</sup> S. J. Lee,<sup>17</sup> R. Mahapatra,<sup>17</sup> J. J. O’Neill,<sup>17</sup> R. Poling,<sup>17</sup> T. Riehle,<sup>17</sup> A. Smith,<sup>17</sup> J. Urheim,<sup>17</sup> S. Ahmed,<sup>18</sup> M. S. Alam,<sup>18</sup> S. B. Athar,<sup>18</sup> L. Jian,<sup>18</sup> L. Ling,<sup>18</sup> A. H. Mahmood,<sup>18,</sup>Permanent address: University of Texas - Pan American, Edinburg TX 78539. M. Saleem,<sup>18</sup> S. Timm,<sup>18</sup> F. Wappler,<sup>18</sup> A. Anastassov,<sup>19</sup> J. E. Duboscq,<sup>19</sup> K. K. Gan,<sup>19</sup> C. Gwon,<sup>19</sup> T. Hart,<sup>19</sup> K. Honscheid,<sup>19</sup> D. Hufnagel,<sup>19</sup> H. Kagan,<sup>19</sup> R. Kass,<sup>19</sup> T. K. Pedlar,<sup>19</sup> H. Schwarthoff,<sup>19</sup> J. B. Thayer,<sup>19</sup> E. von Toerne,<sup>19</sup> M. M. Zoeller,<sup>19</sup> S. J. Richichi,<sup>20</sup> H. Severini,<sup>20</sup> P. Skubic,<sup>20</sup> A. Undrus,<sup>20</sup> S. Chen,<sup>21</sup> J. Fast,<sup>21</sup> J. W. Hinson,<sup>21</sup> J. Lee,<sup>21</sup> N. Menon,<sup>21</sup> D. H. Miller,<sup>21</sup> E. I. Shibata,<sup>21</sup> I. P. J. Shipsey,<sup>21</sup> V. Pavlunin,<sup>21</sup> D. Cronin-Hennessy,<sup>22</sup> Y. Kwon,<sup>22,</sup><sup>§</sup><sup>§</sup>§Permanent address: Yonsei University, Seoul 120-749, Korea. A.L. Lyon,<sup>22</sup> E. H. Thorndike,<sup>22</sup> C. P. Jessop,<sup>23</sup> H. Marsiske,<sup>23</sup> M. L. Perl,<sup>23</sup> V. Savinov,<sup>23</sup> D. Ugolini,<sup>23</sup> X. Zhou,<sup>23</sup> T. E. Coan,<sup>24</sup> V. Fadeyev,<sup>24</sup> Y. Maravin,<sup>24</sup> I. Narsky,<sup>24</sup> R. Stroynowski,<sup>24</sup> J. Ye,<sup>24</sup> T. Wlodek,<sup>24</sup> M. Artuso,<sup>25</sup> R. Ayad,<sup>25</sup> C. Boulahouache,<sup>25</sup> K. Bukin,<sup>25</sup> E. Dambasuren,<sup>25</sup> S. Karamov,<sup>25</sup> S. Kopp,<sup>25</sup> G. Majumder,<sup>25</sup> G. C. Moneti,<sup>25</sup> R. Mountain,<sup>25</sup> S. Schuh,<sup>25</sup> T. Skwarnicki,<sup>25</sup> S. Stone,<sup>25</sup> G. Viehhauser,<sup>25</sup> J.C. Wang,<sup>25</sup> A. Wolf,<sup>25</sup> J. Wu,<sup>25</sup> S. E. Csorna,<sup>26</sup> I. Danko,<sup>26</sup> K. W. McLean,<sup>26</sup> Sz. Márka,<sup>26</sup> and Z. Xu<sup>26</sup>
<sup>1</sup>Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061
<sup>2</sup>Wayne State University, Detroit, Michigan 48202
<sup>3</sup>California Institute of Technology, Pasadena, California 91125
<sup>4</sup>University of California, San Diego, La Jolla, California 92093
<sup>5</sup>University of California, Santa Barbara, California 93106
<sup>6</sup>Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
<sup>7</sup>University of Colorado, Boulder, Colorado 80309-0390
<sup>8</sup>Cornell University, Ithaca, New York 14853
<sup>9</sup>University of Florida, Gainesville, Florida 32611
<sup>10</sup>Harvard University, Cambridge, Massachusetts 02138
<sup>11</sup>University of Hawaii at Manoa, Honolulu, Hawaii 96822
<sup>12</sup>University of Illinois, Urbana-Champaign, Illinois 61801
<sup>13</sup>Carleton University, Ottawa, Ontario, Canada K1S 5B6
and the Institute of Particle Physics, Canada
<sup>14</sup>McGill University, Montréal, Québec, Canada H3A 2T8
and the Institute of Particle Physics, Canada
<sup>15</sup>Ithaca College, Ithaca, New York 14850
<sup>16</sup>University of Kansas, Lawrence, Kansas 66045
<sup>17</sup>University of Minnesota, Minneapolis, Minnesota 55455
<sup>18</sup>State University of New York at Albany, Albany, New York 12222
<sup>19</sup>Ohio State University, Columbus, Ohio 43210
<sup>20</sup>University of Oklahoma, Norman, Oklahoma 73019
<sup>21</sup>Purdue University, West Lafayette, Indiana 47907
<sup>22</sup>University of Rochester, Rochester, New York 14627
<sup>23</sup>Stanford Linear Accelerator Center, Stanford University, Stanford, California 94309
<sup>24</sup>Southern Methodist University, Dallas, Texas 75275
<sup>25</sup>Syracuse University, Syracuse, New York 13244
<sup>26</sup>Vanderbilt University, Nashville, Tennessee 37235
Studies of the evolution of a $`K^0`$ or $`B_d^0`$ into the respective anti-particle, a $`\overline{K}^0`$ or $`\overline{B}_d^0`$, have guided the form and content of the Standard Model, and permitted useful estimates of the masses of the charm and top quark prior to their direct observation. In this Letter, we present the results of a search for the evolution of the $`D^0`$ into the $`\overline{D}^0`$ . Our principal motivation is to observe new physics outside the Standard Model.
A $`D^0`$ can evolve into a $`\overline{D}^0`$ through on-shell intermediate states, such as $`K^+K^{}`$ with mass, $`m_{K^+K^{}}=m_{D^0}`$, or through off-shell intermediate states, such as those that might be present due to new physics. We denote the amplitude through the former (latter) states by $`iy`$ ($`x`$), in units of $`\mathrm{\Gamma }_{D^0}/2`$ . Many predictions for $`x`$ in the $`D^0\overline{D}^0`$ amplitude have been made . The Standard Model contributions are suppressed to $`|x|\mathrm{tan}^2\theta _C5\%`$ because $`D^0`$ decay is Cabibbo-favored; the GIM cancellation could further suppress $`|x|`$ down to $`10^610^2`$. Many non-Standard Models predict $`|x|>1\%`$. Contributions to $`x`$ at this level could result from the presence of new particles with masses as high as 100-1000 TeV . Signatures of new physics include $`|x||y|`$, or charge-parity (CP) violating interference between $`x`$ and $`y`$, or between $`x`$ and a direct decay amplitude. In order to assess the origin of a $`D^0\overline{D}^0`$ mixing signal, the effects described by $`y`$ must be distinguished from those described by $`x`$.
The wrong-sign (WS) process, $`D^0K^+\pi ^{}`$, can proceed either through direct doubly Cabibbo-suppressed decay (DCSD), or through state-mixing followed by the Cabibbo-favored decay (CFD), $`D^0\overline{D}^0K^+\pi ^{}`$. Both processes could contribute to the time integrated WS rate $`R=(f+\overline{f})/2`$, and the inclusive CP asymmetry $`A=(f\overline{f})/(f+\overline{f})`$, where $`f=\mathrm{\Gamma }(D^0K^+\pi ^{})/\mathrm{\Gamma }(\overline{D}^0K^+\pi ^{})`$, and $`\overline{f}`$ is defined by the application of charge-conjugation to $`f`$.
To disentangle the processes that could contribute to $`D^0K^+\pi ^{}`$, we study the distribution of WS final states as a function of the proper decay time $`t`$ of the $`D^0`$. We describe the proper decay time in units of the mean $`D^0`$ lifetime, $`\tau _{D^0}=415\pm 4`$fs . The differential WS rate is
$$r(t)[R_D+\sqrt{R_D}y^{}t+\frac{1}{4}(x^2+y^2)t^2]e^t.$$
(1)
The modified mixing amplitudes $`x^{}`$ and $`y^{}`$ in Eqn. 1 are given by $`y^{}y\mathrm{cos}\delta x\mathrm{sin}\delta `$, $`x^{}x\mathrm{cos}\delta +y\mathrm{sin}\delta `$ and $`R_M\frac{1}{2}\left(x^2+y^2\right)=\frac{1}{2}\left(x^2+y^2\right)`$ where $`\delta `$ is a possible strong phase between the DCSD and CFD amplitudes. The symbol $`R_D`$ ($`R_M`$) represents the DCSD (mixing) rate relative to the CFD rate. There are theoretical arguments that $`\delta `$ is small , which have recently been questioned.
The influence of each of $`x^{}`$, $`y^{}`$, and $`R_D`$ on $`r(t)`$ in Eqn. 1 is distinguishable. Such behavior is complementary to the time dependence of the decay rate to CP eigenstates such as $`D^0K^+K^{}`$ that is primarily sensitive to $`y`$, or that of $`D^0K^+\mathrm{}^{}\overline{\nu }_{\mathrm{}}`$ that is sensitive to $`R_M`$ alone.
We characterize the violation of CP in state-mixing, direct decay, and the interference between those two processes, respectively, by the real-valued parameters $`A_M`$, $`A_D`$, and $`\varphi `$, where, to leading order, both $`x^{}`$ and $`y^{}`$ are scaled by $`(1\pm A_M/2)`$, $`R_DR_D\left(1\pm A_D\right)`$, $`\delta \delta \pm \varphi `$ in Eqn. 1 . The plus (minus) sign is used for an initial $`D^0(\overline{D}^0)`$.
Our data were accumulated between Feb. 1996 and Feb. 1999 from an integrated luminosity of 9.0 fb<sup>-1</sup> of $`e^+e^{}`$ collisions at $`\sqrt{s}10`$GeV provided by the Cornell Electron Storage Ring (CESR). The data were taken with the CLEO II multipurpose detector , upgraded in 1995 when a silicon vertex detector (SVX) was installed and the drift chamber gas was changed from argon-ethane to helium-propane. The upgraded configuration is named CLEO II.V.
We reconstruct candidates for the decay sequence $`D^+\pi _s^+D^0`$, $`D^0K^\pm \pi ^{}`$. The charge of the slow pion ($`\pi _s^+`$ or $`\pi _s^{}`$) identifies the charm state at $`t=0`$ as either $`D^0`$ or $`\overline{D}^0`$. We require the $`D^+`$ momentum, $`p_D^{}`$, to exceed 2.2 GeV, and we require the $`D^0`$ to produce either the final state $`K^+\pi ^{}`$ (WS) or $`K^{}\pi ^+`$ (right-sign (RS)). The broad features of the reconstruction are similar to those employed in the recent CLEO measurement of the $`D`$ meson lifetimes .
The SVX provides precise measurement of the charged particle trajectories, or ‘tracks,’ in three dimensions . We are thus able to refit the $`K^+`$ and $`\pi ^{}`$ tracks with a requirement that they form a common vertex in three dimensions, and require that the confidence level (C.L.) of the refit exceed $`0.01\%`$. We use the trajectory of the $`K^+\pi ^{}`$ system and the position of the CESR luminous region to obtain the $`D^0`$ production point. We refit the $`\pi _s^+`$ track with a requirement that the trajectory intersect the $`D^0`$ production point, and require that the confidence level of the refit exceed $`0.01\%`$.
We reconstruct the energy released in the $`D^+\pi _s^+D^0`$ decay as $`QM^{}Mm_\pi `$, where $`M^{}`$ is the reconstructed mass of the $`\pi _s^+K^+\pi ^{}`$ system, $`M`$ is the reconstructed mass of the $`K^+\pi ^{}`$ system, and $`m_\pi `$ is the charged pion mass. The addition of the $`D^0`$ production point to the $`\pi _s^+`$ trajectory, as well as track-fitting improvements, yields the resolution $`\sigma _Q=190\pm `$6 keV . The use of helium-propane, in addition to improvements in track-fitting, yields the resolution $`\sigma _M=6.4\pm 0.1`$MeV . These resolutions are better than those of earlier studies , and permit improved suppression of background processes.
Candidates must pass two kinematic requirements designed to suppress backgrounds from $`D^0\pi ^+\pi ^{}`$, $`D^0K^+K^{}`$, $`D^0`$multibody, and from cross-feed between WS and RS decays. We evaluate the mass $`M`$ for $`D^0K^+\pi ^{}`$ candidates under the three alternate hypotheses $`D^0\pi ^+\pi ^{}`$, $`D^0K^+K^{}`$, and $`D^0\pi ^+K^{}`$. If any one of the three masses falls within $`4\sigma `$ of the nominal $`D^0`$ mass , the $`D^0K^+\pi ^{}`$ candidate is rejected. A conjugate requirement is made for the RS decays. The second kinematic requirement rejects asymmetric $`D^0`$ decays where the pion candidate has low momentum with the requirement that $`\mathrm{cos}\theta ^{}>0.8`$ where $`\theta ^{}`$ is the angle of the pion candidate in the $`D^0`$ rest frame with respect to the $`D^0`$ boost. The relative efficiency for the CFD to pass the two kinematic requirements is $`84\%`$ and $`91\%`$, respectively.
We require the specific ionization ($`dE/dx`$) measured in the drift chamber for each charged track agree to within 3$`\sigma `$ of the expected value; this is a loose criterion, and we vary the $`dE/dx`$ requirement for systematic studies.
We reconstruct $`t`$ using only the vertical ($`y`$) component of the flight distance of the $`D^0`$ candidate. This reconstruction is effective because the vertical extent of the $`e^+e^{}`$ luminous region has $`\sigma _y=7\mu m`$. The resolution on the $`D^0`$ decay point $`(x_v,y_v,z_v)`$ is typically $`40\mu `$m in each dimension. We measure the centroid of the luminous region $`(x_b,y_b,z_b)`$ with hadronic events in blocks of data with integrated luminosities of several pb<sup>-1</sup>, and an error on $`y_b`$ that is less than $`5\mu m`$. We reconstruct $`t`$ as $`t=M/p_y\times (y_vy_b)/(c\tau _{D^0})`$ where $`p_y`$ is the $`y`$ component of the total momentum of the $`K^+\pi ^{}`$ system. The error in $`t`$, $`\sigma _t`$, is typically $`0.4`$ (in $`D^0`$ lifetimes), although when the $`D^0`$ direction is near the horizontal plane, $`\sigma _t`$ can be large; we reject $`12\%`$ of the CFD by requiring $`\sigma _t<3/2`$. Studies of the plentiful RS sample allow us to determine our resolution function , and show that biases are negligible for the WS results.
Our signal for the WS process $`D^0K^+\pi ^{}`$ is shown in Fig. 1. We determine the background levels by performing a fit to the two-dimensional region of $`0<Q<10`$MeV versus $`1.76<M<1.97`$GeV that has an area 135 times larger than our signal region. Event samples generated by the Monte Carlo (MC) method and fully simulated in our detector corresponding to $`90`$fb<sup>-1</sup> of integrated luminosity are used to estimate the background shapes in the $`QM`$ plane. The normalizations of the background components with distinct distributions in the $`QM`$ plane are allowed to vary in the fit to the data. The background distributions and normalizations in the $`D^0`$ and $`\overline{D}^0`$ samples are consistent and constrained to be identical. We describe the signal shape in the $`QM`$ plane with the RS data that is within $`4\sigma `$ of the nominal CFD value. The results of the fit are displayed in Fig. 1 and summarized in Table I.
TABLE I. Fitted event yields in a region of $`2\sigma `$ centered on the CFD $`Q`$ and $`M`$ values. The total number of candidates is 82. The estimated background is $`37.2\pm 1.8`$. The bottom row describes the normalization sample.
| Component | Yield |
| --- | --- |
| X $`D^0K^+\pi ^{}`$ (WS) | $`44.8_{8.7}^{+9.7}`$ |
| X random $`\pi _s^+`$, $`\overline{D}\text{}^0K^+\pi ^{}`$ | $`16.0\pm 1.6`$ |
| X $`e^+e^{}c\overline{c}`$ bkgd. | $`17.6\pm 0.8`$ |
| X $`e^+e^{}u\overline{u},d\overline{d},s\overline{s}`$ bkgd. | $`3.6\pm 0.4`$ |
| X $`\overline{D}\text{}^0K^+\pi ^{}`$ (RS) | $`13527\pm 116`$ |
The proper decay time distribution is shown in Fig. 2 for WS candidates that are within 2$`\sigma `$ of the CFD signal value in the $`QM`$ plane. We performed maximum-likelihood fits in bins that are 1/20 of the $`D^0`$ lifetime. The background levels are constrained to the levels determined in the fit to the $`QM`$ plane. We use the resolution function in $`t`$ to describe the $`e^+e^{}u\overline{u},d\overline{d},s\overline{s}`$ backgrounds, and an exponential, folded with the resolution function, to describe the $`e^+e^{}c\overline{c}`$ backgrounds. The distribution in $`t`$ of the RS data is used to represent the random $`\pi _s^+`$$`\overline{D}^0K^+\pi ^{}`$ background . The WS signal is described by Eqn. 1, either modified to describe all three forms of CP violation (Fit A), without modification to describe mixing alone (Fit B), or with the mixing parameters constrained to be zero (Fit C). The effect of our resolution is always included.
The reliability of our fit depends upon the simulation of the decay time distribution of the background in the signal region. A comparison of the proper time for the data and MC samples for several sideband regions yields a $`\chi ^2=4.4`$ for 8 degrees of freedom and supports the accuracy of the background simulation .
Our principal results concerning mixing are determined from Fit A. The one-dimensional, 95% confidence intervals for $`x^{}`$, $`y^{}`$ and $`R_D`$, determined by an increase in negative log likelihood ($`\mathrm{ln}`$) of 1.92, are given in Table II. The fits are consistent with an absence of both mixing and CP violation. The small change in likelihood when mixing and CP violation are allowed could be a statistical fluctuation, or an emerging signal.
TABLE II. Results of the fits to the $`D^0K^+\pi ^{}`$ decay time distribution. Fit A allows both $`D^0\overline{D}^0`$ mixing and CP violation. In Fit B, we constrain $`A_M`$, $`A_D`$, and $`\varphi `$ to zero. In Fit C, we constrain $`x^{}`$ and $`y^{}`$ to zero, so $`RR_D`$ and $`AA_D`$. The incremental change in $`\mathrm{ln}`$ for Fit B (Fit C) with respect to the Fit A (Fit B) is 0.07 (1.57).
| Parameter | Best Fit | 95% C.L. |
| --- | --- | --- |
| Fit A a | Most General Fit | |
| $`A^{A^A}`$$`R_D`$ | $`(0.48\pm 0.12\pm 0.04)\%`$ | $`0.24\%<R_D<0.71\%`$ |
| $`y^{}`$ | $`(2.5_{1.6}^{+1.4}\pm 0.3)\%`$ | $`5.8\%<y^{}<1.0\%`$ |
| $`x^{}`$ | $`(0\pm 1.5\pm 0.2)\%`$ | $`|x^{}|<2.9\%`$ |
| $`(1/2)x^2`$ | | $`<0.041\%`$ |
| | CP violating parameters | |
| $`A_M`$ | $`0.23_{0.80}^{+0.63}\pm 0.01`$ | No Limit |
| $`A_D`$ | $`0.01_{0.17}^{+0.16}\pm 0.01`$ | $`0.36<A_D<0.30`$ |
| $`\mathrm{sin}\varphi `$ | $`0.00\pm 0.60\pm 0.01`$ | No Limit |
| Fit B a | CP conserving fit | |
| $`A^{A^A}`$$`R_D`$ | $`(0.47_{0.12}^{+0.11}\pm 0.04)\%`$ | $`0.24\%<R_D<0.69\%`$ |
| $`y^{}`$ | $`(2.3_{1.4}^{+1.3}\pm 0.3)\%`$ | $`5.2\%<y^{}<0.2\%`$ |
| $`x^{}`$ | $`(0\pm 1.5\pm 0.2)\%`$ | $`|x^{}|<2.8\%`$ |
| $`(1/2)x^2`$ | | $`<0.038\%`$ |
| Fit C a | No-mixing fit $`RR_D`$ , $`AA_D`$ | |
| $`R`$ | $`(0.332_{0.065}^{+0.063}\pm 0.040)\%`$ | |
| $`A`$ | $`0.02_{0.20}^{+0.19}\pm 0.01`$ | $`0.43<A<0.34`$ |
| X $`R/\mathrm{tan}^4\theta _C`$ | $`(1.24_{0.24}^{+0.23}\pm 0.15)`$ | |
| $`(D^0K^+\pi ^{})`$ | $`(1.28_{0.25}^{+0.24}\pm 0.15\pm 0.03)\times 10^4`$ | |
We make contours in the two-dimensional plane of $`y^{}`$ versus $`x^{}`$ that contain the true value of $`x^{}`$ and $`y^{}`$ with 95% confidence, for Fit A and Fit B. The contour is where $`\mathrm{ln}`$ increases from the best fit value by 3.0. All fit variables other than $`x^{}`$ and $`y^{}`$ are allowed to vary to best fit values at each point on the contour. The interior of the contour is the tightly cross-hatched region near the origin of Fig. 3. The limits are not substantially degraded when the most general CP violation is allowed, in part because our acceptance, unlike that of earlier studies , is uniform as a function of $`D^0`$ decay time.
Many classes of systematic error cancel due to the similarity of the events that comprise the numerators and denominators of $`R`$ and $`A`$. The dominant systematic errors stem from potential misunderstanding of the shapes and acceptances for our backgrounds. We vary the selection criteria to estimate these systematic errors from the data. The level and composition of the backgrounds are sensitive to the requirements on momentum magnitude and direction, and $`dE/dx`$ of the charged particle trajectories and contribute $`\pm 0.018\%`$, $`\pm 0.018\%`$ and $`\pm 0.026\%`$, respectively, to the systematic error in $`R_D`$. We also include the statistical uncertainty on the MC determination of the proper time for the $`e^+e^{}q\overline{q}`$ backgrounds in the systematic error. We assess a total systematic error on $`R_D`$, $`x^{}`$ and $`y^{}`$ of $`\pm 0.040\%`$, $`\pm 0.2\%`$ and $`\pm 0.3\%`$, respectively. A study of detector-induced and event-reconstruction-induced asymmetries in CFD limits the relative systematic error on $`A`$ to $`<1\%`$.
If we assume that $`\delta `$ is small, then $`x^{}x`$ and we can indicate the impact of our work in limiting predictions of $`D^0\overline{D}^0`$ mixing from extensions to the Standard Model. The 95% C.L. interval for $`x`$ from Fit A has some inconsistency with eighteen of the predictions tabulated in Ref. . A new model invokes SUSY to account for the value of $`ϵ^{}/ϵ`$ and estimates $`D^0\overline{D}^0`$ mixing with $`x=0.6\%`$, somewhat below our sensitivity. Another analysis notes that SUSY could induce $`A_D30`$%, just below our sensitivity.
In conclusion, our data are consistent with no $`D^0\overline{D}^0`$ mixing. We limit the mixing amplitudes, $`x^{}`$ and $`y^{}`$, to be $`(1/2)x^2<0.041\%`$ and $`5.8\%<y^{}<1.0\%`$ at the 95% C.L., without assumptions concerning CP violating parameters. We have observed $`44.8_{8.7}^{+9.7}`$ candidates for the decay $`D^0K^+\pi ^{}`$ corresponding to $`R=(0.332_{0.065}^{+0.063}\pm 0.040)\%`$. We observe no evidence for CP violation. These results are a substantial advance in sensitivity to the phenomena that contribute to the wrong-sign process $`D^0K^+\pi ^{}`$.
We gratefully acknowledge the effort of the CESR staff in providing us with excellent luminosity and running conditions. We wish to acknowledge and thank the technical staff who contributed to the success of the CLEO II.V detector upgrade, including J. Cherwinka and J. Dobbins (Cornell); M. O’Neill (CRPP); M. Haney (Illinois); M. Studer and B. Wells (OSU); K. Arndt, D. Hale, and S. Kyre (UCSB). We thank Y. Nir for his help in the development of our description of CP violating effects. This work was supported by the National Science Foundation, the U.S. Department of Energy, Research Corporation, the Natural Sciences and Engineering Research Council of Canada, the A.P. Sloan Foundation, the Swiss National Science Foundation, and the Alexander von Humboldt Stiftung. |
warning/0001/hep-th0001216.html | ar5iv | text | # Untitled Document
KCL-MTH-00-4
hep-th//0001216
Automorphisms, Non-linear Realizations and Branes
P. West
Department of Mathematics
King’s College, London, UK
Abstract
The theory of non-linear realizations is used to derive the dynamics of the branes of M theory. A crucial step in this procedure is to use the enlarged automorphism group of the supersymmetry algebra recently introduced. The field strengths of the worldvolume gauge fields arise as some of the Goldstone fields associated with this automorphism group. The relationship to the superembedding approach is given.
0. Introduction Many superbranes were constructed using a generalisation of the $`\kappa `$-symmetry first found in the point particle and later in the string. These included the construction of the action for the twobrane of M theory and the D-branes in ten dimension . However, the construction of the dynamics of the most sophisticated brane, the M theory fivebrane, was achieved ,, using the superembedding formalism . In fact, some the first brane actions to be constructed were for superbranes in six and two dimensions using the method of non-linear realizations ,. This method has some advantages, namely the construction is entirely group theoretical and the fields have a very natural interpretation in terms of Goldstone particles for broken symmetries. There now exists a substantial literature on the construction of superbrane actions using this method. However, these works are largely confined to branes in dimensions lower than ten dimensions and the branes in eleven and ten dimensions have not been constructed using this method.
While the relationship between the superembedding approach and that using $`\kappa `$-symmetry is well understood in the sense that the $`\kappa `$-symmetry arises as part of the worldvolume supersymmetry which is automatically present in the superembedding approach, the relationship between these approaches and the approach of non-linear realizations is not clear although some discussion is given in reference .
It is the purpose of this paper to give a systematic method of deriving the dynamics of some of the more commonly used superbranes using the theory of non-linear realizations. A crucial step in this construction is to include in the group not only the usual Spin group, but its generalisation to a much larger automorphism that exists for supersymmetry algebras whose central charges form an arbitrary matrix. This group was introduced in reference and applied within the context of the symmetries of the covariant equations of motion of the M theory fivebrane. In particular, it was shown that the currents for supersymmetry translations and the second rank central charge arose by expressing the equations of motion for the fermion, scalars and second rank tensor gauge field respectively as total derivatives. This suggests that even this brane possessed a construction as a non-linear realization. It was also shown that the fivebrane possessed a further symmetry with a rank three generator that was part of the $`GL(32)`$ automorphism group of the eleven dimensional supersymmetry algebra with all its central charges. The extra generator was associated with the two form gauge field of the fivebrane. These large automorphism groups are only present if one includes sufficient central charges in the supersymmetry algebra. Indeed, if one includes all the possible central charges in the superalgebra, then the automorphism group is $`GL(c_d)`$ where $`c_d`$ are the number of Majorana supercharges. This is the natural generalisation to branes of the Spin and internal symmetry groups that occur if one considers only point particles and so has only a momentum generator in the anti-commutator of the supercharges. In the past, many constructions of branes have not taken into account in the coset construction any automorphisms of the supersymmetry algebra. Some notable exceptions that did include the well known Lorentz and internal groups were given in references which treated some superbranes in six and four dimensions respectively and the point particle in lower dimensions respectively. As we shall see, to construct the branes of M theory using the method of non-linear realisations one must use the much larger automorphism groups. This is consistent with the old result that the topological charges of branes occur as the central charges in the supersymmetry algebra . In the approach of non-linear realizations given in this paper, the worldvolume fields of the superbrane are certain of the Goldstone fields associated with this automorphism group.
We will also find the relationship between the superembedding approach and that of the non-linear realizations. In the latter one first derives the Cartan forms from the group elements and then constructs from them a group invariant dynamics. It was noticed long ago that for certain choices of groups one could set certain of the Cartan forms to vanish in a group invariant manner. This was the so-called inverse Higgs effect . We will show that for the supergroups we consider, we can choose such a constraint on the Cartan forms which is precisely equivalent to the condition assumed in the superembedding formalism.
1. The Theory of Non-linear Realizations We first briefly review the theory of non-linear realizations set out by Coleman, Wess and Zumino and extended by Volkov . Rather than consider a general group we restrict ourselves to the (super) group $`\underset{¯}{G}`$ whose generators can be divided into the two sets $`\underset{¯}{K}`$ and $`\underset{¯}{H}`$. The sets $`\underset{¯}{K}`$ and $`\underset{¯}{H}`$ are both supgroups of $`\underset{¯}{G}`$ and $`\underset{¯}{H}`$ is the automorphism group, of $`\underset{¯}{K}`$. The generators in each of the above two classes are further divided as $`\underset{¯}{K}=\{K,K^{}\}`$ and $`\underset{¯}{R}=\{R,R^{}\}`$. The generators $`K`$ and $`R`$ both form subalgebras of the Lie (super) algebra $`G`$ whose associated groups we denote by $`K`$ and $`H`$ respectively. The generators of $`H`$ are an automorphisms of $`K`$. The division of the generators of $`\underset{¯}{G}`$ into these four classes corresponds to:
$`K`$, unbroken generators associated with positions in (super) space-time,
$`K^{}`$, spontaneously broken generators associated with positions in (super) space-time,
$`R`$, unbroken automorphism generators,
$`R^{}`$, spontaneously broken auotmorphism generators. The automorphism generators include those of the Lorentz group or its covering group the corresponding Spin group, in some cases internal group generators, but also other generators which are not usually considered.
¿From now on we will drop the prefix (super), but it is to be understood to be present. As will be clear, the objects associated with the group $`\underset{¯}{G}`$ carry an underlined indices while those associated with the subgroup $`G`$ carry no underline. The decomposition of the former into the latter and the remainder is achieved using unprimed and primed indices respectively.
We wish to consider the coset $`\underset{¯}{G}/H`$. For simplicity we will use an exponential description of group elements and we may then choose the coset representatives to be
$$g=e^{XK}e^{X^{}K^{}}e^{\varphi ^{}R^{}}$$
$`(1.1)`$
In this equation $``$ denotes the relavent summation over the indices. Under any rigid group transformation $`g_0`$ we find that
$$gg_0g=\widehat{g}=e^{\widehat{X}K}e^{\widehat{X}^{}K^{}}e^{\widehat{\varphi }^{}R^{}}h_o$$
$`(1.2)`$
where $`h_0`$ is an element of $`H`$. The Cartan forms are given by
$$g^1dg=FK+F^{}K^{}+\omega ^{}R^{}+\omega R$$
$`(1.3)`$
Under equation (1.2) the Cartan forms transform as
$$\widehat{g}^1d\widehat{g}=h_0(g^1dg)h_0^1+h_0dh_0^1$$
$`(1.4)`$
It can happen that the Cartan forms carry a reducible representation of $`H`$, in which case, certain of the forms can be set to zero. This is the so called inverse Higgs effect . It has the effect of eliminating some of the Goldstone fields in terms of the others. The action or equations of motion are to be constructed from the Cartan forms in such a way that they are invariant under the transformations of equation (1.2).
The conventional interpretation of the above equations is to regard the $`X`$ as the coordinates of (super) space-time and to take the fields $`X^{}`$ and $`\varphi ^{}`$ to depend on them. This leads to a field theory on the coset space $`G/H`$. This approach has been almost universally adopted. However, when considering branes it is more instructive to consider a more general possibility. The brane is moving through the coset space $`\underset{¯}{G}/\underset{¯}{H}`$ with tangent group $`\underset{¯}{H}`$ and sweeps out a submanifold that has the dimensions of the coset $`G/H`$ and a tangent space group $`H`$. We therefore consider the representatives of the coset $`\underset{¯}{G}/H`$ of equation (1.2) to depend on the variables $`\xi `$ which parametrises the embedded submanifold. Since the Cartan forms involve the exterior derivative $`d`$ they are independent of the coordinate system used. The Cartan forms associated with $`K`$ are given by $`FK=d\xi FK`$ and similarly for the other Cartan forms. We can think of the $`F`$ in this equation as the vielbein on the embedded submanifold. The covariant derivatives of the Goldstone fields associated with $`K^{}`$ are defined by
$$F^{}K^{}F\mathrm{\Delta }X^{}K^{}=d\xi F^1F^{}K^{}.$$
$`(1.5)`$
The $`\mathrm{\Delta }X^{}`$ are independent of reparamenterisations. A similar construction can be made for the covariant derivatives of $`\varphi ^{}`$. When identifying the fields that can be set to zero, i.e. use the inverse Higgs mechanism, we must not only maintain the $`\underset{¯}{G}`$ invariance of equation (1.2) and also reparameterisation invariance. In effect, this means setting only those covariant derivatives of the Goldstone fields in equation (1.5) that transform in a covariant manner under $`h_0`$ to zero. The equations of motion, or action, are to be constructed from the vielbein and the covariant derivatives of the Goldstone fields that remain. In this way one can find a formulation of brane dynamics that is reparameterisation invariant and also invariant under the rigid $`\underset{¯}{G}`$ transformations of equation (1.2). From this approach we can recover the more conventional approach by using the reparameterisation invariance to choose static gauge, i.e. $`X=\xi `$ for those coordinates that lie in the brane directions. 2. Bosonic Branes In this section we will show that the dynamics of bosonic p-branes in a flat background in $`D`$ dimensional space-time arises as a non-linear realization in the sense of the previous section. We take $`\underset{¯}{G}=ISO(D1,1)`$ with $`\underset{¯}{K}=\{P_{\underset{¯}{n}}\}`$ and $`\underset{¯}{H}=\{J_{\underset{¯}{n}\underset{¯}{m}}\}`$ where $`\underset{¯}{n},\underset{¯}{m}=0,1\mathrm{},D1`$ and $`G=ISO(p,1)`$ with $`K=\{P_n\}`$ and $`H=\{J_{nm}\}`$ where $`n,m=0,1,\mathrm{},p+1`$. This is to be expected as the presence of the p-brane clearly breaks the background space-time group $`ISO(D1,1)`$ to $`ISO(p,1)`$. The Lie algebra of $`ISO(D1,1)`$ is given by
$$[J_{\underset{¯}{n}\underset{¯}{m}},J_{\underset{¯}{p}\underset{¯}{q}}]=\eta _{\underset{¯}{n}\underset{¯}{p}}J_{\underset{¯}{m}\underset{¯}{q}}\eta _{\underset{¯}{m}\underset{¯}{q}}J_{\underset{¯}{n}\underset{¯}{p}}+\eta _{\underset{¯}{n}\underset{¯}{q}}J_{\underset{¯}{m}\underset{¯}{p}}+\eta _{\underset{¯}{m}\underset{¯}{p}}J_{\underset{¯}{n}\underset{¯}{q}}$$
$`(2.1)`$
$$[P_{\underset{¯}{n}},J_{\underset{¯}{p}\underset{¯}{q}}]=+\eta _{\underset{¯}{n}\underset{¯}{p}}P_{\underset{¯}{q}}\eta _{\underset{¯}{n}\underset{¯}{q}}P_{\underset{¯}{p}}$$
$`(2.3)`$
We can write the coset representatives in the form
$$g(X,\varphi )=exp(X^nP_n+X^n^{}P_n^{})exp(\varphi ^{nm^{}}J_{nm^{}}+\varphi ^{n^{}m^{}}J_{n^{}m^{}})$$
$$exp(X^{\underset{¯}{n}}P_{\underset{¯}{n}})exp(\varphi J)$$
$`(2.4)`$
We distinguish $`X`$ from $`X^{}`$ by the indices they carry, in other words the prime on the $`X`$ is understood to be present and we just write $`X^n^{}`$. The Cartan forms are given by
$$g^1dge^nP_n+f^n^{}P_n^{}+\mathrm{\Omega }^{nm^{}}J_{nm^{}}+\mathrm{\Omega }^{n^{}m^{}}J_{n^{}m^{}}+w^{nm}J_{nm}$$
$`(2.5)`$
which we may express as
$$g^1dg=exp(\varphi J)(dX^{\underset{¯}{n}}P_{\underset{¯}{n}})exp(\varphi J)+exp(\varphi J)dexp(\varphi J)$$
$`(2.6)`$
A straightforward calculation shows that
$$e^n=dX^{\underset{¯}{m}}\mathrm{\Phi }_{\underset{¯}{m}}{}_{}{}^{n}=2\varphi ^{nm^{}}dX_m^{}+dX^n+\mathrm{}$$
$$f^n^{}=dX^{\underset{¯}{m}}\mathrm{\Phi }_{\underset{¯}{m}}{}_{}{}^{n^{}}=2\varphi ^{m^{}n^{}}dX_m^{}2\varphi ^{n^{}m}dX_m+dX^n^{}$$
$$\mathrm{\Omega }^{nm^{}}=d\varphi ^{nm^{}}+2(\varphi ^{p^{}m^{}}d\varphi _p^{}^n+\varphi ^{p^{}n}d\varphi _p^{}^m^{})$$
$$\mathrm{\Omega }^{n^{}m^{}}=d\varphi ^{n^{}m^{}}+((\varphi ^{p^{}n^{}}d\varphi _p^{}^m^{}+\varphi ^{pn^{}}d\varphi _p^m^{}(n^{}m^{})),w^{nm}=(\varphi ^{p^{}n}d\varphi _p^{}^m(nm))$$
$`(2.7)`$
where $`\mathrm{\Phi }_{\underset{¯}{n}}^{\underset{¯}{m}}`$ is defined by $`exp(\varphi J)P_{\underset{¯}{n}}exp(\varphi J)=\mathrm{\Phi }_{\underset{¯}{n}}{}_{}{}^{\underset{¯}{m}}P_{\underset{¯}{m}}^{}`$ and $`+\mathrm{}`$ means higher order terms in $`\varphi ^{\underset{¯}{n}\underset{¯}{m}}`$.
Under a group transformation, $`g_0g(X,\varphi )=g(\widehat{X},\widehat{\varphi })h_0`$, the Cartan forms transform by the $`J_{nm}`$ transformations with associated parameter $`r^{nm}`$ in accord with equation (1.4). The effect of the $`P_{\underset{¯}{n}}`$ transformations in $`g_0`$ is to simply transform the $`X^{\underset{¯}{n}}`$. Writing $`h_0=1+r^{nm}J_{nm}`$. we find that
$$\widehat{e}^n=e^ne^pr_p^n,\widehat{f}^n^{}=f^n^{},\widehat{\mathrm{\Omega }}^{nm^{}}=\mathrm{\Omega }^{nm^{}}+2r^{pn}\mathrm{\Omega }_p^m^{}2r^{pm^{}}\mathrm{\Omega }_p^n$$
$`(2.8)`$
$$\widehat{w}^{nm}=w^{nm}2(r^{np}w_p^m(nm))+dr^{nm}$$
$`(2.9)`$
to lowest order in $`r^{nm}`$.
Clearly, we can set $`f^n^{}=0`$ and preserve SO(1,D-1) and reparameterisation symmetry. At the linearized level, examining equation (2.7) we find it implies that $`dX^n^{}=2\varphi _m^n^{}dX^m`$ or
$$2\varphi _m^n^{}=\frac{\xi ^p}{X^m}\frac{X^n^{}}{\xi ^p}$$
$`(2.10)`$
If we choose static gauge this equation becomes,
$$2\varphi _m^n^{}=\frac{X^n^{}}{X^m},$$
$`(2.10)`$
While solving for $`f^n^{}=0`$ to all orders may be complicated it is clear that its content is to solve for $`\varphi _m^n^{}`$ in terms of $`_mX^n^{}`$.
What is really of interest to us is the non-linear form of $`e^n`$ once we have solved this constraint. Examining equation (2.7) we find that the vector $`f^{\underset{¯}{n}}(e^n,f^n^{})`$ is related by a Lorentz transformation to the vector $`(dX^n,dX^n^{})=dX^{\underset{¯}{n}}`$. As such,
$$e_p^n\eta _{nm}e_q^m=_pX^{\underset{¯}{n}}_qX^{\underset{¯}{m}}\eta _{\underset{¯}{n}\underset{¯}{m}}$$
$`(2.11)`$
since $`f^n^{}=0`$. The above expression is invariant under $`g_0`$ transformations. A worldvolume reparameterisation and group invariant action is therefore given by
$$d^p\xi 𝑑ete=d^p\xi \sqrt{det(_pX^{\underset{¯}{n}}_qX^{\underset{¯}{m}}\eta _{\underset{¯}{n}\underset{¯}{m}})}$$
$`(2.12)`$
In other words, the well known generalisation of the Nambu action for the string to a p-brane follows in a straightforward consequence of taking the non-linear realization $`ISO(D1,1)`$ with subgroup $`SO(p,1)`$.
Clearly, had we not included the Lorentz group in our coset and introduced the corresponding Goldstone bosons the veilbein on the brane would have been trivial and we would not have found the above action. 3. Superbranes For superbranes we take the supergroup $`\underset{¯}{G}`$ to be that considered in and applied there to the context of M theory. In accord with section one it has the two subgroups $`\underset{¯}{K}`$ and $`\underset{¯}{H}`$. The generators of $`\underset{¯}{K}`$ obey the relations
$$\{Q_{\underset{¯}{\alpha }},Q_{\underset{¯}{\beta }}\}=Z_{\underset{¯}{\alpha }\underset{¯}{\beta }},[Q_{\underset{¯}{\gamma }},Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}]=0,[Z_{\underset{¯}{\alpha }\underset{¯}{\beta }},Z_{\underset{¯}{\gamma }\underset{¯}{\delta }}]=0,$$
$`(3.1)`$
while those of the automorphism group $`\underset{¯}{H}`$ satisfy
$$[Q_{\underset{¯}{\alpha }},R_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}]=\delta _{\underset{¯}{\alpha }}{}_{}{}^{\underset{¯}{\delta }}Q_{\underset{¯}{\gamma }}^{},[Z_{\underset{¯}{\alpha }\underset{¯}{\beta }},R_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}]=\delta _{\underset{¯}{\alpha }}{}_{}{}^{\underset{¯}{\delta }}Z_{\underset{¯}{\gamma }\underset{¯}{\beta }}^{}+\delta _{\underset{¯}{\beta }}{}_{}{}^{\underset{¯}{\delta }}Z_{\underset{¯}{\alpha }\underset{¯}{\gamma }}^{}.$$
$`(3.2)`$
The generators $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$ are Grassmann even generators labeled by spinor indices. They are obviously symmetric in these indices. It is easy to verify that such an algebra obeys the generalized super Jacobi identities. Let us assume that the generators $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$ form the most general symmetric matrix. Expanding this matrix out in terms of the enveloping algebra of the generators of the relevant Clifford algebra we find that $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$ contains a set of totally anti-symmetric tensorial generators which constitute the central charges:
$$Z_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}=\underset{p}{}\underset{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}{}(\gamma ^{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}C^1)_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}Z_{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}^{}.$$
$`(3.3)`$
The central charges include the momenta $`P_{\underset{¯}{n}}Z_{\underset{¯}{n}}`$. Both the eleven dimensional and the six dimensional superalgebras associated with the fivebrane equations of motion have a supersymmetry algebra in which the central charges form the most general matrix . In the latter case one must take an appropriate index range for the indices $`\underset{¯}{\alpha },\underset{¯}{\beta }\mathrm{}`$. The corresponding superalgebras in the IIA and IIB theories for which the $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$ form the most general matrix in ten dimensions are well known.
The automorphism group is $`GL(c_d)`$ where $`c_d`$ is the number of supercharges. The precise properties of the matrices under complex conjugation being given by implementing the Majorana or other properties of the spinorial supercharge on the commutator relation of the supercharge with the automorphisms. For the case of the eleven dimensional superalgebra it is $`GL(32)`$
To gain a more familiar set of generators we may expand $`R_{\underset{¯}{\gamma }}^{\underset{¯}{\delta }}`$ out in terms of the elements of the enveloping Clifford algebra
$$R_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}=\underset{p}{}\underset{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}{}(\gamma ^{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p})_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}R_{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}^{}.$$
$`(3.4)`$
If the only central charge is the momentum then the right-hand side of the anti-commutator of two supercharges is $`\gamma ^{\underset{¯}{n}}P_{\underset{¯}{n}}`$. In this case, the most general automorphism group is by definition the spin group. Hence, the automorphism group $`GL(32)`$ is the natural generalisation to branes of the spin group relavent to point particles. In the above expansion the generators $`R_{\underset{¯}{n}\underset{¯}{m}}`$ are those of the Spin Lie algebra.
The automorphism group has been found to be relavent to the dynamics of the fivebrane, In particular, its covariant equations of motion have been found to possess additional conserved currents beyond those associated with Spin transformations, central charges and supersymmetry. These new currents lead to conserved charges that generate, at the classical linearized level, a subalgebra of the automorphism group.
We may also consider a subgroup of the full automorphism group. One such is given by the set of symmetric matrices, these form the Lie algebra of Sp($`c_d`$). The generators of this subalgebra are $`S_{\underset{¯}{\gamma }\underset{¯}{\delta }}=R_{\underset{¯}{\gamma }\underset{¯}{\delta }}+R_{\underset{¯}{\delta }\underset{¯}{\gamma }}`$ where $`R_{\underset{¯}{\gamma }\underset{¯}{\delta }}=R_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\beta }}(C^1)_{\underset{¯}{\beta }\underset{¯}{\delta }}^{}`$. Clearly, in the decomposition of equation (3.4) this means keeping only those terms for which the symmetric matrices enter. In eleven dimensions these are the generators $`R_{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}`$ of ranks one, two and five. The commutator of the generators $`S_{\underset{¯}{\gamma }\underset{¯}{\delta }}`$ with the supercharges is given by
$$[Q_{\underset{¯}{\alpha }},S_{\underset{¯}{\gamma }\underset{¯}{\delta }}]=(C^1)_{\underset{¯}{\alpha }\underset{¯}{\delta }}Q_{\underset{¯}{\gamma }}+(C^1)_{\underset{¯}{\alpha }\underset{¯}{\gamma }}Q_{\underset{¯}{\delta }}.$$
$`(3.5)`$
We divide the generators in $`\underset{¯}{K}`$ into $`Q_{\underset{¯}{\alpha }}=(Q_\alpha ,Q_\alpha ^{})`$ and $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}=(Z_{\alpha \beta },Z_{\alpha \beta ^{}},Z_{\alpha ^{}\beta ^{}})`$ and the generators of $`\underset{¯}{H}`$ as $`R_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\beta }}=(R_\gamma {}_{}{}^{\beta },R_\gamma ^{}{}_{}{}^{\beta },R_\gamma {}_{}{}^{\beta ^{}},R_\gamma ^{}{}_{}{}^{\beta ^{}})`$. We take the Lie superalgebra of the group $`G`$ to contain the generators $`Q_\alpha `$ and $`Z_{\alpha \beta }`$ as well as some of the generators $`R_\gamma ^\delta `$
Initially, we will take the non-linear realization approach to the superbrane in which the coset representatives are parameterised by the coordinates on the brane and the transverse fields. Proceeding in this way, we will find the expressions for the superbrane in static gauge. As such, the coset representatives of $`\underset{¯}{G}/H`$ are written as
$$g=e^{X^{\alpha \beta }Z_{\alpha \beta }+\theta ^\alpha Q_\alpha }e^{X^{\alpha \beta ^{}}Z_{\alpha \beta ^{}}+X^{\alpha ^{}\beta ^{}}Z_{\alpha ^{}\beta ^{}}+\mathrm{\Theta }^\alpha ^{}Q_\alpha ^{}}e^{\varphi ^{}R^{}}$$
$`(3.6)`$
where $`\varphi ^{}R^{}`$ is a sum that includes all the generators in $`\underset{¯}{H}`$, except for those in $`H`$. The Cartan forms are given by
$$g^1dg=e^{\varphi ^{}R^{}}de^{\varphi ^{}R^{}}+$$
$$e^{\varphi ^{}R^{}}(Y^{\alpha \beta }Z_{\alpha \beta }+Y^{\alpha \beta ^{}}Z_{\alpha \beta ^{}}+Y^{\alpha ^{}\beta ^{}}Z_{\alpha ^{}\beta ^{}}+d\theta ^\alpha Q_\alpha +d\mathrm{\Theta }^\alpha ^{}Q_\alpha ^{})e^{\varphi ^{}R^{}}$$
$$F^{\alpha \beta }Z_{\alpha \beta }+F^{\alpha \beta ^{}}Z_{\alpha \beta ^{}}+F^{\alpha ^{}\beta ^{}}Z_{\alpha ^{}\beta ^{}}+F^\alpha Q_\alpha +F^\alpha ^{}Q_\alpha ^{}+e^{\varphi ^{}R^{}}de^{\varphi ^{}R^{}},$$
$`(3.7)`$
where $`Y^{\alpha \beta }=dX^{\alpha \beta }+\frac{1}{4}(\theta ^\alpha d\theta ^\beta +\theta ^\beta d\theta ^\alpha )`$, $`Y^{\alpha \beta ^{}}=dX^{\alpha \beta ^{}}d\theta ^\alpha d\mathrm{\Theta }^\beta ^{}`$ and $`Y^{\alpha ^{}\beta ^{}}=dX^{\alpha ^{}\beta ^{}}+\frac{1}{4}(\mathrm{\Theta }^\alpha ^{}d\mathrm{\Theta }^\beta ^{}+\mathrm{\Theta }^\beta ^{}d\mathrm{\Theta }^\alpha ^{})`$
In order to gain some understanding of the physical content of the formalism we will begin by giving a linearized analysis. To lowest order in $`\varphi _{\underset{¯}{\delta }}^{\underset{¯}{\gamma }}`$ the Cartan forms are given by
$$F^\alpha =d\theta ^\alpha +d\mathrm{\Theta }^\delta ^{}\varphi _\delta ^{}{}_{}{}^{\alpha }+d\theta ^\delta \varphi _\delta {}_{}{}^{\alpha },F^{\alpha \beta }=Y^{\alpha \beta }+Y^{\alpha \delta ^{}}\varphi _\delta ^{}{}_{}{}^{\beta }+Y^{\alpha \delta }\varphi _\delta {}_{}{}^{\beta }+Y^{\beta \delta }\varphi _\delta {}_{}{}^{\alpha },$$
$$F^\alpha ^{}=d\mathrm{\Theta }^\alpha ^{}+d\mathrm{\Theta }^\delta ^{}\varphi _\delta ^{}{}_{}{}^{\alpha ^{}}+d\theta ^\delta \varphi _\delta {}_{}{}^{\alpha ^{}},$$
$$F^{\alpha ^{}\beta ^{}}=Y^{\alpha ^{}\beta ^{}}+Y^{\alpha ^{}\delta ^{}}\varphi _\delta ^{}{}_{}{}^{\beta ^{}}++Y^{\beta ^{}\delta ^{}}\varphi _\delta ^{}{}_{}{}^{\alpha ^{}}+Y^{\alpha ^{}\delta }\varphi _\delta {}_{}{}^{\beta ^{}}+Y^{\beta ^{}\delta }\varphi _\delta {}_{}{}^{\alpha ^{}},$$
$$F^{\alpha \beta ^{}}=Y^{\alpha \beta ^{}}+Y^{\alpha \delta ^{}}\varphi _\delta ^{}{}_{}{}^{\beta ^{}}++Y^\beta ^{}\varphi _\delta ^{}{}_{}{}^{\alpha }+Y^{\alpha \delta }\varphi _\delta {}_{}{}^{\beta ^{}}+Y^{\delta \beta ^{}}\varphi _\delta ^\alpha $$
$`(3.8)`$
We wish to place constraints on the Cartan forms that respect the group $`\underset{¯}{G}`$ and lead to the equations of motion of the superbrane. Taking the Goldstone fields to be of order one and the coordinates $`X^{\alpha \beta }`$ and $`\theta ^\alpha `$ to be of order zero, we will first carry out a linearized analysis. As explained in section one, the equation of motion are given by setting certain of the covariant derivatives in the Goldstone fields to vanish. As such, we will only require the supervielbein to order zero. It is straightforward to read it off from equation (3.8) and we find that
$$F=\left(\begin{array}{cc}\frac{1}{2}(\delta _\gamma ^\alpha \delta _\delta ^\beta +\delta _\gamma ^\beta \delta _\delta ^\alpha )& 0\\ \frac{1}{4}(\delta _\gamma ^\alpha \theta ^\beta +\delta _\gamma ^\beta \theta ^\alpha )& \delta _\gamma ^\alpha \end{array}\right)$$
$`(3.9)`$
We can also read off the linearized Cartan forms from equation (3.8) and, multiplying the result by the inverse supervielbein, we find that the covariant derivatives of the Goldstone fields are given by
$$_\alpha \mathrm{\Theta }^\alpha ^{}=D_\alpha \mathrm{\Theta }^\alpha ^{}+\varphi _\alpha {}_{}{}^{\alpha ^{}},_{\gamma \delta }\mathrm{\Theta }^\alpha ^{}=_{\gamma \delta }\mathrm{\Theta }^\alpha ^{},$$
$$_\gamma X^{\alpha \beta ^{}}=D_\gamma X^{\alpha \beta ^{}}\delta _\gamma ^\alpha \mathrm{\Theta }^\beta ^{},_{\gamma \delta }X^{\alpha \beta ^{}}=_{\gamma \delta }X^{\alpha \beta ^{}}+\frac{1}{2}(\delta _\gamma ^\alpha \varphi _\delta {}_{}{}^{\beta ^{}}+\delta _\delta ^\alpha \varphi _\gamma {}_{}{}^{\beta ^{}})$$
$`(3.10)`$
where
$$D_\alpha _\alpha +\frac{1}{2}\theta ^\delta _{\gamma \delta },_{\gamma \delta }\frac{}{X^{\gamma \delta }}$$
$`(3.11)`$
It is a consequence of the linearized analysis that $`\varphi _\gamma ^\beta ^{}`$ is the only part of $`\varphi _{\underset{¯}{\gamma }}^{\underset{¯}{\beta }}`$ that enters any of these expressions. The covariant derivatives of the other Goldstone fields will not interest us here.
We now choose the group $`H`$ to be just the Lorentz group. The Cartan forms carry a very reducible representation with respect to this subgroup and hence with respect to $`\underset{¯}{G}`$. Consequently, we may choose many of the covariant derivatives of the Goldstone fields to vanish. In fact, we will choose the set
$$_\gamma \mathrm{\Theta }^\alpha ^{}=0,$$
$`(3.12)`$
$$_\gamma X^{\alpha \beta ^{}}(\gamma ^n^{}C^1)_{\alpha \beta ^{}}=0$$
$`(3.13)`$
and
$$_{\gamma \delta }\mathrm{\Theta }^\alpha ^{}\frac{2}{r_p}(\gamma ^nC^1)_{\gamma \delta }(\gamma ^nC^1)^{\alpha \beta }_{\alpha \beta }\mathrm{\Theta }^\alpha ^{}=0$$
$`(3.14)`$
where $`r_p`$ is the number of supercharges in $`G`$. The superspaces defined by the above coset space construction contain many central charge associated coordinates in addition to the usual Minkowski coordinates. The Grassmann even coordinates of the background superspace $`\underset{¯}{G}/\underset{¯}{H}`$ include the coordinates $`X^{\underset{¯}{n}}X^{\underset{¯}{\alpha }\underset{¯}{\beta }^{}}(\gamma ^{\underset{¯}{n}}C^1)_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$ of Minkowski space, but also all the other coordinates corresponding to the expansion of the supercharge of equation (3.3). Similarly, the coset space $`G/H`$ on which the above superfields are defined includes the Minkoswki space coordinates $`X^nX^{\alpha \beta ^{}}(\gamma ^nC^1)_{\alpha \beta }`$ as well as a corresponding number of other Grassmann even coordinates associated with the central charges in $`G`$ other than the momentum.
Equation (3.14) sets the central charge dependence of $`\mathrm{\Theta }^\beta ^{}`$ to be trivial except for the dependence on the coordinates $`X^n`$. Equation (3.13) then implies that $`X^n^{}`$ has the same dependence. As such, the above equations are equivalent to the conditions
$$D_\alpha \mathrm{\Theta }^\beta ^{}=\varphi _\alpha ^\beta ^{}$$
$`(3.15)`$
and
$$D_\gamma X^n^{}=(\gamma ^n^{}C^1)_{\gamma \delta ^{}}\mathrm{\Theta }^\delta ^{}$$
$`(3.16)`$
where all the superfields $`\mathrm{\Theta }^\beta ^{}`$, $`X^n^{}`$ and $`\varphi _\alpha ^\beta ^{}`$ depend only on $`X^n`$ and $`\theta ^\alpha `$. It is known from the superembedding approach that these equation imply the correct field content and equations of motion for most of the superbranes including the 2 and 5-branes of M theory.
Let us study the consequences of these equations for the branes in M theory. We must decompose the eleven dimensional Clifford algebra to one that keeps manifest the Clifford algebra appropriate to the brane. This results in a corresponding decomposition of the spinor index $`\underset{¯}{\alpha }=(\alpha ,i)`$. For the fivebrane $`\alpha =1,\mathrm{},4`$ are the Weyl projected spinor indices of Spin(1,5) and $`i=1,\mathrm{},4`$ are the indices of the internal group $`Usp(4)=Spin(5)`$ while for the twobrane $`\alpha =1,2`$ are the spinor indices of Spin(1,2) and $`i=1,\mathrm{},8`$ are indices of $`Spin(8)`$. The analysis of equation (3.16) depends only on the properties of this decomposition of the eleven dimensional Clifford algebra. For the fivebrane of M theory it implies that
$$D_{\alpha i}\mathrm{\Theta }_\beta {}_{}{}^{j}(\text{/})_{\alpha \beta }(\gamma _n^{})_i{}_{}{}^{j}X_{}^{n^{}}+\delta _i^j(\gamma _{n_1n_2n_3})_{\alpha \beta }h^{n_1n_2n_3}$$
$`(3.17)`$
where $`h_{n_1n_2n_3}`$ is the self-dual gauge field strength of the fivebrane. At the linearized level, the field $`h_{n_1n_2n_3}`$ obeys the Bianchi identity implying it is the curl of a second rank gauge field in addition to being self-dual. This relation follows from the constraint of equation (3.16) by applying further covariant derivatives.
For the twobrane we find that
$$D_{\alpha i}\mathrm{\Theta }_\beta {}_{}{}^{j}(\text{/})_{\alpha \beta }(\gamma _n^{})_i{}_{}{}^{j}X_{}^{n^{}}$$
$`(3.18)`$
Equation (3.15) implies that the Goldstone field $`\varphi _\alpha ^\beta ^{}`$ associated with the automorphism group $`\underset{¯}{H}`$ is non-vanishing. Expanding
$$\varphi _{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}=\underset{p}{}\underset{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}{}(\gamma ^{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p})_{\underset{¯}{\gamma }}{}_{}{}^{\underset{¯}{\delta }}\varphi _{\underset{¯}{n}_1\mathrm{}\underset{¯}{n}_p}^{}.$$
$`(3.19)`$
we find that both for the fivebrane and the twobrane that $`\varphi _n{}_{}{}^{m^{}}_nX^m^{}`$ which is the same as the condition of equation (2.10) for the bosonic brane. It relates the Goldstone fields for the broken Lorentz generators to the derivative of Goldstone fields of broken translations. However, for the fivebrane we find in addition that the field strength for the self-dual gauge field is related to the Goldstone fields for broken generators of the automorphism group GL(32), in particular $`h_{n_1n_2n_3}\varphi _{n_1n_2n_3}`$. We note that these generators are the $`R_{nmp}`$ contained in equation (3.4) and that these generators are not contained in the subgroup $`Sp(32)`$. This is consistent with the existence of the additional rank three conserved current found in the covariant equations of motion of the fivebrane.
The non-linear realization described above contains more Goldstone superfields than those of equations (3.15) and (3.16). One could consider imposing more constraints generalizing that of equation (3.16) to include the covariant derivatives of the Goldstone fields associated with the central charges. These would determine these additional fields in terms of the fields that we have already. In particular, one could impose a constraint the implies that the Goldstone field associated with the central charge $`Z_{nm}`$ is the second rank gauge field of the fivebrane.
It is not clear that the choice of subgroup $`H`$ and the related constraints of equations (3.12) to (3.14) are the unique choice that leads to the correct dynamics. Indeed, it would be interesting to explore other choices. In particular, one could also consider constraints which allow the superfields to depend on the central charges of the six dimensional subalgebra, but in a rather restricted way.
To make contact with the superembedding formalism, and to carry out an all orders analysis, it will be convenient to adopt a different parameterisation of the coset representatives. As explained in section one, we can take the Goldstone fields to depend on the coordinates of the superembedded manifold and treat the transverse and longitudinal coordinates of the superbrane on an equal footing. This allows us to work in a superreparameterisation invariant manner and not choose superstatic gauge as above. As such, we choose our coset representatives to be
$$g=e^{X^{\underset{¯}{\alpha }\underset{¯}{\beta }}Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}+\mathrm{\Theta }^{\underset{¯}{\alpha }}Q_{\underset{¯}{\alpha }}}e^{\varphi ^{}R^{}}$$
$`(3.20)`$
Where the coordinates $`Z^{\underset{¯}{M}}(X^{\underset{¯}{\alpha }\underset{¯}{\beta }},\mathrm{\Theta }^{\underset{¯}{\alpha }})`$, are not be confused with the central charges $`Z_{\underset{¯}{\alpha }\underset{¯}{\beta }}`$, depend on the coordinates, $`z^M=(\xi ^{\alpha \beta },\theta ^\alpha )`$ that parameterise the superbrane. The transformation from the fields of equation (3.6) to that in the above equation is obtained by choosing superstatic gauge $`X^{\alpha \beta }=\xi ^{\alpha \beta },\mathrm{\Theta }^\alpha =\theta ^\alpha `$ and making the shift
$$X^{\alpha \beta ^{}}X^{\alpha \beta ^{}}+\frac{1}{2}\theta ^\alpha \mathrm{\Theta }^\beta ^{},$$
$`(3.21)`$
with all the other fields being the same. Constructing the Cartan forms as before, we find that
$$g^1dg=e^{\varphi ^{}R}(E^{\underset{¯}{A}}K_{\underset{¯}{A}})e^{\varphi ^{}R^{}}+e^{\varphi ^{}R^{}}de^{\varphi ^{}R^{}}$$
$$F^{\underset{¯}{A}}K_{\underset{¯}{A}}+e^{\varphi ^{}R^{}}de^{\varphi ^{}R^{}}$$
$`(3.22)`$
In this equation the index $`\underset{¯}{A}=(\underset{¯}{\gamma }\underset{¯}{\delta },\underset{¯}{\gamma })`$ and the one forms $`E^{\underset{¯}{A}}dZ^{\underset{¯}{\mathrm{\Pi }}}E_{\underset{¯}{M}}^{\underset{¯}{A}}`$ contain the supervielbeins $`E_{\underset{¯}{M}}^{\underset{¯}{A}}`$ of the background superspace and have the form
$$E_{\underset{¯}{M}}{}_{}{}^{\underset{¯}{A}}=\left(\begin{array}{cc}\frac{1}{2}(\delta _{\underset{¯}{\gamma }}^{\underset{¯}{\alpha }}\delta _{\underset{¯}{\delta }}^{\underset{¯}{\beta }}+\delta _{\underset{¯}{\gamma }}^{\underset{¯}{\beta }}\delta _{\underset{¯}{\delta }}^{\underset{¯}{\alpha }})& 0\\ \frac{1}{4}(\delta _{\underset{¯}{\gamma }}^{\underset{¯}{\alpha }}\mathrm{\Theta }^{\underset{¯}{\beta }}+\delta _{\underset{¯}{\gamma }}^{\underset{¯}{\beta }}\mathrm{\Theta }^{\underset{¯}{\alpha }})& \delta _{\underset{¯}{\gamma }}^{\underset{¯}{\alpha }}\end{array}\right)$$
$`(3.23)`$
Defining $`e^{\varphi ^{}R^{}}(K_{\underset{¯}{A}})e^{\varphi ^{}R^{}}\mathrm{\Phi }_{\underset{¯}{A}}{}_{}{}^{\underset{¯}{B}}K_{\underset{¯}{B}}^{}`$ we find that
$$F^{\underset{¯}{A}}=E^{\underset{¯}{B}}\mathrm{\Phi }_{\underset{¯}{B}}^{\underset{¯}{A}}$$
$`(3.24)`$
The dynamics of the non-linear theory should be given by imposing the constraints of equations (3.12) to (3.14) where the derivatives are now defined for the full theory. In the next section we will argue that these constraints correctly encode the non-linear dynamics of the brane. It would be interesting to examine the implications of these constraints in detail and to see in particular what they imply for all the Goldstone fields associated with the automorphism. For those superbranes that have a straightforward action, it is possible that the supervielbein on the superbrane that arises in the non-linear realization will play an essential role in defining the action as it does for the case of the bosonic brane considered in section one. 4. Relation to the Superembedding Formalism In the superembedding approach we consider one supermanifold $`M`$ embedded in another $`\underset{¯}{M}`$. The preferred vector fields on the former are given by $`E_A=E_A{}_{}{}^{M}_{M}^{}`$, where $`E_A^M`$ are the inverse supervielbein, while those on the latter are $`E_{\underset{¯}{A}}=E_{\underset{¯}{A}}{}_{}{}^{\underset{¯}{M}}_{\underset{¯}{M}}^{}`$. The fields on $`M`$ must point somewhere in $`\underset{¯}{M}`$ as is expressed by the equation
$$E_A{}_{}{}^{M}_{M}^{}=E_A{}_{}{}^{\underset{¯}{A}}E_{\underset{¯}{A}}^{}{}_{}{}^{\underset{¯}{M}}_{\underset{¯}{M}}^{}.$$
$`(4.1)`$
The assumption in the superembedding approach is that
$$E_\alpha {}_{}{}^{\underset{¯}{a}}=0.$$
$`(4.2)`$
This states that the Grassmann odd preferred vector fields in $`M`$ can be expressed in terms of only the Grassmann odd vector fields in $`\underset{¯}{M}`$. Although the elegant simplicity of this assumption is apparent, its deeper meaning is still unclear. For many branes, and certainly many of the most interesting ones, this single assumption is sufficient to put the theory on-shell and determine its complete equations of motion . This is the case for the twobrane and fivebrane of M theory.
We may re-express equation (4.1) as
$$dz^ME_M{}_{}{}^{A}E_{A}^{}{}_{}{}^{\underset{¯}{B}}=dz^M_MZ^{\underset{¯}{\mathrm{\Lambda }}}E_{\underset{¯}{\mathrm{\Lambda }}}^{\underset{¯}{B}}$$
$`(4.3)`$
Multiplying by $`\mathrm{\Phi }_{\underset{¯}{B}}^{\underset{¯}{C}}`$, identifying the supervielbien on the embedded super manifold with the Cartan form in the $`K_A`$ directions, i.e. $`E_M{}_{}{}^{A}=F_M^A`$, and recognizing $`F^{\underset{¯}{A}}`$ from equation (3.22) we conclude that
$$F_A{}_{}{}^{\underset{¯}{B}}E_A{}_{}{}^{M}F_{M}^{}{}_{}{}^{\underset{¯}{B}}=_AZ^{\underset{¯}{B}}=E_A{}_{}{}^{\underset{¯}{C}}\mathrm{\Phi }_{\underset{¯}{C}}^{}^{\underset{¯}{B}}$$
$`(4.4)`$
The $`\mathrm{\Phi }`$ defined above equation (3.24) are constructed from the Goldstone fields associated with the automorphism group. They form an invertible matrix and also must have indices such that they are always Grassmann even; implying for example that $`\mathrm{\Phi }_{\underset{¯}{\beta }}{}_{}{}^{\underset{¯}{b}}=0`$. It follows from equation (4.4) that the inverse Higgs condition
$$F_\alpha ^{\underset{¯}{a}}=0,$$
$`(4.5)`$
is the necessary and sufficient condition for the embedding condition of equation (4.2). However, the superembedding formalism uses superfields that do not depend on the central charges other than the usual coordinates of Minkowski space. Hence, provided we adopt the inverse Higgs conditions of equations (3.12), (3.13) and (3.15) we should find that they describe the dynamics of the full non-linear theory. This follows because equation (3.13) is just the condition of equation (4..5), equation (3.14) eliminates the dependence of the superfields on the coordinates associated with the central charges and equation (3.12) identifies the automorphism Goldstone fields in terms of the fields of the brane. We note that the first two inverse Higgs conditions can be written as $`F_\alpha {}_{}{}^{\underset{¯}{A}}=\delta _\alpha ^{\underset{¯}{A}}`$.
The relationship between the inverse Higgs mechanism and the superembedding condition has been previously discussed in the context of superbranes in lower dimensions and their construction as non-linear realizations , . However, in these cases the dynamics of the branes are not determined by the embedding condition of equation (4.5). It would be interesting to apply the methods developed in this paper to branes of this type.
Identifying the supervielbien $`F_A^M`$ in the non-linear realization approach with the supervielbein of the superembedding approach implies that $`F_A{}_{}{}^{B}=\delta _A^B`$ and, as a consequence, $`E_A{}_{}{}^{\underset{¯}{B}}\mathrm{\Phi }_{\underset{¯}{B}}^{}{}_{}{}^{B}=\delta _A^B`$. Taking fermonic indices, and including the above condition of equation (4.5), we conclude that $`E_\alpha {}_{}{}^{\underset{¯}{\beta }}\mathrm{\Phi }_{\underset{¯}{\beta }}^{}{}_{}{}^{\gamma }=\delta _\alpha ^\gamma `$. We can however choose a set of frames normal to the superbrane. By expressing these vector fields in terms of the preferred frames of the background superspace, we can define an extension of the $`E_A^{\underset{¯}{B}}`$ to $`E_{\underset{¯}{A}}^{\underset{¯}{B}}`$. Making this extension in an appropriate way we conclude that $`E_{\underset{¯}{A}}^{\underset{¯}{B}}`$ is the inverse of the $`\mathrm{\Phi }_{\underset{¯}{A}}^{\underset{¯}{B}}`$. Thus the Goldstone fields associated with the automorphism group specify the relation between preferred frames associated with the branes and those in the background superspace.
Although the superfields do not in the end depend on the central charges, without the central charges in the original algebra one would not have the large automorphism group which is essential for the theory to contain the fields of the brane. This is apparent as a consequence of the discussion of the above paragraph, however, another way to observe this fact is to note that the supervierbeins on the embedded manifold would be essentially trivial if we had not included the Goldstone fields for the automorphism group. 5. Conclusion We have shown that the branes of M theory and, by reduction, those in ten dimensions can be described by the theory of non-linear realization provided we use a superalgebra that has the large automorphism group considered in . Indeed, the field strengths of the worldvolume fields of the brane arise as the Goldstone bosons for the automorphisms. It is a widely held belief that branes correspond to the low energy effects of a defect in space-time the breaks certain symmetries. We have shown in this paper that their dynamics is essentially determined by a knowledge of which symmetries are broken.
In reference it was explained how the fields of the branes arise as zero modes of the the corresponding background supergravity theory. It would be interesting understand the connections between that work and that of this paper. It would also be interesting to find if there are connections between the non-linear realization proposed in this paper and the coset consider in reference which was based on the Osp groups.
The necessity of including at least parts of the GL(32) automorphism group in the construction of the non-linear realization of the branes of M theory supports the idea proposed in that this symmetry may play an important role in M theory. Acknowledgment The author would like to thank Jon Bagger for a number of useful discussions at an early stage of this work and the hospitality of the Ecole Normale Superieure, where some of this work was carried out. This work was supported in part by the EU network on Integrability, Non-perturbative effects, and Symmetry in Quantum Field theory (FMRX-CT96-0012).
References
1 W.Siegel, Phys. Lett. B128 (1983) 397.
2 E. Bergshoeff, E. Sezgin and P. K. Townsend, Phys. Lett. 189B (1987) 75; Ann. of Phys. 185 1988 330.
3 M. Cederwall, A. von Gussich, B. E. W. Nilsson, A. Westerberg, Nucl.Phys. B490 1997 163; M. Cederwall, A. von Gussich, B. E. W. Nilsson, P. Sundell and A. Westerberg, Nucl.Phys. B490 (1997) 179; M. Aganagic, C. Popescu, J.H. Schwarz, Phys. Lett. B393 1997 311; E. Bergshoeff, P.K. Townsend, Nucl.Phys. B490 (1997) 145; M. Aganagic, J. Park, C. Popescu and J. H. Schwarz, Nucl. Phys. B496 (1997) 191.
4 P. S. Howe, E. Sezgin, and P. C. West. Covariant field equations of the M theory five-brane. Phys. Lett., B399:49–59, 1997. hep-th/9702008.
5 I. Bandos, K. Lechner, A. Nurmagambetov, P. Pasti, D. Sorokin and M. Tonin, Phys. Rev. Lett. 78 (1997) 4332; I. Bandos, K. Lechner, A. Nurmagambetov, P. Pasti, D. Sorokin and M. Tonin, Phys. Lett. B408 (1997) 135.
6 M. Perry and J. H. Schwarz, Nucl. Phys. 498 (1997) 47; J. H. Schwarz, Phys. Lett. B395 (1997) 191; Mina Aganagic, Jaemo Park, Costin Popescu, John H. Schwarz, World-Volume Action of the M Theory Five-Brane, hep-th/9701166, Nucl.Phys. B496 (1997) 191-214
7 For two reviews see P. S. Howe, E. Sezgin and P. C. West, Aspects of Superembeddings, D. V. Volkov Memorial Volume, Lecture notes in physics , Vol. 509, p. 64, Springer–Verlag, Berlin, Heidelberg 1998. (1988) 709; and D. Sorotkin, Superbranes and Superembeddings, hep-th/9906142.
8 J. Hughes, J. Polchinski, Nucl. Phys. B 278 (1986) 147; J. Hughes, J. Liu, J. Polchinski, Phys. Lett. B 180 (1986) 370.
9 S. Coleman, J. Wess and B. Zumino, Phys. Rev. 177 (1969) 2239; K. Callan, S. Coleman, J. Wess and B. Zumino, Phys. Rev. 177 (1969) 2247.
10 D. V. Volkov, Sov. J. Part. Nucl. 4 (1973) 3; D. V. Volkov and V. P. Akulov, JETP Letters 16 (1972) 438; Phys. Lett. B46 (1973) 109.
11 J. Bagger, A. Galperin, Phys. Lett. B 336 (1994) 25; Phys. Rev. D 55 (1997) 1091; Phys. Lett. B 412 (1997) 296.
12 F. Gonzalez-Rey, I.Y. Park, M. Roček, Nucl. Phys. B 544 (1999) 243; M. Roček, A. Tseytlin, Phys. Rev. D 59 (1999) 106001. S. Bellucci, E. Ivanov, S. Krivonos, Phys. Lett. B 460 (1999) 348; E. Ivanov, S. Krivonos, Phys. Lett. B 453 (1999) 237.
13 S. Belucci, E. Ivanov and S. Krivonos, Partial breaking $`N=4`$ to $`N=2`$: hypermultiplet as a Goldstone superfield, hep–th/9809190. S. Belucci, E. Ivanov and S. Krivonos, Partial breaking of $`N=1`$, $`D=10`$ supersymmetry, hep–th/9811244; F. Gonzalez–Rey, I. Y. Park and M. Roc̃ek, Nucl. Phys. B544 (1999) 243; M. Roc̃ek and A. Tseytlin, Phys. Rev. D59 (1999) 106001; E. Ivanov and S. Krivonos, Phys. Lett. B453 (1999) 237; S. V. Ketov, Mod. Phys. Lett. A14 (1999) 501; Born-Infeld-Goldstone superfield actions for gauge-fixed D-5- and D-3-branes in 6d, hep-th/9812051.
14 T. Adawi, M. Cederwall, U. Gran, M. Holm and B. E. W. Nilsson, Int. J. Mod. Phys. A13 (1998) 4691.
15 O. Barwald and P. West, Brane Rotating symmetries and the fivebrane equations of motion hep-th/9912226.
16 E. A. Ivanov and V. I. Ogievetsky, Teor. Mat. Fiz. 25 (1975) 164.
17 P. S. Howe and E. Sezgin, Phys. Lett. B390 (1997) 133; P. S. Howe and E. Sezgin, Phys. Lett. B394 (1997) 62.
18 D. Sorokin and P. K. Townsend. M theory superalgebra from the M five-brane.; Phys. Lett., B412 (1997) 265, hep-th/9708003.
19 T. Adawi, M. Cederwall, U. Gran, B. E. W. Nilsson and B. Razaznejad, JHEP 9902 (1999) 001.
20 E. Ivanov, S. Krivonos, N=1 D=4 supermembrane in the coset approach Phys.Lett. B453 (1999) 237, hep-th/9901003.
21 F. Delduc, E. Ivanov, S. Krivonos, 1/4 PBGS and Superparticle Actions, hep-th/9912292; F Delduc, E. Ivanov, S. Krivonos, 1/4 Partial Breaking of Global Supersymmetry and New Superparticle Actions, hep-th/9912222
22 Igor Bandos, Jerzy Lukierski, Dmitri Sorokin, The OSp(1/4) Superparticle and Exotic BPS States, hep-th/9912264 ; Paolo Pasti, Dmitri Sorokin, Mario Tonin, Branes in Super-AdS Backgrounds and Superconformal Theories, hep-th/9912076 ; Igor Bandos, Kharkov, Ukraine, Jerzy Lukiersk, Dmitri Sorokin, Generalized Superconformal Symmetries and Supertwistor Dynamimics, hep-th/9912051
23 J. A. de Azcarraga, J. P. Gauntlett, J. M. Izquierdo and P. K. Townsend; Topological extensions of the superalgebra for extended objects; Phys. Rev. Lett., B63 (1989) 2443.
24 J. P. Gauntlett, K,Itoh and P. K. Townsend, Superparticle with extrinsic curvature, Phys. Lett. B (1990) 65; J. P. Gauntlett, J. Gomis and P. K. Townsend, Particle actions as Wess-Zumino terms for space-time (super)symmetry groups, Phys. Lett. B (1990) 255. |
warning/0001/hep-th0001002.html | ar5iv | text | # 1 The scalar potential () as a function of ϕ for the 𝐷=5 case. The structure of the potential for other dimensions is similar.
CTP TAMU-01/00
UPR-868-T
hep-th/0001002
December 1999
Domain Walls and Massive Gauged Supergravity Potentials
M. Cvetič$`^1`$, H. Lü$`^1`$ and C.N. Pope$`^2`$
$`^{}`$Department of Physics and Astronomy
University of Pennsylvania, Philadelphia, Pennsylvania 19104
$`^{}`$Center for Theoretical Physics
Texas A&M University, College Station, Texas 77843
ABSTRACT
We point out that massive gauged supergravity potentials, for example those arising due to the massive breathing mode of sphere reductions in M-theory or string theory, allow for supersymmetric (static) domain wall solutions which are a hybrid of a Randall-Sundrum domain wall on one side, and a dilatonic domain wall with a run-away dilaton on the other side. On the anti-de Sitter (AdS) side, these walls have a repulsive gravity with an asymptotic region corresponding to the Cauchy horizon, while on the other side the runaway dilaton approaches the weak coupling regime and a non-singular attractive gravity, with the asymptotic region corresponding to the boundary of space-time. We contrast these results with the situation for gauged supergravity potentials for massless scalar modes, whose supersymmetric AdS extrema are generically maxima, and there the asymptotic regime transverse to the wall corresponds to the boundary of the AdS space-time. We also comment on the possibility that the massive breathing mode may, in the case of fundamental domain-wall sources, stabilize such walls via a Goldberger-Wise mechanism.
$`^1`$ Research supported in part by DOE grant DE-FG02-95ER40893
$`^2`$ Research supported in part by DOE grant DE-FG03-95ER40917.
The scalar potentials of gauged supergravity theories provide a natural testing-ground for studying domain-wall configurations within the framework of a basic theory. In general, such scalar potentials have isolated supersymmetric extrema with a negative cosmological constant. Within the AdS/CFT correspondence, these supersymmetric (BPS) domain walls play a role in elucidating the renormalization group flows and bound-state spectra of strongly coupled gauge theories (see, for example, - and references therein). A typical feature of gauged supergravity potentials is such that the supersymmetric extrema are maxima of the potential. The domain walls are therefore typically those with negative tension, and the metric transverse to the wall asymptotically ($`z\mathrm{}`$) approaches the boundary of the AdS space-time . Another feature of these solutions is that the region near the wall ($`z0`$) is in general singular; both the scalar field and the curvature generically exhibit singular behavior and thus the continuation across the wall region on the side $`z<0`$ involves (within the effective theory) a continuation across a singular domain-wall regime (c.f. ).
On the other hand, in recent months there has been a resurgence in the study of domain walls in asymptotically AdS space-times in $`D=5`$ gravity theories. For special examples of such static domain walls, the gravity effects transverse to the wall are suppressed, which has interesting implications for the phenomenology of the world on the brane. (See, for example, - and references therein.) Non-static walls in $`D=5`$ were also recently considered. (See - and references therein.) <sup>1</sup><sup>1</sup>1It turns out that the local and global space-time structure of vacuum domain walls ($`(D2)`$-brane configurations) in $`D`$ dimensions is universal, and thus the previous studies of domain walls in $`D=4`$ (see, and references therein) are completely parallel to the domain-wall solutions in any other dimension $`D`$.
A particular focus is on infinitely thin, static, $`Z_2`$-symmetric domain-wall solutions, constructed in a pure AdS gravity theory (the Randall-Sundrum scenario).<sup>2</sup><sup>2</sup>2Another proposal for the origin of the five-dimensional domain wall was made in , which is dilatonic and can be viewed as M5-branes wrapped arond the two-cycles of a Calabi-Yau manifold. (Generalizations that incorporate the effects of additional compactified dimensions were given in .) These solutions have a repulsive gravity , for which the asymptotic regions ($`z\pm \mathrm{}`$) corresponding Cauchy horizons . They satisfy a specific relation between the domain-wall tension $`\sigma `$ and the cosmological constant $`\mathrm{\Lambda }`$ of the AdS vacuum; this latter condition was subsequently shown to be a consequence of supersymmetry. (These results are again completely parallel with supersymmetric domain walls of N=1 supergravity theories in $`D=4`$.) These types of wall are of Type II in the classification scheme of refs. .
The main motivation of this paper is to provide a framework within gauged supergravity theories that has a chance of implementing the Randall-Sundrum scenario(s). As mentioned above, gauged supergravity theories tend to have potentials for the massless scalar modes that have isolated supersymmetric maxima and not minima. Thus the supersymmetric domain walls have negative tension (whose magnitude is the same as the tension of Type II walls). They have attractive gravity transverse to the wall, with the asymptotic regions ($`z\pm \mathrm{}`$) corresponding to AdS space-time boundaries . These types of walls are referred to as Type IV walls and are complementary to Type II walls.
In order to obtain Type II domain-wall solutions of the Randall-Sundrum scenario, the gauged supergravity potential would have to have two isolated supersymmetric minima. Since the potentials for the massless scalar fields in a gauged supergravity do not seem to have this feature, we turn in our analysis to include other scalar fields that do not lie in the massless supermultiplet.
We shall focus on the special classes of gauged supergravities that arise from sphere reductions of M-theory or string theory, with particular emphasis on the $`D=5`$ case. For examples in the Kaluza-Klein reduction of Type IIB string theory on a five-sphere ($`S^5`$), there will be an infinite tower of massive supermultiplets in addition to the massless multiplet, and so one could consider the potentials for one or more of the massive scalar fields. In general, one cannot focus attention on a single such field in isolation, on account of its couplings to other fields. However, in certain special cases a consistent truncation to a single massive scalar can be performed. One such example is the “breathing mode” that parameterises the overall volume of the compactifying $`S^5`$. (Unlike the breathing mode in a toroidal reduction, which is massless, the breathing mode in a spherical reduction is a member of a massive supermultiplet.)
The scalar potentials for the breathing-mode scalars in various Kaluza-Klein spherical reductions were studied in . Although the breathing mode is a member of a massive multiplet, the truncation is nonetheless consistent since it is a singlet under the isometry group of the internal sphere. (It would not in general be consistent to turn on a finite subset of other fields as well.)
The resulting $`D`$-dimensional Lagrangians all turn out to have the following form:
$$_D=eR\frac{1}{2}e(\varphi )^2eV,$$
(1)
where the potential is given by
$$V=\frac{1}{2}g^2(\frac{1}{a_1^2}e^{a_1\varphi }\frac{1}{a_1a_2}e^{a_2\varphi }).$$
(2)
The positive constants $`a_1`$ and $`a_2`$ are given by
$$a_1^2=\frac{4}{N}+\frac{2(D1)}{D2},a_1a_2=\frac{2(D1)}{D2},$$
(3)
where $`N`$ is a certain positive integer. For $`D=4`$, 7 and 5, this integer takes the value $`N=1`$. These cases correspond to the $`S^7`$ and $`S^4`$ reductions of $`D=11`$ supergravity, and the $`S^5`$ reduction of type IIB supergravity respectively. For $`D=3`$ the integer $`N`$ can be equal to 1, 2 or 3, corresponding to the $`S^1`$ reduction of the Freedman-Schwarz model,<sup>3</sup><sup>3</sup>3The reduction in this case is of a generalised Scherk-Schwarz type, where the axion is allowed a linear dependence on the reduction coordinate. the $`S^3`$ reduction of $`D=6`$ simple (chiral) supergravity, and the $`S^2`$ reduction of $`D=5`$ simple supergravity respectively. The explicit dilaton coupling constants $`a_1`$ and $`a_2`$ for the above cases are given in Table 1.
| D | N | $`a_1`$ | $`a_2`$ | $`\frac{a_1}{a_2(D1)}`$ |
| --- | --- | --- | --- | --- |
| 7 | 1 | $`4\sqrt{\frac{2}{5}}`$ | $`\frac{3}{\sqrt{10}}`$ | $`\frac{4}{9}`$ |
| 5 | 1 | $`2\sqrt{\frac{5}{3}}`$ | $`\frac{4}{\sqrt{15}}`$ | $`\frac{5}{8}`$ |
| 4 | 1 | $`\sqrt{7}`$ | $`\frac{3}{\sqrt{7}}`$ | $`\frac{7}{9}`$ |
| 3 | 1 | $`2\sqrt{2}`$ | $`\sqrt{2}`$ | $`1`$ |
| 3 | 2 | $`\sqrt{6}`$ | $`2\sqrt{\frac{2}{3}}`$ | $`\frac{3}{4}`$ |
| 3 | 3 | $`\frac{4}{\sqrt{3}}`$ | $`\sqrt{3}`$ | $`\frac{2}{3}`$ |
Table 1: The values of the parameters $`N`$, $`a_1`$ and $`a_2`$ in diverse dimensions $`D`$
xxxxxxxxxxxxx that enter the scalar potential (2).
Since $`a_1>a_2>0`$, the potential has a minimum at $`\varphi =0`$, with
$$V_{\mathrm{min}}=\frac{g^2(D2)}{N(D1)a_1^2}.$$
(4)
See Figure 1 for the shape of the potential. The potential can be expressed in terms of a “superpotential” $`W_+`$ or $`W_{}`$ as follows:
$$V=(\frac{W_\pm }{\varphi })^2\frac{D1}{2(D2)}W_\pm ^2,$$
(5)
where
$$W_\pm =\sqrt{\frac{N}{2}}g(\frac{1}{a_1}e^{a_1\varphi /2}\pm \frac{1}{a_2}e^{a_2\varphi /2}).$$
(6)
Let us now consider the following ansatz for a domain-wall metric:
$$ds^2=e^{2A}dx^\mu dx^\nu \eta _{\mu \nu }+dz^2.$$
(7)
The equations of motion are
$`\varphi ^{\prime \prime }+(D1)A^{}\varphi ^{}={\displaystyle \frac{V}{\varphi }}`$
$`A^{\prime \prime }+(D1)(A^{})^2=(D1)A^{\prime \prime }+(D1)(A^{})^2+\frac{1}{2}(\varphi ^{})^2={\displaystyle \frac{V}{D2}}.`$ (8)
These admit a first integral, given by
$$\varphi ^{}=\sqrt{2}\frac{W_\pm }{\varphi },A^{}=\frac{1}{\sqrt{2}(D2)}W_\pm .$$
(9)
Here, we shall consider the choice $`W_{}`$ for the superpotential, since it has a supersymmetric minimum, i.e. $`_\varphi W_{}=0`$, at $`\varphi =0`$. From (9), we shall therefore have
$`\varphi ^{}`$ $`=`$ $`\sqrt{{\displaystyle \frac{N}{4}}}g(e^{a_1\varphi /2}e^{a_2\varphi /2}),`$
$`A^{}`$ $`=`$ $`={\displaystyle \frac{g\sqrt{N}}{2(D2)}}[{\displaystyle \frac{1}{a_1}}e^{a_1\varphi /2}{\displaystyle \frac{1}{a_2}}e^{a_2\varphi /2}].`$ (10)
Solving for $`\varphi `$ and $`A`$, we find that $`A`$ can be expressed as a function of $`\varphi `$, namely
$$e^{(D1)A}=c\frac{W}{\varphi }e^{\frac{1}{2}(a_1+a_2)\varphi },$$
(11)
where $`c`$ is an integration constant. For $`D>3`$, the solution for $`\varphi `$ is given by
$$zz_0=\frac{4}{a_2g\sqrt{N}}e^{\frac{1}{2}a_2\varphi }{}_{2}{}^{}F_{1}^{}[\frac{a_2}{a_2a_1},1,1+\frac{a_2}{a_2a_1};e^{\frac{1}{2}(a_1a_2)\varphi }].$$
(12)
(For our specific examples mentioned above, we shall have $`N=1`$ and $`D=4`$, 5 or 7.) For $`D=3`$, we find that $`\varphi `$ is given by
$`N=1:`$ $`zz_0={\displaystyle \frac{\sqrt{8}}{g}}\left(e^{\frac{1}{\sqrt{2}}\varphi }+\mathrm{log}(e^{\frac{1}{\sqrt{2}}\varphi }1)\right),`$
$`N=2:`$ $`zz_0={\displaystyle \frac{\sqrt{3}}{g}}(e^{\sqrt{\frac{2}{3}}\varphi }++2e^{\frac{1}{\sqrt{6}}\varphi }+2\mathrm{log}(e^{\frac{1}{\sqrt{6}}\varphi }1)),`$ (13)
$`N=3:`$ $`zz_0={\displaystyle \frac{2}{g}}\left(e^{\frac{1}{\sqrt{3}}\varphi }+\frac{2}{3}e^{\frac{\sqrt{3}}{2}\varphi }+2e^{\frac{1}{2\sqrt{3}}\varphi }+2\mathrm{log}(e^{\frac{1}{2\sqrt{3}}\varphi }1)\right),`$
(The analogous solutions constructed using $`W_+`$ rather than $`W_{}`$ can also be easily obtained, but they seem not to be directly relevant for our present purposes.) These supersymmetric domain walls in a different coordinate system were given in , where their higher dimensional origins as M-branes and D3-branes were discussed. Note that the hypergeometric function $`{}_{2}{}^{}F_{1}^{}[a,1,1+a,x]`$ appearing in (12) is the Lerch transcendent $`a\mathrm{\Phi }(x,1,a)`$. In fact, the solutions for $`D>3`$ and for $`D=3`$ can all be given by a single formula using the Lerch transcendent, namely
$$zz_0=\frac{a_1\sqrt{N}}{g}e^{\frac{1}{2}a_2\varphi }\mathrm{\Phi }(e^{\frac{1}{2}(a_1a_2)\varphi },1,\frac{a_2}{a_2a_1}).$$
(14)
The $`W_{}`$ solutions above all have two different branches. In one branch, $`\varphi `$ runs from 0 to $`+\mathrm{}`$, with $`z`$ runnning from $`z=\mathrm{}`$ to $`z=0`$, where we have chosen the integration constant $`z_0`$ to be
$`D>3:`$ $`z_0={\displaystyle \frac{\pi a_1\sqrt{N}}{g}}\left(\mathrm{i}+\mathrm{cot}({\displaystyle \frac{\pi a_1}{a_1a_2}})\right),`$
$`D=3:`$ $`z_0={\displaystyle \frac{\pi a_1\sqrt{N}}{g}}.`$ (15)
(The imaginary part cancels the imaginary additive constant in the expressions (12) and (13).) When $`\varphi `$ is large, the solution takes the form
$`e^{\frac{1}{2}a_1\varphi }`$ $``$ $`\frac{1}{4}a_1\sqrt{N}gz,`$
$`e^{(D1)A}`$ $``$ $`c\sqrt{{\displaystyle \frac{N}{8}}}ge^{\frac{1}{2}a_2\varphi }c\sqrt{{\displaystyle \frac{N}{8}}}g\left(\frac{1}{4}a_1\sqrt{N}gz\right)^{\frac{a_2}{a_1}}.`$ (16)
In this branch, when the coordinate $`z`$ reaches its limit at $`z=0`$, the factor $`e^{2A}`$ in the metric therefore goes to zero, and there is a power-law naked curvature singularity. (Note that in this regime the solution extends into large positive values of the potential (2) with a large cost to the energy density of the wall, and it thus terminates at a finite value of the transverse coordinate.)
As $`z`$ approaches $`\mathrm{}`$, the functions $`\varphi `$ and $`A`$ become
$$\varphi e^{\frac{g}{a_1\sqrt{N}}z},A\frac{g}{a_1\sqrt{N}(D1)}z.$$
(17)
The metric asymptotically approaches the AdS space-time, described in horospherical coordinates with $`z\mathrm{}`$ corresponding to the Cauchy horizon . Note that on that side of the wall the gravity is repulsive and provides “one half” of the Randall-Sundrum wall.
In the second branch, $`\varphi `$ runs from 0 to $`\mathrm{}`$, while $`z`$ runs from $`z=\mathrm{}`$ to $`z=+\mathrm{}`$. The behaviour of the solution near $`z=\mathrm{}`$ is the same as in the branch discussed previously, with the metric approaching asymptotically AdS. As $`z`$ approaches $`+\mathrm{}`$, the solution becomes
$`e^{\frac{1}{2}a_2\varphi }`$ $``$ $`\frac{1}{4}a_2\sqrt{N}gz,`$
$`e^{(D1)A}`$ $``$ $`c\sqrt{{\displaystyle \frac{N}{8}}}ge^{\frac{1}{2}a_1\varphi }c\sqrt{{\displaystyle \frac{N}{8}}}g\left(\frac{1}{4}a_2\sqrt{N}gz\right)^{\frac{a_1}{a_2}}.`$ (18)
(The constant $`c`$ is negative in this case.) This side describes one-side of a supersymmetric dilatonic domain wall . Interestingly, it has no curvature singularity; as $`z`$ tends to $`+\mathrm{}`$ the curvature falls off as $`1/z^2`$, while the diverging dilaton $`\varphi \mathrm{}`$ approaches the weak coupling limit. Gravity on this side is attactive and for the null geodesics the affine parameter $`\tau `$ is infinite. Namely, $`\tau ^+\mathrm{}e^Adzz^{1\frac{a_1}{a_2(D1)}}|^+\mathrm{}`$. Since for all the cases under consideration the ratio $`\frac{a_1}{a_2(D1)}1`$ (see Table 1), $`\tau `$ is indeed infinite and $`z+\mathrm{}`$ corresponds to the boundary of the space-time. ($`D=3`$, $`N=1`$ case is borderline with the affine parameter diverging logarithmically.)
For the purpose of constructing a domain-wall universe, it is the second of the two branches that is relevant. Thus this solution is a hybrid of the Type II vacuum domain wall and the dilatonic wall. The thickness of the wall is of the order of $`1/g`$. It is a non-singular solution, with repulsive gravity on the AdS-side ($`z<0`$) and attractive gravity one on the dilatonic-side. While $`z=\mathrm{}`$ corresponds to the AdS Cauchy horizons, $`z+\mathrm{}`$ is the time-like boundary of the space-time. The solutions for both the metric coefficient $`e^{2A}`$ and the breathing-mode scalar $`\varphi `$, as functions of $`z`$, are sketched in Figure 2 below.
Thus within a pure field-theoretic framework, i.e. employing only the breathing-mode scalar field to construct the domain wall solution, we have been only partially successful; the massive gauged supergravity potential gave us one “supersymmetric AdS minimum” and another “run-away vacuum”, thus yielding a hybrid domain wall solution, and not the pure Type II vacuum domain wall that we were really after. Somewhat disappointing is the fact that on the dilatonic domain wall side gravity is attractive, and thus these domain walls cannot provide a phenomenologically viable scenario with a large transverse direction $`z=\{\mathrm{},+\mathrm{}\}`$; only the domain $`z<0`$ can be taken large.
We may also explore another possibility, by adding a singular domain-wall source to this potential. The breathing-mode potential then provides a framework for implementing the Goldberger-Wise scenario . In this case, in the second branch the diverging behaviour of the dilaton is cancelled by a delta-function source for the domain wall at some finite value of $`z`$, say $`z=z_{}`$. (Note that the source tension has to precisely balance that of the scalar contribution at the wall .) Then, the solution for $`z>z_{}`$ can be replaced by a reflection of the solution for $`z<z_{}`$, so that
$$|zz_{}|+z_{}=\frac{4}{a_2g\sqrt{N}}e^{\frac{1}{2}a_2\varphi }{}_{2}{}^{}F_{1}^{}[\frac{a_2}{a_2a_1},1,1+\frac{a_2}{a_2a_1};e^{\frac{1}{2}(a_1a_2)\varphi }].$$
(19)
The metric function $`A`$ is again given by substituting $`\varphi `$ into (11). (The reason why a solution can be constructed in this way is because the original equations of motion (8) are invariant under $`zz`$, and $`zz+`$constant.) Since $`A`$ is continuous at $`z=z_{}`$, but its first derivative is not, it follows that there will be a delta-function curvature singularity there. This can be balanced by a domain-wall source term, in precisely the same way that one can balance the delta-function singularity on an electric string or $`p`$-brane soliton with an appropriate source term. The functions $`e^{2A}`$ and $`\varphi `$ as a function of $`z`$ for this solution are plotted in Figure 3 below.
To summarise, in this paper we set out to explore the possibility of finding a supersymmetric AdS domain-wall solution, relevant for $`D=5`$ for the Randall-Sundrum scenario, within massive gauged supergravity theories. By employing the potential for the massive breathing-mode scalar of the compactifying sphere in M-theory or string theory in diverse dimensions, we arrived at static (supersymmetric) domain walls which are of a hybrid type. On one side they correspond to the Randall-Sundrum wall with repulsive gravity, and on the other side they are supersymmetric dilatonic walls .
Although these supergravity solutions per se do not possess all of the features needed for a Randall-Sundrum scheme, one can obtain a more satisfactory result by including also a fundmental (singular) domain-wall source. The massive scalar mode acts as a modulus stabilizing the domain wall (as in the Goldberger-Wise scenario), and it provides a repulsive (AdS) gravity transverse to the wall, as required for implementing the Randall-Sundrum scenario. |
warning/0001/hep-th0001078.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Riemann normal coordinates (abbreviated as RNC from now on) have a vast field of applications to Physics. In particular they constitute the main ingredient of the background field method widely used for nonlinear $`\sigma `$\- model calculations in curved spacetime. They have the appealing feature that the geodesics passing through the origin have the same form as the equations of straight lines passing through the origin of a Cartesian system of coordinates in Euclidean geometry.
The purpose of this letter is to bridge the gap one confronts when attempts to perform computations that involve expansion of the metric tensor beyond the fourth order . We present the general method for breaking the barrier of the fourth order and at the same time we supply the reader with formulae adequate to reach the order of its choice.
The starting point will be a geodesic, on a compact n-dimensional Riemannian manifold, parametrized by $`x^l(s)`$ and satisfying the differential equation:
$`{\displaystyle \frac{d^2x^l}{ds^2}}+\mathrm{\Gamma }_{jk}^l(x){\displaystyle \frac{dx^j}{ds}}{\displaystyle \frac{dx^k}{ds}}=0`$ (1)
where s is the arc length and $`\mathrm{\Gamma }_{jk}^l`$ denotes the Christoffel symbol for the Levi-Civita connection. Any integral curve of (1) is completely determined by a point $`𝒫(x_0^1,x_0^2,\mathrm{},x_0^n)`$ and a direction defined by the tangent vector $`\xi ^l=\left(\frac{dx^l}{ds}\right)_𝒫`$. The power series solution of (1) is:
$`x^l(s)=x_0^l+{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{k!}}\left({\displaystyle \frac{d^kx^l}{ds^k}}\right)_𝒫s^k`$ (2)
and its coefficients can be replaced using successive differentiations by:
$`\left({\displaystyle \frac{dx^l}{ds}}\right)_𝒫`$ $`=`$ $`\xi ^l`$
$`\left({\displaystyle \frac{d^2x^l}{ds^2}}\right)_𝒫`$ $`=`$ $`\left(\mathrm{\Gamma }_{i_1i_2}^l\right)_P\xi ^{i_1}\xi ^{i_2}`$
$`\left({\displaystyle \frac{d^3x^l}{ds^3}}\right)_𝒫`$ $`=`$ $`\left[_{(i_1}\mathrm{\Gamma }_{i_2i_3)}^l2\mathrm{\Gamma }_{(i_1\rho }^l\mathrm{\Gamma }_{i_2i_3)}^\rho \right]_P\xi ^{i_1}\xi ^{i_2}\xi ^{i_3}`$
$`\left({\displaystyle \frac{d^4x^l}{ds^4}}\right)_P`$ $`=`$ $`[_{(i_1}_{i_2}\mathrm{\Gamma }_{i_3i_4)}^l_\rho \mathrm{\Gamma }_{(i_1i_2}^l\mathrm{\Gamma }_{i_3i_4)}^\rho 4_{(i_1}\mathrm{\Gamma }_{i_2\rho }^l\mathrm{\Gamma }_{i_3i_4)}^\rho 2_{(i_1}\mathrm{\Gamma }_{i_2i_3}^\rho \mathrm{\Gamma }_{\rho i_4)}^l`$
$`+`$ $`4\mathrm{\Gamma }_{k(i_1}^l\mathrm{\Gamma }_{i_2\rho }^k\mathrm{\Gamma }_{i_3i_4)}^\rho +2\mathrm{\Gamma }_{\rho \sigma }^l\mathrm{\Gamma }_{(i_1i_2}^\rho \mathrm{\Gamma }_{i_3i_4)}^\sigma ]_P\xi ^{i_1}\xi ^{i_2}\xi ^{i_3}\xi ^{i_4}.`$
$`\mathrm{}`$ (3)
where dots in (3) represent higher derivative terms at the point $`𝒫`$ and symmetrization exclusively relates the lower i indices <sup>*</sup><sup>*</sup>*Proof of the fourth expression in (3) would require the use of the identity $`\frac{dx^\mu }{ds}D_\mu \frac{dx^\nu }{ds}=0`$. Apparently the domain of convergence of (2) depends on the components $`g_{ij}`$ and the values of $`\xi ^i`$. However, for sufficiently small values of s it defines an integral curve of (1).
## 2 The computational procedure
In RNC the geodesics through $`𝒫`$ are straight lines defined by:
$`y^l=\xi ^ls.`$ (4)
This set of coordinates cannot be used to cover the whole manifold and is valid only for a small neighbourhood of the point $`𝒫`$ where the conditions of the existence and uniqueness theorem of differential equation of the geodesics are satisfied. In this region no two geodesics through $`𝒫`$ intersect due to one-one correspondence of $`x^l`$ and $`y^l`$. Incidently, this displays the local nature of RNC.
Substituting (4) into eq. (2) one has:
$`x^l=x_0^l+{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{k!}}\left(\mathrm{\Gamma }_{i_1i_2\mathrm{}i_k}^l\right)_𝒫y^{i_1}y^{i_2}\mathrm{}y^{i_k}`$ (5)
where $`\left(\mathrm{\Gamma }_{i_1i_2\mathrm{}i_k}^l\right)_𝒫`$ are the “generalized Christoffel symbols” at the point $`𝒫`$ used in eq. (3). The Jacobian $`\left|\frac{x^l}{y^m}\right|_𝒫0`$ and thus the series (5) can be inverted.
For this local system of coordinates the geodesic equation can be written as:
$`\overline{\mathrm{\Gamma }}_{ij}^l(y)y^iy^j=0`$ (6)
and the power series solution becomes:
$`y^l={\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{k!}}\left(\overline{\mathrm{\Gamma }}_{i_1i_2\mathrm{}i_k}^l\right)_𝒫\xi ^{i_1}\xi ^{i_2}\mathrm{}\xi ^{i_k}s^k.`$ (7)
Reduction of (7) to (4) for arbitrary values of $`\xi ^l`$ implies that at the origin $`𝒫`$ holds:
$`\overline{\mathrm{\Gamma }}_{(i_1i_2\mathrm{}i_k)}^l=0`$ (8)
or, by induction, one can easily prove that eq. (8) is equivalent to:
$`_{(i_1}_{i_2}\mathrm{}_{i_{k2}}\overline{\mathrm{\Gamma }}_{i_{k1}i_k)}^l=0`$ (9)
Paraphrasing eq. (9) means that all symmetric derivatives of the affine connection vanish at the origin in RNC.
Generally speaking, a covariant second rank tensor field on a manifold can be expanded according to:
$`T_{k_1k_2}(\stackrel{~}{\varphi })={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{n!}}\left[{\displaystyle \frac{}{\xi _{i_1}}}{\displaystyle \frac{}{\xi _{i_2}}}\mathrm{}{\displaystyle \frac{}{\xi _{i_n}}}\right]T_{k_1k_2}(\varphi )\xi _{i_1}\xi _{i_2}\mathrm{}\xi _{i_n}.`$ (10)
The coefficients of the Taylor expansion are tensors and can be expressed in terms of the components $`R_{mnp}^l`$ Our conventions are: $`R_{mnp}^l=_n\mathrm{\Gamma }_{mp}^l+\mathrm{\Gamma }_{mp}^k\mathrm{\Gamma }_{kn}^l(np)`$, $`R_{\mu \nu }=R_{\mu \alpha \nu }^\alpha `$ and $`R=R_\mu ^\mu `$. of the Riemann curvature tensor and the covariant derivatives $`D_kT_{lm}`$ and $`D_kR_{mnp}^l`$. One without much effort can prove that:
$`_{(i_1}_{i_2}\mathrm{}_{i_{n1}}\overline{\mathrm{\Gamma }}_{i_n)k}^l=`$ $``$ $`\left({\displaystyle \frac{n1}{n+1}}\right)[D_{(i_1}D_{i_2}\mathrm{}D_{i_{n2}}\overline{R}_{i_{n1}ki_n)}^l`$ (11)
$`+`$ $`_{(i_1}_{i_2}\mathrm{}_{i_{n2}}\left(\overline{\mathrm{\Gamma }}_{i_{n1}k}^\alpha \overline{\mathrm{\Gamma }}_{\alpha i_n)}^l\right)`$
$``$ $`_{(i_1}_{i_2}\mathrm{}_{i_{n3}}(\overline{\mathrm{\Gamma }}_{i_{n2}\alpha }^l\overline{R}_{i_{n1}ki_n)}^\alpha l\alpha ,\alpha k)`$
$``$ $`_{(i_1}_{i_2}\mathrm{}_{i_{n4}}(\overline{\mathrm{\Gamma }}_{i_{n3}\alpha }^lD_{i_{n2}}\overline{R}_{i_{n1}ki_n)}^\alpha l\alpha ,\alpha k)`$
$`\mathrm{}`$
$``$ $`(_{(i_1}\overline{\mathrm{\Gamma }}_{i_2\alpha }^lD_{i_3}\mathrm{}D_{i_{n2}}\overline{R}_{i_{n1}ki_n)}^\alpha l\alpha ,\alpha k)]`$
where the interchange of covariant and contravariant indices act independently. Expression (11) reproduces for various values of nIn there is a misprint for the $`n=4`$ case. A minus sign is needed in front of the $`\frac{2}{9}`$-term. the following results:
$`_{(i_1}\overline{\mathrm{\Gamma }}_{i_2)k}^l=`$ $`{\displaystyle \frac{1}{3}}\overline{R}_{(i_1i_2)k}^l`$
$`_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3)k}^l=`$ $``$ $`{\displaystyle \frac{1}{2}}D_{(i_1}\overline{R}_{i_2ki_3)}^l`$
$`_{(i_1}_{i_2}_{i_3}\overline{\mathrm{\Gamma }}_{i_4)k}^l=`$ $``$ $`{\displaystyle \frac{3}{5}}\left[D_{(i_1}D_{i_2}\overline{R}_{i_3ki_4)}^l+{\displaystyle \frac{2}{9}}\overline{R}_{(i_1i_2\alpha }^l\overline{R}_{i_3i_4)k}^\alpha \right]`$
$`_{(i_1}_{i_2}_{i_3}_{i_4}\overline{\mathrm{\Gamma }}_{i_5)k}^l=`$ $``$ $`{\displaystyle \frac{2}{3}}\left[D_{(i_1}D_{i_2}D_{i_3}\overline{R}_{i_4ki_5)}^lD_{(i_1}\overline{R}_{i_2ki_3}^\alpha \overline{R}_{i_4i_5)\alpha }^l\right]`$
$`_{(i_1}_{i_2}_{i_3}_{i_4}_{i_5}\overline{\mathrm{\Gamma }}_{i_6)k}^l=`$ $``$ $`{\displaystyle \frac{5}{7}}[D_{(i_1}\mathrm{}D_{i_4}\overline{R}_{i_5ki_6)}^l`$
$``$ $`{\displaystyle \frac{1}{5}}\left(7D_{(i_1}D_{i_2}\overline{R}_{i_3\alpha i_4}^l\overline{R}_{i_5i_6)k}^\alpha +D_{(i_1}D_{i_2}\overline{R}_{i_3ki_4}^\alpha \overline{R}_{i_5i_6)\alpha }^l\right)`$
$`+`$ $`{\displaystyle \frac{3}{2}}D_{(i_1}\overline{R}_{i_2ki_3}^\alpha D_{i_4}\overline{R}_{i_5\alpha i_6)}^l{\displaystyle \frac{16}{45}}\overline{R}_{(i_1i_2\alpha }^l\overline{R}_{i_3i_4\beta }^\alpha \overline{R}_{i_5i_6)k}^\beta ]`$
$`\mathrm{}`$ (12)
The coefficients of (10) can be rewritten as:
$`_{(i_1}_{i_2}\mathrm{}_{i_n)}\overline{T}_{k_1k_2}=`$ $`D_{(i_1}D_{i_2}\mathrm{}D_{i_n)}\overline{T}_{k_1k_2}`$ (13)
$`+`$ $`_{(i_1}_{i_2}\mathrm{}_{i_{n1})}[\overline{\mathrm{\Gamma }}_{i_nk_1)}^\alpha \overline{T}_{\alpha k_2}+k_1k_2]`$
$`+`$ $`_{(i_1}_{i_2}\mathrm{}_{i_{n2})}[\overline{\mathrm{\Gamma }}_{i_{n1}k_1}^\alpha D_{i_n)}\overline{T}_{\alpha k_2}+k_1k_2]`$
$`\mathrm{}`$
$`+`$ $`_{(i_1}_{i_2}[\overline{\mathrm{\Gamma }}_{i_3k_1}^\alpha D_{i_4}\mathrm{}D_{i_n)}\overline{T}_{\alpha k_2}+k_1k_2]`$
$`+`$ $`_{(i_1}[\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha D_{i_3}\mathrm{}D_{i_n)}\overline{T}_{\alpha k_2}+k_1k_2].`$
Expressions (11) and (13) compose the building blocks of the current recursive method which produces the following results for different values of n:
$`_{(i_1}_{i_2)}\overline{T}_{k_1k_2}`$ $`=`$ $`D_{(i_1}D_{i_2)}\overline{T}_{k_1k_2}{\displaystyle \frac{1}{3}}[\overline{R}_{(i_1k_1i_2)}^\rho \overline{T}_{\rho k_2}+k_1k_2]`$
$`_{(i_1}_{i_2}_{i_3)}\overline{T}_{k_1k_2}=`$ $`D_{(i_1}D_{i_2}D_{i_3)}\overline{T}_{k_1k_2}`$
$`+`$ $`\left[_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3)k_1}^\rho \overline{T}_{\rho k_2}+2_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\rho D_{i_3)}\overline{T}_{\rho k_2}+k_1k_2\right]`$
$``$ $`{\displaystyle \frac{1}{3}}(\overline{R}_{(i_1k_1i_2}^\rho D_{i_3)}\overline{T}_{\rho k_2}+k_1k_2)`$
$`_{(i_1}_{i_2}_{i_3}i_4)\overline{T}_{k_1k_2}=`$ $`D_{(i_1}D_{i_2}D_{i_3}D_{i_4)}\overline{T}_{k_1k_2}`$
$`+`$ $`[_{(i_1}_{i_2}_{i_3}\overline{\mathrm{\Gamma }}_{i_4)k_1}^\rho \overline{T}_{\rho k_2}+3_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3k_1}^\rho D_{i_4)}\overline{T}_{\rho k_2}`$
$`+`$ $`3_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\rho (D_{i_3}D_{i_4)}\overline{T}_{\rho k_2}{\displaystyle \frac{1}{3}}(\overline{R}_{i_3\rho i_4)}^\sigma \overline{T}_{\sigma k_2}+\rho k_2))+k_1k_2]`$
$`_{(i_1}_{i_2}_{i_3}i_4_{i_5)}\overline{T}_{k_1k_2}=`$ $`D_{(i_1}D_{i_2}D_{i_3}D_{i_4}D_{i_5)}\overline{T}_{k_1k_2}`$
$``$ $`[{\displaystyle \frac{10}{3}}\overline{R}_{(i_1k_1i_2}^\alpha D_{i_3}\mathrm{}D_{i_5)}\overline{T}_{\alpha k_2}+5D_{(i_1}\overline{R}_{i_2k_1i_3}^\alpha D_{i_4}D_{i_5)}\overline{T}_{\alpha k_2}`$
$`+`$ $`3D_{(i_1}D_{i_2}\overline{R}_{i_3k_1i_4}^\alpha D_{i_5)}\overline{T}_{\alpha k_2}+{\displaystyle \frac{2}{3}}D_{(i_1}D_{i_2}D_{i_3}\overline{R}_{i_4k_1i_5)}^\alpha \overline{T}_{\alpha k_2}`$
$``$ $`{\displaystyle \frac{2}{3}}D_{(i_1}\overline{R}_{i_2k_1i_3}^\rho \overline{R}_{i_4i_5)\rho }^\alpha \overline{T}_{\alpha k_2}+D_{(i_1}\overline{R}_{i_2k_1i_3}^\alpha (\overline{R}_{i_4\alpha i_5)}^\rho \overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`{\displaystyle \frac{2}{3}}(D_{(i_1}\overline{R}_{i_2\alpha i_3}^\rho \overline{T}_{\rho k_2}+\alpha k_2)\overline{R}_{i_4i_5)k_1}^\alpha \pm k_1k_2]`$
$`_{(i_1}_{i_2}_{i_3}i_4_{i_5}_{i_6)}\overline{T}_{k_1k_2}=`$ $`D_{(i_1}\mathrm{}D_{i_6)}\overline{T}_{k_1k_2}`$ (14)
$`+`$ $`[_{(i_1}\mathrm{}_{i_5}\overline{\mathrm{\Gamma }}_{i_6)k_1}^\alpha \overline{T}_{\alpha k_2}+6_{(i_1}\mathrm{}_{i_4}\overline{\mathrm{\Gamma }}_{i_5k_1}^\alpha D_{i_6)}\overline{T}_{\alpha k_2}`$
$`+`$ $`15_{(i_1}\mathrm{}_{i_3}\overline{\mathrm{\Gamma }}_{i_4k_1}^\alpha D_{i_5}D_{i_6)}\overline{T}_{\alpha k_2}+20_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3k_1}^\alpha D_{i_4}\mathrm{}D_{i_6)}\overline{T}_{\alpha k_2}`$
$`+`$ $`14_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha D_{i_3}\mathrm{}D_{i_6)}\overline{T}_{\alpha k_2}`$
$`+`$ $`10_{(i_1}\mathrm{}_{i_3}\overline{\mathrm{\Gamma }}_{i_4k_1}^\alpha (_{i_5}\overline{\mathrm{\Gamma }}_{i_6)\alpha }^\rho \overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`10_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3k_1}^\alpha (_{i_4}_{i_5}\overline{\mathrm{\Gamma }}_{i_6)\alpha }^\rho \overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`36_{(i_1}_{i_2}\overline{\mathrm{\Gamma }}_{i_3k_1}^\alpha (_{i_4}\overline{\mathrm{\Gamma }}_{i_5\alpha }^\rho D_{i_6)}\overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`5_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha (_{i_3}\mathrm{}_{i_5}\overline{\mathrm{\Gamma }}_{i_6)\alpha }^\rho \overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`24_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha (_{i_3}_{i_4}\overline{\mathrm{\Gamma }}_{i_5\alpha }^\rho D_{i_6)}\overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`45_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha (_{i_3}\overline{\mathrm{\Gamma }}_{i_4\alpha }^\rho D_{i_5}D_{i_6)}\overline{T}_{\rho k_2}+\alpha k_2)`$
$`+`$ $`15[_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha _{i_3}\overline{\mathrm{\Gamma }}_{i_4\alpha }^\rho (_{i_5}\overline{\mathrm{\Gamma }}_{i_6)\rho }^\sigma \overline{T}_{\sigma k_2}+\rho k_2)+\alpha k_2]`$
$`+`$ $`_{(i_1}\overline{\mathrm{\Gamma }}_{i_2k_1}^\alpha D_{i_3}\mathrm{}D_{i_6)}\overline{T}_{\alpha k_2}+k_1k_2].`$
If the second rank tensor with components $`\overline{T}_{k_1k_2}`$ is replaced by the metric components $`\overline{g}_{k_1k_2}`$ then the related covariant derivatives (provided we deal with a torsion free affine connection) vanish and the above expressions are simplified. One could derive for $`n=5`$ the result:
$`_{(i_1}\mathrm{}_{i_5)}\overline{g}_{k_1k_2}={\displaystyle \frac{4}{3}}[D_{i_1}\mathrm{}D_{i_3}\overline{R}_{k_1i_4i_5k_2}+2(D_{i_1}\overline{R}_{k_1i_2i_3\rho }\overline{R}_{i_4i_5k_2}^\rho +k_1k_2)].`$ (15)
On the other hand for $`n=6`$ one gets:
$`_{(i_1}_{i_2}_{i_3}i_4_{i_5}_{i_6)}\overline{g}_{k_1k_2}=`$ $`{\displaystyle \frac{10}{7}}D_{(i_1}\mathrm{}D_{i_4}\overline{R}_{k_1i_5i_6)k_2}+{\displaystyle \frac{34}{7}}(D_{(i_1}D_{i_2}\overline{R}_{k_1i_3i_4\rho }\overline{R}_{i_5i_6)k_2}^\rho +k_1k_2)`$ (16)
$`+`$ $`{\displaystyle \frac{55}{7}}D_{(i_1}\overline{R}_{k_1i_2i_3\rho }D_{i_4}\overline{R}_{i_5i_6)k_2}^\rho +{\displaystyle \frac{16}{7}}\overline{R}_{k_1(i_1i_2\rho }\overline{R}_{i_3i_4l}^\rho \overline{R}_{i_5i_6)k_2}^l.`$
Thus, plugging into (10) expressions (15) and (16) we end up with the following expansion of the metric tensor in RNC:
$`g_{k_1k_2}=`$ $`\overline{g}_{k_1k_2}+{\displaystyle \frac{1}{2!}}{\displaystyle \frac{2}{3}}\overline{R}_{k_1i_1i_2k_2}\xi ^{i_1}\xi ^{i_2}`$ (17)
$`+`$ $`{\displaystyle \frac{1}{3!}}D_{i_1}\overline{R}_{k_1i_2i_3k_2}\xi ^{i_1}\mathrm{}\xi ^{i_3}`$
$`+`$ $`{\displaystyle \frac{1}{4!}}{\displaystyle \frac{6}{5}}\left[D_{i_1}D_{i_2}\overline{R}_{k_1i_3i_4k_2}+{\displaystyle \frac{8}{9}}\overline{R}_{k_1i_1i_2m}\overline{R}_{i_3i_4k_2}^m\right]\xi ^{i_1}\mathrm{}\xi ^{i_4}`$
$`+`$ $`{\displaystyle \frac{1}{5!}}{\displaystyle \frac{4}{3}}[D_{i_1}\mathrm{}D_{i_3}\overline{R}_{k_1i_4i_5k_2}+2(D_{i_1}\overline{R}_{k_1i_2i_3\rho }\overline{R}_{i_4i_5k_2}^\rho +k_1k_2)]\xi ^{i_1}\mathrm{}\xi ^{i_5}`$
$`+`$ $`{\displaystyle \frac{1}{6!}}{\displaystyle \frac{10}{7}}[D_{i_1}\mathrm{}D_{i_4}\overline{R}_{k_1i_5i_6k_2}+{\displaystyle \frac{17}{5}}(D_{i_1}D_{i_2}\overline{R}_{k_1i_3i_4\rho }\overline{R}_{i_5i_6k_2}^\rho +k_1k_2)`$
$`+`$ $`{\displaystyle \frac{11}{2}}D_{i_1}\overline{R}_{k_1i_2i_3\rho }D_{i_4}\overline{R}_{i_5i_6k_2}^\rho +{\displaystyle \frac{8}{5}}\overline{R}_{k_1i_1i_2\rho }\overline{R}_{i_3i_4l}^\rho \overline{R}_{i_5i_6k_2}^l]\xi ^{i_1}\mathrm{}\xi ^{i_6}`$
$`+`$ $`O(\xi ^{i_1}\mathrm{}\xi ^{i_7}).`$
The expansion (17) is in perfect agreement with that quoted in . The inverse of the metric tensor (obeying $`g_{k_1k_2}g^{k_2k_3}=\delta _{k_3}^{k_1}`$) can be found to be:
$`g^{k_1k_2}=`$ $`\overline{g}^{k_1k_2}{\displaystyle \frac{1}{2!}}{\displaystyle \frac{2}{3}}\overline{R}_{i_1i_2}^{k_1k_2}\xi ^{i_1}\xi ^{i_2}`$ (18)
$``$ $`{\displaystyle \frac{1}{3!}}D_{i_1}\overline{R}_{i_2i_3}^{k_1k_2}\xi ^{i_1}\mathrm{}\xi ^{i_3}`$
$``$ $`{\displaystyle \frac{1}{4!}}{\displaystyle \frac{6}{5}}\left[D_{i_1}D_{i_2}\overline{R}_{i_3i_4}^{k_1k_2}{\displaystyle \frac{4}{3}}\overline{R}_{i_1i_2m}^{k_1}\overline{R}_{i_3i_4}^{mk_2}\right]\xi ^{i_1}\mathrm{}\xi ^{i_4}`$
$``$ $`{\displaystyle \frac{1}{5!}}{\displaystyle \frac{4}{3}}[D_{i_1}\mathrm{}D_{i_3}\overline{R}_{i_4i_5}^{k_1k_2}3(D_{i_1}\overline{R}_{i_2i_3\rho }^{k_1}\overline{R}_{i_4i_5}^{\rho k_2}+k_1k_2)]\xi ^{i_1}\mathrm{}\xi ^{i_5}`$
$``$ $`{\displaystyle \frac{1}{6!}}{\displaystyle \frac{10}{7}}[D_{i_1}\mathrm{}D_{i_4}\overline{R}_{i_5i_6}^{k_1k_2}5(D_{i_1}D_{i_2}\overline{R}_{i_3i_4\rho }^{k_1}\overline{R}_{i_5i_6}^{\rho k_2}+k_1k_2)`$
$``$ $`{\displaystyle \frac{17}{2}}D_{i_1}\overline{R}_{i_2i_3\rho }^{k_1}D_{i_4}\overline{R}_{i_5i_6}^{\rho k_2}+{\displaystyle \frac{16}{3}}\overline{R}_{i_1i_2\rho }^{k_1}\overline{R}_{i_3i_4l}^\rho \overline{R}_{i_5i_6}^{lk_2}]\xi ^{i_1}\mathrm{}\xi ^{i_6}`$
$`+`$ $`O(\xi ^{i_1}\mathrm{}\xi ^{i_7}).`$
As a simple check one could consider the symmetric space $`V_n`$ in RNC for which $`D_k\overline{R}_{lmnp}=0`$ and prove that indeed the R terms in (17) satisfy:
$`g_{\alpha \beta }=\overline{g}_{\alpha \beta }+{\displaystyle \frac{1}{2}}{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}(1)^h{\displaystyle \frac{2^{2k+2}}{(2k+2)!}}f_\alpha ^{\sigma _1}f_{\sigma _2}^{\sigma _1}\mathrm{}f_{\sigma _{k1}\beta },h=\{\begin{array}{cc}k+1\hfill & \text{if k is even}\hfill \\ k\hfill & \text{if k is odd}\hfill \end{array}`$ (21)
where $`f_\alpha ^{\sigma _1}=\overline{R}_{\alpha i_1i_2}^{\sigma _1}\xi ^{i_1}\xi ^{i_2}`$ and $`f_{\sigma _{k1}\beta }=\overline{R}_{i_p\sigma _{k1}i_{p+1}\beta }\xi ^{i_p}\xi ^{i_{p+1}}`$.
## 3 Conclusions
We have shown the method one could rely on to evaluate the expansion of the components of the metric tensor in RNC at a specific order. The recursive structure permits an answer the derivation of which becomes cumbersome when one attempts to calculate higher order terms. |
warning/0001/cond-mat0001140.html | ar5iv | text | # Stretching necklaces
## I Introduction
Recently it has become possible to manipulate single polymer chains and to experimentally study their elastic response . Our aim here is to describe theoretically the single chain elasticity in the specific case of polyelectrolytes in poor solvents, typically in water, a poor solvent for most organic polyelectrolyte backbones. Beside the direct experimental measurements of single chain elasticity, there are at least two other motivations for this study: the possible relation with drag reduction , and the relation to the elastic behavior of more complex structures such as polyelectrolyte gels.
Charged polymers in solution, polyampholytes and polyelectrolytes, are known to reduce drag very efficiently in turbulent flows. This property is shared, to some extent, by neutral polymers. For neutral polymers, drag reduction has sometimes been explained by the dissipation associated to the stretch / coil transition where the polymer chains undergo a discontinuous elongation. One would also anticipate a large dissipation for a collapsed polymer globule in a poor solvent: even for fairly small variations of the stress there is a significant hysteresis loop in the globule/stretched chain transition. Of course, the concentration of polymer globules in a poor solvent must be extremely low in order to avoid phase separation. It has been suggested recently that polyelectrolytes in poor solvents may exhibit a pearl-necklace structure where collapsed globules, the pearls, are connected by stretched polymer strings. One may look at this system as a way of solubilizing (connected) dense polymer globules at finite concentrations, these structures could then be good candidates to enhance drag reduction.
If the necklaces comprise many pearls (which, we think, might be difficult to achieve for single chains in the accessible experimental range of parameters), thermal fluctuations smear out the single pearl features and the force/elongation curve is continuous. The continuous elasticity of these necklaces may be used as a starting point to model the elasticity of more complex, systems : polyelectrolyte gels will be described in a separate paper.
The behavior of polyelectrolyte chains in a Theta solvent has been discussed thoroughly theoretically and it is fairly well understood (we will not need in this paper to consider the bond angle correlations and the persistence length that remain controversial issues). We consider a chain of $`N`$ monomers of size $`b`$, with a fraction $`f`$ of charged monomers. The repulsive Coulombic interaction between charged monomers is written as $`V(𝐫)=k_\mathrm{B}Tl_\mathrm{B}/r`$, $`l_\mathrm{B}=e^2/(4\pi ϵk_\mathrm{B}T)`$ being the Bjerrum length where $`ϵ`$ is the dielectric constant of the solvent and $`e`$ the elementary charge. The chain radius can be estimated by a balance between entropy and electrostatic energy that is equivalent to the electrostatic blob model : the chain may be viewed as a stretched string of Gaussian isotropic blobs each containing $`(l_Bf^2/b)^{2/3}`$ monomers and of radius $`(l_Bf^2/b)^{1/3}b`$; the overall chain length is then $`R=bN\left(\frac{l_\mathrm{B}f^2}{b}\right)^{1/3}`$. The electrostatic interaction is relevant for high enough charge fraction i.e., $`f>N^{3/4}`$, otherwise the chain remains essentially Gaussian isotropic (comprising less than one blob).
If the chain backbone is in a poor solvent (which is often the case in experiments), there is a competition between surface tension effects tending to minimize the polymer/water(solvent) contact area and electrostatics tending to maximize the overall chain size. In the following, the poor solvent conditions are characterized by a negative excluded volume $`v=\tau b^3`$, where $`\tau =(\theta T)/\theta `$ is the relative distance to the compensation temperature $`\theta `$. In the absence of electrostatic interactions, the chain collapses into a dense globule that may be viewed as a small region of dense polymer phase at co-existence with free solvent; the finite size of the globule is associated with an extra energy penalty due to the polymer-water surface tension. The balance of the osmotic pressure between the dense and dilute phases (of almost vanishing concentration) yields the concentration inside the globule $`c=\tau /b^3`$. Alternatively, the dense phase may be described by a close packing of thermal blobs of size $`\xi _t=b/\tau `$, containing $`g=1/\tau ^2`$ monomers. A globule of $`N`$ monomers has then a radius $`R=b\left(N/\tau \right)^{1/3}`$. The corresponding surface tension is $`\gamma =k_\mathrm{B}T/\xi _t^2\tau ^2/b^2`$. For a charged chain, the balance between surface tension and electrostatics was first considered by Khokhlov assuming a cylindrical globular shape. This yields a highly elongated globule of length $`L=Nb(l_\mathrm{B}f^2/b)^{2/3}/\tau `$ and radius $`R_c=b(l_\mathrm{B}f^2/b)^{1/3}`$ provided that the radius of the chain in the absence of electrostatic interactions, in a spherical globular conformation, is larger than $`R_c`$; for smaller chains, the electrostatic interaction is only a perturbation. The elongated cylindrical globule can be looked at as a stretched string of spherical globules of radius $`R_c`$.
Recently, it has been proposed by Kantor and Kardar that the large spherical globule instability is rather similar to the Rayleigh instability of charged liquid droplets. A large liquid droplet carrying an electrostatic charge, breaks up into marginally stable smaller droplets for which surface tension and electrostatic self-energy are of the same order of magnitude. For a ”liquid of thermal blobs” with individual size $`\xi _t`$, the size of the marginal stable droplet (the pearl) is again $`R_c`$. However, in order to minimize the total free energy, the droplets must be infinitely far apart. Due to the chain connectivity, the best separation that can be achieved in the polyelectrolyte problem is to connect the pearls with stretched polyelectrolyte strings. This leads to the pearl-necklace model introduced by Dobrynin, Rubinstein and Obukhov . The string monomers are then in equilibrium with the pearl monomers. An equilibrium between a collapsed chain and a globule is also found when an external force is applied on a globule. As shown by Halperin and Zhulina , in a first approximation, (discarding finite size corrections) the strings must have a radius equal to $`\xi _t`$ (the thermal correlation length in the pearls), and a tension $`\sigma =\frac{k_\mathrm{B}T}{\xi _t}`$. The string tension is qualitatively due to the repulsion between adjacent pearls $`k_\mathrm{B}Tl_\mathrm{B}Q^2/l^2`$ (a logarithmic factor corresponding to pearls farther apart is omitted), where $`Q`$ is the pearl charge and $`l`$ the length of a string. The string length is determined by the force balance $`\sigma =k_\mathrm{B}T\tau /b=k_\mathrm{B}Tl_\mathrm{B}Q^2/l^2`$. It is easily checked that, if the solvent is poor enough, most monomers belong to pearls and that the stretched strings dominate the total length $`L=(f^2l_\mathrm{B}/b)^{1/2}Nb\tau ^{1/2}`$.
It is worthwhile to return to the case of an ordinary simple liquid, the Rayleigh problem, where a large drop splits into non-interacting smaller drops far apart. In order to remain close to the polyelectrolyte problem, we consider a ”formal” liquid of disconnected thermal blobs. The free energy of an assembly of $`\stackrel{~}{N}=N\tau ^2`$ blobs grouped into drops of $`\stackrel{~}{m}=m\tau ^2`$ blobs each is the sum of the electrostatic and interfacial energies of the individual drops. It reads $`F/k_\mathrm{B}T=0.5\stackrel{~}{N}(\mathrm{\Lambda }/2)^{1/3}\text{f}(x)`$ where $`x=2\stackrel{~}{m}/\mathrm{\Lambda }`$ is the number of blobs in the marginal droplet. Note that $`\mathrm{\Lambda }=\tau ^3/(l_\mathrm{B}f^2/b)`$ (hence $`\mathrm{\Lambda }>1`$ is required). The function $`\text{f}(x)=2x^{1/3}+x^{2/3}`$ is plotted in Fig.1:
* It has a minimum at $`x=1`$, the preferred droplet size contains $`\stackrel{~}{m}=\mathrm{\Lambda }/2`$ blobs.
* It changes curvature at $`x=4`$: an assembly of identical droplets smaller than four times the preferred size is locally stable against polydispersity.
* For a size larger than about 1.4 times the preferred size, it is favorable to split a droplet into two identical droplets.
* The overall free energy $`\text{f}(x)`$ is rather flat and thermal fluctuations are anticipated to become important for mesoscopic systems. For an assembly of 1000 blobs (a rather large number) and a preferred droplet size of 100 blobs, the barrier against splitting all 10 droplets is roughly $`4k_\mathrm{B}T`$ only. The fluctuations become less important with decreasing solvent quality.
Though we are discussing here a liquid of disconnected blobs only, we may anticipate that steepest descent calculations where only the most probable necklace conformation is accounted for are not always very accurate.
Let us briefly make a remark on single chain stretching in general. There came up a number of experimental studies . Some of them deal also with DNA stretching and defolding under forces . Theoretical papers describe stretching of DNA molecules . The forces we are going to discuss in the case of polyelectrolytes in poor solvent are much lower than appearing there. In most cases related to the DNA overstretching plays a significant role, in contrast to the problem described here.
The paper is organized as follows: in the next section we first summarize briefly the known results on the pearl-necklace structure ; we then discuss the structure of necklaces under an external force; we start with the continuous limit, corresponding to many pearls connected by strings and show that the pearls are ”unwinded” by the external force; when the number of pearls becomes small a more detailed treatment is necessary, we study in details the dissolution of a necklace of two pearls. Section III is devoted to a more general discussion of the discreteness of the number of pearls. A somewhat simpler description, that we think still captures the essential physics, is developed to describe individual pearl unwinding for necklaces involving many pearls. Fluctuations progressively smear out force plateaus associated to individual pearls when the number of pearls increases. This whole analysis ignores the role of counterions of the chain. The pearl surface potential however, can be rather high even for moderate charge fractions and some charge regulation due to counterion condensation on the pearls is to be expected. Large pearls inducing counterion condensation and very poor solvents are discussed in section IV where we show that charge regulation can suppress the large globule instability. We also consider very poor solvents for which there is hardly any solvent in the pearl core and the dielectric constant is low; the polyelectrolyte behaves then as an ionomer <sup>*</sup><sup>*</sup>*A short account of this work has been presented at the St. Petersburg conference on ”Molecular Mobility and Order in Polymer Systems” June 1999, ..
## II Polyelectrolytes in poor solvent under external forces
### A Equilibrium necklaces, $`\phi =0`$
Polyelectrolyte chains in a poor solvent show a rich conformational phase behavior because competing interactions are ruling the chain statistics: surface tension drives the globule toward small polymer/water interfaces whilst electrostatics drive it toward large overall extensions. Therefore, the globular structures for large enough charge fractions are no longer spherical as in the case of neutral polymers, but elongated. This was first recognized by Khokhlov who proposed a cylindrical globular shape (assuming that there is only one characteristic length). More precisely he finds that a globule of radius $`R`$ becomes elongated when $`F_{\mathrm{surface}}=\gamma R^2>F_{\mathrm{el}}=k_\mathrm{B}T(l_\mathrm{B}f^2N^2)/R`$, i.e., $`\tau >(l_\mathrm{B}/b)f^2N`$. For larger charge fractions the globule elongates to a thickness $`D`$ and length $`L`$. This can be seen from a simple free energy balance. The free energy of the cylindrical globule is (remember that the concentration is given by $`c=\tau /b^3=N/(D^2L)`$)
$$\frac{F}{k_\mathrm{B}T}=\tau ^{3/2}N^{1/2}\left(\frac{L}{b}\right)^{1/2}+\frac{l_\mathrm{B}f^2N}{L}$$
(1)
which yields upon minimization, the length of the cylinder $`L=b(N/\tau )\left(l_\mathrm{B}f^2/b\right)^{2/3}`$ and the diameter comparable to the size $`R_cb(l_\mathrm{B}f^2/b)^{1/3}`$ of the marginally stable spherical globule, as already mentioned in the introduction.
Kantor and Kardar have shown that the elongated globule structure is unstable and breaks up into marginally stable pearls by analogy with the Rayleigh instability of charged liquid droplets. Due to the chain connectivity these globules are separated by stretched strings. The details of the Rayleigh-like instability for the necklace structure have been worked out by Dobrynin et al. . The elongated globule breaks up into pearls and strings. The pearls of size $`R`$, contain $`m=c_0R^3`$ monomers, where $`c_0=\tau /b^3`$ is the equilibrium concentration of the globule. The elongated strings of length $`l`$ and diameter $`d`$, contain $`m_s`$ monomers. For completeness and later use, we write down the free energy in the following form
$$\frac{F}{k_\mathrm{B}T}=\frac{N}{m}\left\{\gamma R^2+\frac{l_\mathrm{B}f^2m^2}{R}+\frac{F_{\mathrm{elastic}}(l)}{k_\mathrm{B}T}+\gamma ld+\frac{l_\mathrm{B}f^2m_s^2}{l}\right\}+\frac{lbf^2N^2}{L}$$
(2)
The terms correspond respectively to the surface energy of the pearls, to the electrostatic self-energy of the pearls, to the elastic free energy of the strings, to their surface energy, to their electrostatic energy and to the entire (smeared out) electrostatic energy of the whole chain. Within the limitations of this model, the necklace structure can be determined by minimization of eq.(2). The size of the pearls is $`RR_\mathrm{c}=b/((l_\mathrm{B}/b)f^2)^{1/3}`$ and each pearl contains $`m=(\tau b)/(l_\mathrm{B}f^2)`$ monomers. The string diameter is $`d=\xi _t`$ and the string length $`l=b(b\tau /l_\mathrm{B}f^2)^{1/2}`$. The total length of the chain is $`L=bN(l_\mathrm{B}f^2/b)^{1/2}\tau ^{1/2}`$. This last result shows that the chain is indeed stretched. The physical picture can be summarized as follows: pearls of size $`R_c`$ are separated by strings with a tension equal to the critical tension for unwinding pearls $`\sigma `$. This tension comes from the long range electrostatic repulsions between pearls. The string length obtained from this argument $`l`$ agrees with the minimization of the free energy. It is easily checked that most monomers are packed in pearls and that the stretched strings dominate the total length $`L`$.
### B Necklace stretching in the continuous limit, $`\phi 0`$
We first discuss qualitatively, the effect of an external force $`\phi `$ to the necklace chain. When the polyelectrolyte necklace is stretched under an external force, the pearl size and the string tension are essentially unaffected as they are fixed by the Rayleigh instability and by the equilibrium between pearl and string monomers respectively. The string length $`l`$, is given by a force balance on a ”half” necklace, relating the tension $`\sigma `$, the external force $`\phi `$ and the electrostatic force exerted by the other half necklace, $`\sigma =k_\mathrm{B}T\tau /b=\phi +k_\mathrm{B}Tl_\mathrm{B}Q^2/l^2`$; it increases with the external force and thus some pearls are converted into stretched strings (see Fig.2).
This argument yields the pearl separation length $`l(\phi )`$ as a function of the external force. For moderate external forces where most monomers remain in pearls, we obtain the simple force law
$$\frac{\phi }{k_\mathrm{B}T}=\frac{\tau }{b}\left(1\left(\frac{L(0)}{L}\right)^2\right)$$
(3)
The corresponding force profile is represented in Fig.3. This equation holds for small enough values of $`L`$, i.e., $`L<Nb\tau `$ and logarithmic factors have been omitted. For larger forces, the hydrophobic forces are not important and the elasticity is the Gaussian elasticity of the strings as for a polyelectrolyte in a Theta solvent. The pearl number is treated here as a continuous variable. At small scales the force curve is expected to show plateaus corresponding to the unwinding of individual pearls, however for a large number of pearls, the curve is rounded by fluctuations. The typical fluctuation of the pearl number must then be larger than one.
To study the deformation behavior of isolated necklaces in more details we use an explicit free energy close to eq.(2). A more detailed expression for the interactions between strings and pearls is introduced, and we further account for an imposed external forces $`\phi `$.
$`{\displaystyle \frac{G}{k_\mathrm{B}T}}`$ $`=`$ $`p\left({\displaystyle \frac{\tau }{b}}\right)^2R^2+(p1)\left({\displaystyle \frac{\tau }{b}}\right)^2dl+p{\displaystyle \frac{l_\mathrm{B}f^2m^2}{R}}+{\displaystyle \frac{1}{2}}{\displaystyle \frac{l_\mathrm{B}f^2M^2}{L}}\mathrm{log}(L/d)+`$ (4)
$`+`$ $`2{\displaystyle \frac{l_\mathrm{B}f^2mM}{L}}\mathrm{log}(L/R)+p{\displaystyle \frac{l_\mathrm{B}f^2m^2}{l}}\mathrm{log}p+{\displaystyle \frac{L^2}{2Mb^2}}\phi L`$ (5)
We have used here the geometrical relationships for the necklace structure, the string length between two pearls is $`l=L/(p1)`$, where $`L`$ is the total length of the necklace chain and $`p`$ the number of pearls; the number of monomers in one pearl is $`m=(NM)/p`$, where $`N`$ is the total number of monomers in the entire chain and $`M=(p1)m_s`$ is the total number of monomers in the strings; the diameter of the strings is given by $`d^2=Mb^3/L\tau `$; finally the radius of the pearls is $`R=((NM)/p)^{1/3}b/\tau ^{1/3}`$. It is useful to note that despite we have written out the logarithmic corrections here, we do not derive with respect to the chain length them in the following. We added them to keep control on the physical origin different terms. Therefore we treat them as constants. Precise numerical factors, however, are omitted as well.
It turns out to be convenient to use again the parameter $`\mathrm{\Lambda }=b\tau ^3/l_\mathrm{B}f^2`$ proportional to the number of thermal blobs in a pearl and the following dimensionless variables, a reduced tension $`t=L/(Mb\tau )`$ and a reduced external force, $`\stackrel{~}{\phi }=\phi b/\tau `$. This amounts to rescale every quantity by the thermal blob. The physically relevant cases correspond to $`\mathrm{\Lambda }>>1`$, thus $`1/\mathrm{\Lambda }`$ can be used as an expansion parameter. With these definitions we rewrite the free energy as
$`{\displaystyle \frac{G}{k_\mathrm{B}T}}`$ $`=`$ $`p^{1/3}(\stackrel{~}{N}\stackrel{~}{M})^{2/3}+{\displaystyle \frac{1}{\mathrm{\Lambda }p^{2/3}}}(\stackrel{~}{N}\stackrel{~}{M})^{5/3}+`$ (6)
$`+`$ $`{\displaystyle \frac{1}{2\mathrm{\Lambda }t}}\stackrel{~}{M}\mathrm{log}\left(L/d\right)+2{\displaystyle \frac{1}{\mathrm{\Lambda }t}}(\stackrel{~}{N}\stackrel{~}{M})\mathrm{log}(L/R)+`$ (7)
$`+`$ $`{\displaystyle \frac{1}{\mathrm{\Lambda }t}}{\displaystyle \frac{(\stackrel{~}{N}\stackrel{~}{M})^2}{\stackrel{~}{M}}}\mathrm{log}p`$ (8)
$`+`$ $`\stackrel{~}{M}t^{1/2}+\stackrel{~}{M}t^2\stackrel{~}{\phi }\stackrel{~}{M}t`$ (9)
The free energy is expressed only in terms of scaled variables and we introduced $`(\stackrel{~}{N},\stackrel{~}{M})=(N,M)\tau ^2`$. The independent variables for the minimization are $`\{p,\stackrel{~}{M},t\}`$. After minimization of the free energy, we obtain
$$\frac{\stackrel{~}{N}\stackrel{~}{M}}{p}=\frac{\mathrm{\Lambda }}{2}$$
(10)
$`{\displaystyle \frac{2}{3}}\left({\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{p}}\right)^{1/3}{\displaystyle \frac{5}{3\mathrm{\Lambda }}}\left({\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{p}}\right)^{2/3}\left({\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{\stackrel{~}{M}}}\right)^2{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\mathrm{log}p2{\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{\stackrel{~}{M}}}{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\mathrm{log}p`$ (11)
$`+\sqrt{t}+{\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\mathrm{log}(L/d){\displaystyle \frac{2}{\mathrm{\Lambda }t}}\mathrm{log}(L/R)+t^2\stackrel{~}{\phi }t=0`$ (12)
$`2t+{\displaystyle \frac{1}{2\sqrt{t}}}{\displaystyle \frac{1}{\mathrm{\Lambda }t^2}}\left(2\left({\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{\stackrel{~}{M}}}\right)\mathrm{log}(L/R)+\left({\displaystyle \frac{\stackrel{~}{N}\stackrel{~}{M}}{\stackrel{~}{M}}}\right)^2\mathrm{log}p\right)\stackrel{~}{\phi }{\displaystyle \frac{1}{2\mathrm{\Lambda }t^2}}\mathrm{log}(L/d)=0`$ (13)
The first equation determines the number of blobs in one pearl, in agreement with our previous arguments. The second derivative with respect to $`p`$ allows to evaluate the fluctuation of the number of pearls, we find a r.m.s fluctuation $`\delta p=\mathrm{\Lambda }^{1/3}/p^{1/2}`$ (we neglect here the coupling between the fluctuations of $`p`$ and the fluctuations of other variables but this does not qualitatively change the result). The continuous description is consistent for $`p>\mathrm{\Lambda }^{2/3}`$ when the fluctuation of the pearl number is larger than one. The two other equations determine the effective tension $`t`$ and the number of monomers in the strings or equivalently the ratio $`\mathrm{\Delta }`$ of the number of monomers in the pearls to that in the strings $`\mathrm{\Delta }=(\stackrel{~}{N}\stackrel{~}{M})/\stackrel{~}{M}`$.
$`{\displaystyle \frac{3}{2}}\left({\displaystyle \frac{2}{\mathrm{\Lambda }}}\right)^{1/3}{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\mathrm{\Delta }(\mathrm{\Delta }+2)\mathrm{log}p+\sqrt{t}+{\displaystyle \frac{1}{2\mathrm{\Lambda }t}}\mathrm{log}(L/d)+{\displaystyle \frac{2}{\mathrm{\Lambda }t}}\mathrm{log}(L/R)+t^2\stackrel{~}{\phi }t=0`$ (14)
$`{\displaystyle \frac{1}{\mathrm{\Lambda }t^2}}\mathrm{\Delta }\left\{\mathrm{\Delta }\mathrm{log}p+2\mathrm{log}(L/R)\right\}+2t+{\displaystyle \frac{1}{2\sqrt{t}}}\stackrel{~}{\phi }{\displaystyle \frac{1}{2\mathrm{\Lambda }t^2}}\mathrm{log}(L/d)=0`$ (15)
The external force can be eliminated and
$$\frac{1}{\mathrm{\Lambda }t}\mathrm{log}(L/d)+\frac{2\mathrm{\Delta }}{\mathrm{\Lambda }t}\mathrm{log}\left(\frac{L}{Rp}\right)+\frac{2}{\mathrm{\Lambda }t}\mathrm{log}(L/R)t^2+\frac{\sqrt{t}}{2}\frac{3}{2}\left(\frac{2}{\mathrm{\Lambda }}\right)^{1/3}=0$$
(16)
For the free necklace , $`\phi =0`$, the fraction $`\mathrm{\Delta }`$ of pearl monomers to string monomers is of order $`\mathrm{\Delta }\sqrt{\mathrm{\Lambda }/\mathrm{log}p}`$. When a finite force is applied, $`\mathrm{\Delta }`$ decreases. In the limit $`\mathrm{\Lambda }\mathrm{}`$, the reduced tension is $`t=t_0=(1/2)^{2/3}`$, which can be used as a basis for an expansion to find approximate solutions. Following eq. (16), the leading correction to $`t_0`$ comes from the the pearl self-energy; it is negative and proportional to $`\mathrm{\Lambda }^{1/3}`$. It is a typical finite pearl size correction proportional to $`\xi _t/R`$ and comprising both surface tension and electrostatic contributions of the same order. The next order correction comes from the electrostatic interactions, it is proportional to $`\mathrm{\Delta }`$ and positive; at small force, it is proportional to $`\mathrm{\Lambda }^{1/2}\mathrm{log}(L/R)/\mathrm{log}p^{1/2}`$. At moderate force, the force extension curve to leading order in $`1/\mathrm{\Lambda }`$, is given by
$$\frac{\phi }{k_\mathrm{B}T}=(2t_0+1/2\sqrt{t_0})\frac{\tau }{b}\left(1\alpha \mathrm{\Lambda }^{1/3}\right)\frac{l_\mathrm{B}f^2N^2}{L^2}\mathrm{log}p$$
(17)
where $`\alpha `$ is a positive numerical constant, depending on $`t_0`$ only. This last equation corresponds now to eq.(3). Physically it describes the unwinding of pearls by increasing external forces. Two new features appear in comparison with the simple argument above. The first is that a new term which corrects the (bare) line tension $`\sigma =k_\mathrm{B}T\tau /b`$ by a term proportional to $`\mathrm{\Lambda }^{1/3}`$. This term has its physical origin in the finite size of the pearls, it is proportional to $`(\xi _t/R)^{1/3}`$ as for the unwinding of a neutral globule but has both surface tension and electrostatic contributions of the same order. The other point is the appearance of the logarithmic term, which comes from the interactions between pearls that are dominant. Under the logarithm, the number of pearls can be approximated by $`p=(N\tau ^2L\tau /b)/\mathrm{\Lambda }`$.
When a finite fraction of the pearls are unwinded $`\mathrm{\Delta }`$ becomes of order unity. The force law eq.(17) can no longer be applied when $`L/\tau Nb1`$. The force is then a second order polynomial in $`\mathrm{\Delta }`$ and, due to the correction to $`t_0`$ (see eq.16), the linear term in $`\mathrm{\Delta }`$ is positive; $`\phi `$ is of the form $`\phi =AB\mathrm{\Delta }^2+C\mathrm{\Delta }`$ where all three constants are positive and $`A`$ and $`B`$ are of the same order. This leads to a spinodal instability when about half of the pearls are unwinded. We thus expect the polymer to jump from a metastable necklace to a single string unwinding about half of the pearls at once. Note that our argument relies on small correction terms to the force and tension (as $`\mathrm{\Delta }1`$); our free energy, involving several approximations, may be too crude in this case. However when half of the pearls are unwinded the elasticity is dominated by the string elasticity even in the absence of any transition.
### C Discrete pearls: two pearl-necklace
When there are only a few pearls in the necklace, the fact that the number of pearls is discrete cannot be ignored. This is of particular importance if the total number of pearls is small as shown by many numerical simulations . In this subsection, we discuss the special case of two pearls separated by a string. This is the simplest model which is physically meaningful. For energetic reasons based on the here used free energy, the ends of the necklace chain consist of pearls (a chain with one pearl in the middle and two strings at the end is energetically not favorable). This minimal model has the advantage that the dissolution of the pearls can be computed in details. The number of pearls is fixed, in contrast to the continuous model. If the necklace had more than two pearls, we would have to take into account the fact that the pearls at the chain end, are larger than the central pearls, simply because of smaller electrostatic interactions (one missing neighbor). For a small number of pearls, we expect a discontinuous stretching with force plateaus corresponding to the unwinding of a pearl.
In order to describe the jump from 2 to 0 pearls, we start from a free energy that takes into account the interactions more precisely and includes all the surface and electrostatic free energies.
$`G=2\gamma R^2`$ $`+`$ $`2{\displaystyle \frac{l_\mathrm{B}f^2m^2}{R}}+\gamma Ld+{\displaystyle \frac{l_\mathrm{B}f^2M}{L}}\mathrm{log}(L/d)+`$ (18)
$`+`$ $`2{\displaystyle \frac{l_\mathrm{B}f^2mM}{L}}\mathrm{log}(L/R)+{\displaystyle \frac{l_\mathrm{B}f^2m^2}{L}}+`$ (19)
$`+`$ $`{\displaystyle \frac{L^2}{2Mb}}\phi L`$ (20)
The free energy can be rescaled using dimensionless variables ($`\stackrel{~}{m}=\tau ^2m`$):
$`{\displaystyle \frac{G}{k_\mathrm{B}T\stackrel{~}{N}}}`$ $`=`$ $`{\displaystyle \frac{1}{\stackrel{~}{N}}}\left(2\stackrel{~}{m}^{2/3}+{\displaystyle \frac{2}{\mathrm{\Lambda }}}\stackrel{~}{m}^{5/3}\right)+{\displaystyle \frac{1}{\mathrm{\Lambda }t}}{\displaystyle \frac{2\stackrel{~}{m}}{\stackrel{~}{N}}}\mathrm{log}(L/R)+`$ (21)
$`+`$ $`{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\left(1{\displaystyle \frac{2\stackrel{~}{m}}{\stackrel{~}{N}}}\right)\mathrm{log}(L/d)+\left(1{\displaystyle \frac{2\stackrel{~}{m}}{\stackrel{~}{N}}}\right)t^{1/2}+\left(1{\displaystyle \frac{2\stackrel{~}{m}}{\stackrel{~}{N}}}\right)t^2+`$ (22)
$`+`$ $`{\displaystyle \frac{1}{\mathrm{\Lambda }t}}\left({\displaystyle \frac{\stackrel{~}{m}}{\stackrel{~}{N}}}\right)^2{\displaystyle \frac{1}{\left(1\frac{2\stackrel{~}{m}}{\stackrel{~}{N}}\right)}}\stackrel{~}{\phi }t\left(1{\displaystyle \frac{2\stackrel{~}{m}}{\stackrel{~}{N}}}\right)`$ (23)
As the number of pearls is fixed, there are only two variables: the number of thermal blobs in a pearl $`\stackrel{~}{m}`$ and the reduced tension $`t`$. Two simple limits are described by this equation: $`\stackrel{~}{m}0`$ corresponds to a string with no pearls and the chain is an extended string of thermal blobs in the so-called Pincus regime ; in the absence of external force, $`\phi =0`$, we describe a free two-pearl necklace.
The free energy map $`\text{G}(t,m)`$ eq.(21) is shown in Fig.4, for $`\stackrel{~}{N}=2.05\mathrm{\Lambda }/2`$ with $`\mathrm{\Lambda }=10^3`$. The plot is typical for high $`\mathrm{\Lambda }`$-values where the description by the most likely necklace (steepest descent) is meaningful.
At low force, there is only one minimum in the $`G`$-map that corresponds to a necklace with roughly the optimal pearl size. For higher applied force the pearls unwind. At some value of the force, situation (a), a second minimum corresponding to a stretched string without pearls appears as a metastable state. With increasing force the free energy $`G`$ of the stretched string becomes closer to the necklace free energy, the two minima in the free energy are equal in situation (b); they are separated by a saddle point significantly higher in energy. For even higher forces the necklace becomes metastable; the necklace minimum disappears in situation (b) where roughly half of the pearl content is unwinded. From these simple considerations, we plot an elongation/force diagram in Fig.5with two branches. The stability exchange (situation (b) in the contour plots) is shown by the dashed line.
Below the dashed line the stretched string branch is metastable whereas above this line, the necklace branch is metastable. In a fast stretch/collapse/stretch cycle we expect a hysteresis loop following most of the metastable branches. In a quasi static extension procedure under imposed increasing force a pseudo-plateau close to the dashed line in the force-extension curve is expected. A better description of the pseudo-plateau requires a more complete theory including thermal fluctuations. A first approximation for the chain partition function is:
$$\mathrm{\Xi }=𝑑m𝑑L\mathrm{exp}(G(m,t)/k_\mathrm{B}T)$$
(24)
where the integral is carried out over the previous $`t,m`$ map. The average length is obtained as the derivative of the logarithm of the partition function $`L=\frac{\mathrm{log}\mathrm{\Xi }}{\phi }`$. The force-extension curve is plotted in Fig.6.
A pseudo-plateau is indeed observed. According to linear response theory, the length fluctuations are obtained from the slope of this curve $`L^2L^2=kT\frac{L}{\phi }`$. The RMS fluctuation of the length is of the order of the average length on the pseudo-plateau and much weaker outside the pseudo-plateau. This description remains crude as we did not allow for unlike pearl sizes. As shown in the simple discussion of the Rayleigh instability for mesoscopic drops, pearl size fluctuations may be important for not too high values of $`\mathrm{\Lambda }`$. In the following, we give a simpler description where the tension of the string is kept constant during the pearl unwinding (a reasonable assumption in view of the numerics). It is then possible to get a simple picture also in the case when the necklace has several pearls.
## III Discrete pearl model
So far, we have ignored pearl size fluctuations; in the following we propose a simplified description of the necklace structure that allows to take these fluctuations into account. We describe approximately the necklace by a row of pearls, and we fix the structure of the strings i.e. their tension and their elastic energy per unit length. If the total number of monomers in the strings is $`M`$ as above and the total length of the chain (dominated by the strings) is $`L`$, $`M/L=\tau /b`$ and the total elastic energy of the strings is $`F_{\mathrm{string}}\frac{1}{2}k_\mathrm{B}TL\tau /b`$.
We then distribute the $`P=NM`$ remaining monomers into $`p`$ pearls with masses $`m_i`$, which are not necessarily identical. The total free energy includes the pearl self-energy, the string electrostatic energy and the smeared out electrostatic energy $`k_\mathrm{B}Tl_\mathrm{B}f^2N^2/L`$ between pearls and strings. In this model, all the contributions to the free energy except for the pearl self-energy are identical to those of the continuous model. We do not include here the logarithmic factors in the electrostatic term but they could be included as before. For a fixed length of the chain, the number of monomers in each pearl and the fluctuations of this number are essentially determined by the pearl self-energy, we thus ignore the fluctuations in all the other terms of the free energy. The main approximation of our model is that the length of the chain is only due to the strings and thus that the total number of monomers in the strings and therefore in the pearls are fixed. This approximation may not be quantitative but we expect it to give the correct physical picture and the correct scaling behavior.
In this section, in contrast to previous considerations, we impose the length of the chain and compute the force. In order to discuss this model qualitatively, we further simplify it by assuming that all pearls are identical. In a second step the fluctuations in the number of monomers of the individual pearls are explicitly taken into account.
The free energy of a single pearl is
$$F_p=\gamma R^2+k_\mathrm{B}T\frac{l_\mathrm{B}f^2m^2}{R}=k_\mathrm{B}T\left(N^{2/3}\tau ^{4/3}+\frac{l_\mathrm{B}}{b}f^2\tau ^{1/3}m^{5/3}\right)$$
(25)
where $`m`$ is the number of monomers in the pearl. The free energy per pearl monomer is then
$$\stackrel{~}{f}=m^{1/3}\tau ^{4/3}+\frac{l_\mathrm{B}}{b}f^2\tau ^{1/3}m^{2/3}$$
(26)
The free energy per monomer $`\stackrel{~}{f}`$ plotted as a function of the pearl mass $`m`$ has a minimum at $`m_0=\mathrm{\Lambda }/(2\tau ^2)`$. The second derivative is positive for $`m=m_0`$, it vanishes for $`m=4m_0`$. In the following, in order to take into account the deviations from the preferred value $`m=m_0`$, we approximate the free energy per monomer by a parabolic form where the parabola has a spring constant $`k=(^2\stackrel{~}{f}/m^2)|_{m_0}=m_0^1(l_\mathrm{B}f^2/b)^{4/3}`$. The total free energy of the chain can then be written as
$$F=F_{\mathrm{cont}}(m_0)+\frac{1}{2}k_\mathrm{B}Tkm_0\underset{i=1}{\overset{p}{}}(m_im_0)^2.$$
(27)
where $`F_{\mathrm{cont}}`$ is the free energy discussed earlier in the continuous limit with equal numbers of monomers $`m_0`$ in all pearls $`(F_{\mathrm{cont}}`$ is in a first approximation given by eq. (2) or by eq. (5) where the force term has been omitted) and the sum is over the $`p`$ pearls.
As a first step, we consider that all the pearls have the same number of monomers $`m`$ that can be different from the preferred value $`m_0`$. When the necklace is stretched, the number of monomers in each pearl decreases. At some point however, it becomes energetically more favorable to have one pearl less. In this crude argument, we assume that the jump from $`p`$ to $`p1`$ pearls occurs abruptly for a given imposed necklace length, this is equivalent to a large energy scale in the Boltzmann weights and hence to large pearls ($`\mathrm{\Lambda }>>1`$). The stability exchange from the necklace containing $`p`$ pearls to the necklace containing $`p1`$ pearls corresponds, in the parabolic potential, to opposite deviations from the preferred pearl size: $`\delta m_p=\delta m_{p1}`$. On the other hand, due to mass conservation, (the length is imposed), $`pm_p=(p1)m_{p1}`$. The necklace thus jumps from $`p`$ to $`p1`$ pearls for $`\delta m_p=m_0/(2p1)`$. For a chain containing $`p`$ pearls, we get
$$\delta m=\frac{1}{p}\left(N\frac{L}{b\tau }pm_0\right)$$
The force is obtained by derivation of the total free energy, it is equal to
$$\phi _p=\phi _{\mathrm{cont}}\frac{1}{p}\frac{km_0}{\tau b}\left(N\frac{L}{b\tau }pm_0\right).$$
(28)
The first term is the force calculated previously in the continuous limit and the second term is the contribution due to the discreteness of the pearls and to the fact that they do not have the preferred size; this ”individual pearl” contribution is non monotonic (see Fig.7). If the number of pearls in the necklace is small enough, the total force also exhibits downwards jumps.
This occurs if the distance $`l`$ between pearls is large enough ($`l>l_0\mathrm{\Lambda }^{2/9}m_0\tau b`$). The branches where the force decreases with length are unstable and the force actually shows a plateau. The plateau is obtained from a Maxwell-like construction on the total force. In practice, for non quasi-static transformations, the system rather follows the metastable branch and, when the applied force increases, the length jumps at the force extrema.
The force jumps obtained upon imposing the chain length are rounded by fluctuations of both the number of monomers in a pearl and the number of pearls. In order to take into account the fluctuations, we compute the partition function of a necklace of $`p`$ pearls
$`Z_p=Z_0{\displaystyle \underset{i}{}dm_i\mathrm{exp}\left\{\frac{1}{2}km_0\left(m_im_0\right)^2\right\}\delta \left(N\frac{L}{\tau b}\underset{i=1}{\overset{p}{}}m_i\right)}.`$ (29)
The delta function ensures the mass conservation. The bare partition function is associated to the free energy calculated in the continuous limit $`Z_0=\mathrm{exp}(F_{\mathrm{cont}}/k_\mathrm{B}T)`$. Formally we are looking for the partition function of a row of $`p`$ harmonic springs with a fixed total length. The result is
$$Z=Z_0\left(\frac{2\pi }{km_0}\right)^{p/2}\left(\frac{2\pi km_0}{p}\right)^{1/2}\mathrm{exp}\left\{\frac{km_0}{2p}\left(N\frac{L}{\tau b}pm_0\right)^2\right\}$$
(30)
The total partition function of the necklace chain is $`Z=_pZ_p`$ and the free energy is $`F=k_\mathrm{B}T\mathrm{log}Z`$. The force is then obtained by derivation
$`\phi =\phi _{con}+k_\mathrm{B}T{\displaystyle \frac{_p\left(\phi _p\frac{(l_\mathrm{B}f^2/b)^{2p/3}}{p^{1/2}}\mathrm{exp}(\frac{\mathrm{\Lambda }^{2/3}}{p}(N/m_0L/(\tau bm_0)p))^2\right)}{_p\left(\frac{(l_\mathrm{B}f^2/b)^{2p/3}}{p^{1/2}}\mathrm{exp}(\frac{\mathrm{\Lambda }^{2/3}}{p}(N/m_0L/(\tau bm_0)p))^2\right)}},`$ (31)
where $`\phi _{con}`$ is the force obtained in the continuous approximation 17 and
$$\phi _p=2\frac{\tau }{b}\frac{\mathrm{\Lambda }^{1/3}}{p}(N/m_0L/\tau bm_0p)$$
(32)
The fluctuations of the pearl number smear out the force oscillations for $`p>\mathrm{\Lambda }^{2/3}`$ as found previously. The fluctuations of the pearl sizes are reflected by the prefactors $`p^{1/2}(l_\mathrm{B}f^2/b)^{2p/3}`$. The sums are evaluated numerically for certain values. The result is displayed on Fig.8. As in our simple argument where the fluctuations were ignored, the force shows oscillations with unstable decreasing branches that should be replaced by plateaus or pseudo-plateaus using the Maxwell construction. In the limit of very large $`\mathrm{\Lambda }`$, the decrease of the force becomes very steep and the force on each branch tends to the value given by equation 28.
## IV Localized counterions: very poor solvent and counterion condensation on the pearls
So far, the counterions of the necklace chain have not been taken into account and have been considered as forming a free ideal solution. There are however cases where this assumption fails.
When the solvent is very poor, the pearls are very dense and the local dielectric constant inside a pearl is lower than that of water by typically one order of magnitude, there are essentially no dissociated charges in the pearl core. Charged systems with low dielectric constants, called ionomers, have been extensively studied . We follow in this section the lines of a simple model proposed by Dobrynin .
The electrostatic potential at the surface of a pearl $`\tau (l_\mathrm{B}/f)^{1/3}`$ increases with decreasing solvent quality and decreasing charge fraction. A condensation of the counterions and thus a regulation of the charge of the pearls is expected when this potential becomes of order $`k_\mathrm{B}T`$ i.e. when $`\tau ^3>fb/l_\mathrm{B}`$. The necklace becomes then unstable and collapses to form a compact globule. We now discuss this instability and the effect of an external force.
### A Condensation on the pearls
We assume that the counterions condense to a certain amount on the pearls but that they do not condense on the strings, i.e., the linear charge density along the strings is smaller than the Bjerrum length, which leads to $`f<\tau (b/l_\mathrm{B})`$. Counterion condensation then only takes place on the pearls. In the simplest description , the pearls have an effective charge $`Q_\mathrm{e}`$ determined by equilibrating the chemical potential of the counterions to bulk value of the order of the thermal energy $`k_\mathrm{B}T`$.
The pearl-necklace structure is based on a Rayleigh instability. We first discuss the effect of condensation for a ”liquid of disconnected blobs”. Let us call $`\alpha k_\mathrm{B}T`$ the chemical potential of the free counterions and $`Q_\mathrm{e}`$ the net charge of a droplet of radius $`R`$ and of nominal charge $`Q=f\tau R^3/b^3`$, the free energy of an assembly of $`\stackrel{~}{N}`$ blobs forming p monodisperse non-interacting droplets of radius $`R`$ is
$$F=p\gamma R^2+k_\mathrm{B}Tp\frac{Q_\mathrm{e}^2l_\mathrm{B}}{R}+p\alpha k_\mathrm{B}T(QQ_\mathrm{e}).$$
It includes the surface tension energy, the electrostatic energy of each drop and a chemical potential term for the counterions. After minimization with respect to the effective charge and taking into account the fact that the effective charge must be smaller than the nominal charge we obtain
$`F\tau ^3/\stackrel{~}{N}k_\mathrm{B}T`$ $`=`$ $`{\displaystyle \frac{\tau ^2b}{R}}+{\displaystyle \frac{l_Bf^2R^2\tau ^2}{2b^3}}(R<R_{co})`$ (33)
$`F\tau ^3/\stackrel{~}{N}k_\mathrm{B}T`$ $`=`$ $`{\displaystyle \frac{\tau ^2b}{R}}{\displaystyle \frac{\alpha ^2b^3}{2l_BR^2}}+f\tau \alpha (R>R_{co})`$ (34)
charge regulation takes place for $`R>R_{\mathrm{co}}`$ where $`R_{\mathrm{co}}^2=(\alpha b^3)/(f\tau l_\mathrm{B})`$.
If the fraction of charged monomers is large enough, $`\alpha ^3bf/l_b\tau ^3>1`$, in the regime where there is no counterion condensation, the free energy per blob has a minimum at $`R_{min}=b(l_Bf^2/b)^{1/3}`$ (the electrostatic blob size). The associated droplet radius is such that $`R_{min}<R_{co}`$ and the minimum actually exists. In the same range of parameters, in the regime where the counterions condense on the spheres the free energy per blob has a maximum in the charge regulated regime at $`R_{\mathrm{max}}=\alpha ^2b^2/\tau ^2l_\mathrm{B}`$. The minimum at the preferred pearl size is then stable against the infinite globule. In the opposite limit $`\alpha ^3bf/l_\mathrm{B}\tau ^3<1`$, the free energy per blob decreases monotonically with the radius of the droplets and the largest possible globule is stable. In this case, counterion condensation prevents the Rayleigh instability. The energy per blob is sketched in Fig.9 for both cases.
The structure of a necklace when there is no counterion condensation on the pearls has been described in the previous sections. The argument on the Rayleigh instability suggests that charge regulation and collapse occur simultaneously, as stated by Dobrynin et al. and Schiessel and Pincus following similar arguments. The situation is also reminiscent of that of cylindrical globules studied by Khokhlov where Manning condensation provokes collapse.
We now turn to the case of a polyelectrolyte chain and consider a necklace structure with condensed counterions on the pearls. The total free energy can be written as
$`{\displaystyle \frac{G}{k_\mathrm{B}T}}=p\gamma R^2+(p1)\gamma ld+p{\displaystyle \frac{l_\mathrm{B}Q_\mathrm{e}^2}{R}}+`$ (35)
$`+(p1){\displaystyle \frac{l_\mathrm{B}f^2m_s^2}{l}}`$ $`+`$ $`{\displaystyle \frac{l_\mathrm{B}p^2Q_\mathrm{e}^2}{L}}\mathrm{log}p+{\displaystyle \frac{L^2}{Mb^2}}+\alpha p(QQ_\mathrm{e})\phi L`$ (36)
where the first term is the surface contribution of the $`p`$ pearls, the second term the surface free energy of $`p1`$ strings of length $`l`$ and thickness $`d`$, the third term is the electrostatic self-energy of the pearls, followed by the electrostatic self-energy of the strings containing $`m_s`$ monomers. The fifth term is the total electrostatic interaction between the pearls where the total length of the necklace is $`L(p1)l`$. The total elastic free energy contribution yields the Gaussian term and finally the next term is the contribution of the free counterions, the last term is the contribution of the external force. The minimization of the free energy eq.(35) with respect to $`Q_\mathrm{e}`$ gives the effective charge as a function of the radius $`R`$ of the pearls and of the distance $`l`$ between pearls , the onset of charge regulation defined as $`Q_\mathrm{e}=Q`$ is located at $`Q=\frac{\alpha R}{l_B(1+\frac{2R}{l+R}\mathrm{log}p)}`$. Due to the interactions between pearls, the condensation threshold is shifted to a lower pearl charge; the effect depends on the force through $`l`$ and $`p`$ and decreases for increasing forces. However this is a small correction for high values of $`\mathrm{\Lambda }`$. Ignoring these corrections, the free energy can be minimized using the same set of variables as in section (II-B). The free energy does not have a minimum at any finite value of $`p`$ and is minimum for $`p=1`$ or $`p=0`$. As for the disconnected liquid of blobs, the necklace structure is unstable and as soon as the counterions condense on the pearls, the polyelectrolyte collapses. Within this description there is no regime of necklace stabilized by an external force. We thus conclude that at low charge fraction $`f<\tau ^3`$ where there would be charge regulation by the pearls, the dense globule is stable and stretches discontinuously when the force is equal to $`k_\mathrm{B}T\tau /b`$, the necklace structure is never stable.
### B Very poor solvent
In the preceding sections, we have assumed that the solvent is poor, but that the reduced excluded volume $`\tau `$ is smaller than unity, i.e., the correlation length remains much larger than the monomer size. In this case, one can assume that locally the dielectric constant is that of water, that the Bjerrum length has the water value and that the charges are dissociated. Consequently the counterions can be distributed inside the globule, almost freely. In a poorer solvent, the correlation length $`\xi _t`$ is of the order of the monomer size $`b`$ and this assumption on the dielectric constant cannot be made. Inside the dense phase, the solution is close to a polymer melt and the Bjerrum length decreases by one order of magnitude. This is also true well inside the pearls; the charges are then not dissociated and a typical ionomer behavior is expected. In the following brief discussion we assume that ion dissociation takes place only at the surface of the globules, and that only those groups localized in a thin shell of the order of the thermal blob size $`\xi _t`$ are dissociated and contribute to the pearl net charge.
The effective charge therefore becomes
$$Q_\mathrm{e}=fR^2\xi _tc=f\frac{R^2}{b^2},$$
(37)
where the concentration inside the globule is $`cb^3`$. The balance between the electrostatic free energy due to this effective charge and the surface free energy yields the size of the pearls
$$R=b\frac{b}{l_\mathrm{B}}\left(\frac{\tau }{f}\right)^2.$$
(38)
which enables us to compute the effective pearl charge and pearl electrostatic potential:
$`Q_\mathrm{e}=\left({\displaystyle \frac{b}{l_\mathrm{B}}}\right)^2\tau ^4f^3,`$ (39)
$`U_\mathrm{s}=(l_\mathrm{B}Q_\mathrm{e})/R{\displaystyle \frac{\tau ^2}{f}}`$ (40)
These expressions are only valid in the range of values of $`\tau `$ close to 1 and if part of the dissociated ions do not condense in the outer pearl shell, i.e., $`f>\tau ^2`$. On the other hand, the pearl size $`\xi _t`$ must be larger than the string, diameter i.e., the thermal blob size $`\xi _t`$, which yields an upper bound of the form $`f<\tau ^{3/2}(b/l_\mathrm{B})^{1/2}`$. The charge fraction must thus be in the range
$$\tau ^2<f<\tau ^{3/2}(b/l_\mathrm{B})^{1/2}.$$
(41)
The number of monomers in the pearl is then $`m=R^3c`$, giving
$$m=\left(\frac{b}{l_\mathrm{B}}\right)^3\frac{\tau ^7}{f^6}.$$
Again the string length between two pearls can be computed by the force balance. It is given by
$$l=b\left(\frac{l_\mathrm{B}}{b}\right)^{3/2}\left(\frac{\tau ^7}{f^6}\right)^{1/2}$$
and the calculation of the mass of the strings $`m_\mathrm{s}=ld^2c`$ shows that still most charges are indeed in the pearls. If an external force is applied, the force balance has to be modified and the force-extension law is
$$L=N\left(\frac{b}{\tau /b\phi }\right)^{1/2}\left(\frac{l_\mathrm{B}}{b}\right)^{3/2}\frac{f^6}{\tau ^6}.$$
(42)
This last equation is only valid as long as the pearl mass is much larger than the string mass, i.e., as long as
$$b\left(\frac{\tau }{b}\right)\phi >\left(\frac{l_\mathrm{B}}{b}\right)^3\frac{f^6}{\tau ^8}$$
For $`f<\tau ^2`$ however part of the counterions have to condense onto the outer pearl shell of thickness $`\xi _t`$ in order to decrease the surface potential back to $`k_\mathrm{B}T`$. In that case the results of the previous section where charge regulation by the pearls was already considered apply. Additionally we remark that the range of parameters is small for the validity of these scaling arguments. Within the presented approach, we cannot guarantee that the necklaces are stable. This remark does not at all affect the regimes discussed, e.g., in
## V Conclusion
A polyelectrolyte globule in a poor solvent has been shown by Kantor and Kardar and Dobrynin and Rubinstein to elongate with increasing charge and to have a necklace structure comprising smaller ”pearls” separated by stretched strings quite similarly to the so-called Rayleigh instability of charged liquid droplets. If we neglect the interaction with the charged monomers of the strings and of the other drops,, the pearl size is fixed by a Rayleigh criterion, $`l_BQ^2/R\tau R^2/b^2`$. Most monomers are in pearls whereas the strings make most of the chain length. In this paper we discussed theoretically pearl-necklace chain elasticity.
We first developed a continuous picture, valid for a necklace comprising many pearls and where single pearl features are ignored. When the necklace elongates under the action of an external force, the pearls are unwinded and converted into stretched strings. During this process the thermodynamic equilibrium between pearl and string monomers is preserved; this imposes the string tension to $`k_\mathrm{B}T\tau /b`$ and the string diameter, ($`b/\tau )`$. At equilibrium the tension in a string balances the repulsion between half-necklaces, and the external force. This force balance leads to the simple force/length relation: $`\phi =k_\mathrm{B}T(\tau /bl_\mathrm{B}f^2N^2/L^2)`$. During the unwinding process the length increases by a factor $`\mathrm{\Lambda }^{1/2}`$ where $`\mathrm{\Lambda }>>1`$ measures the number of thermal blobs in a pearl (or equivalently $`\mathrm{\Lambda }`$ is the inverse of the electrostatic interaction interaction between thermal blobs). This is in contrast with the earlier cylindrical model which predicts that when a force is applied to a spherical globule, the globule becomes unstable when it is elongated by a factor of order two.
In practice, the number of pearls is not always very large (as seen in several numerical simulations) and the discreteness of the pearls becomes an important factor. We have discussed in details a necklace comprising two pearls. When the external force is increased, a well defined metastable single string state appears in addition to the two pearl necklace. For a somewhat higher force, the single string and the two-pearl necklace exchange stability. At this critical force, the pearl size is still close to the pearl size in the absence of external force. When about half of the pearls is unwinded, the metastable two-pearls state disappears and the single string state is the only equilibrium conformation. For a fast stretch-collapse cycle, a hysteresis loop describing most of the metastable branches is expected. For quasi-static cycles, a force plateau is anticipated at coexistence (due to the one dimensional character of the problem the plateau is not associated to a real phase transition). For necklaces containing more pearls, our description uses a somewhat simpler model, where the intensive string properties, tension etc… are kept constant and where the only variables which are allowed to fluctuate are the number of monomers in a pearl and the number of pearls; this allows a transfer of the monomers from the pearls towards the strings as the applied force is increased. In this model the number of pearls is bound to be an integer. In a first approximation neglecting thermal fluctuations, an abrupt decrease of the pearl number by one unit is predicted for a given length (sequence of lengths). This formally corresponds to a abrupt decrease of the force. There is thus a sequence of Maxwell loops, corresponding to force plateaus in a quasi static transformation, or to a series of spinodal length jumps under increasing applied force. This picture is altered by fluctuations which are usually expected to be important and that smoothen the plateaus.
The pearl-necklace structure of polyelectrolytes relies on a steepest descent description which becomes exact in the thermodynamic limit for macroscopic systems. In the usual Rayleigh instability of small (yet macroscopic) droplets the energy scale, say typically the surface energy of a marginally stable droplet, is large compared to $`k_\mathrm{B}T`$. For mesoscopic systems such as the polyelectrolyte necklace, this is not always the case and thermal fluctuations are often important. In the case of a two pearl necklace we have computed the average length , and the length fluctuation when the necklace is subject to an external force. A pseudo-plateau is indeed found for the force but even for a pearl self-energy as high as $`100k_\mathrm{B}T`$ the plateau is strongly rounded and the length fluctuations in the plateau are roughly as large as the average length. If the necklace contains many pearls, the fluctuations have also been taken into account. The force is the sum of a continuous component that can be calculated from the continuous model and a component arising from the discreteness of the pearl number. This last contribution is non monotonic and can lead to Maxwell loops. However if the self-energy of a pearl is small or if the network comprises many pearls the plateaus are strongly rounded by fluctuations and significant force (pseudo) plateaus are no longer expected.
In the pearl-necklace structure, the electrostatic charge on the pearls is rather high and, even though there is no Manning condensation on the strings or on the smeared out necklace charge, condensation on the pearls (or charge regulation) can occur. In agreement with previous work, charge regulation is found for $`f<\tau ^3`$ and the polyelectrolyte collapses into a large globule whenever counterion condensation becomes important. If an external force is applied, the collapsed globule discontinuously stretches at a critical tension of order $`k_\mathrm{B}T\tau /b`$. The extreme case of a very poor solvent where there is no ion dissociation in the (ionomer-like) dense phase has also been discussed as it is of some experimental relevance. In this case stable necklaces are obtained and the pearls unwind under an external force.
As a remark aside we would like to mention, that we ignored the effect of salt which might be added to the solution. This will be part of a different work. However, we expect that the results presented here are valid as long as the persistence length (including the electrostatic contribution) together with the Debye screening length is larger than $`l`$, i.e., the distance between two pearls.
Besides direct single chain elasticity measurements and implications for drag reduction in polyelectrolyte solutions and related topics, the single chain elasticity studied in this paper can be used as a basis to investigate more complex systems. In a separate work, we will present results on polyelectrolyte gel elasticity in a poor solvent following Katchalsky’s arguments.
### Acknowledgments
T.A.V. acknowledges the financial support of the LEA. The hospitality of the Institut Charles Sadron is gratefully appreciated. The authors acknowledge enjoy- and helpful discussions with Phil Pincus. They are also grateful to H. Schiessel, who sent his preprint prior to publication.
Note added: After submission of this work, we received a preprint by H. Schiessel . In this work the same topic is discussed. The free energies are discussed numerically and the results are complementary to those discussed here. |
warning/0001/hep-th0001218.html | ar5iv | text | # References
hep-th/0001218
BFV–BRST analysis of equivalence between noncommutative and ordinary gauge theories
Ömer F. DAYI
Physics Department, Faculty of Science and Letters, Istanbul Technical University, 80626 Maslak, Istanbul, Turkey
and
Feza Gürsey Institute, P.O.Box 6, 81220 Çengelköy, Istanbul, Turkey. <sup>1</sup><sup>1</sup>1E-mails: dayi@itu.edu.tr and dayi@gursey.gov.tr.
Constrained hamiltonian structure of noncommutative gauge theory for the gauge group $`U(1)`$ is discussed. Constraints are shown to be first class, although, they do not give an Abelian algebra in terms of Poisson brackets. The related BFV-BRST charge gives a vanishing generalized Poisson bracket by itself due to the associativity of $``$–product. Equivalence of noncommutative and ordinary gauge theories is formulated in generalized phase space by using BFV-BRST charge and a solution is obtained. Gauge fixing is discussed.
Noncommutative geometry is one of the most attractive subjects of physics and mathematics. Attention to it is increased considerably after the observation that in string theories noncommutativity of space appears in a natural way. Seiberg and Witten used the ideas about the noncommutativity in string theory and showed that noncommutative and ordinary gauge theories are equivalent. Following them there appeared several works on noncommutative gauge theories. BRST symmetry of a noncommutative gauge theory in first order Lagrangian formulation was discussed in .
When one deals with a gauge theory, Lagrangian framework is suitable for perturbative calculations. However, understanding its Hamiltonian structure is essential to perform canonical quantization which can be used to derive some features like the correct measure of path integral and physical states.
In Hamiltonian framework, an ordinary gauge theory action leads to first class constraints which decrease the number of linearly independent phase space variables. For keeping track of gauge invariance, instead of getting rid of the unphysical degrees of freedom, one enlarges the phase space by introducing ghost fields possessing the opposite statistics of the constraints and write the BVF–BRST charge. Now, invariance of the theory under the transformation of variables generated by the constraints (reminiscent of gauge invariance) replaced by the invariance under the transformations generated by the BFV–BRST charge. Action and measure of the related path integrals should be invariant under the BFV–BRST transformations. Once the canonical commutation relations between phase space fields are imposed, they can be written in terms of their normal modes. Thus, one can find the quantum BFV–BRST charge which is defined to be nilpotent and whose cohomology gives the physical states.
One expects that canonical formulation of noncommutative gauge theories can be studied in a similar manner, as far as the time coordinate is kept classical. We show that, indeed, this is the case, although, it was not guaranteed a priori: Gauge invariance of noncommutative gauge theory is not an invariance of the Lagrange density, in contrary to the ordinary case, but an invariance of the action,
Hamiltonian structure of ordinary gauge theories are well studied. Thus, we hope that showing the equivalence between ordinary and noncommutative gauge theories in terms of the generalized phase space variables will be useful in understanding features like canonical quantization of the latter.
Constrained hamiltonian structure of noncommutative gauge theory for the gauge group $`U(1)`$ is studied. This is the simplest gauge group, however, the main features of noncommuting gauge theories are already present. The related BFV-BRST charge which gives a vanishing generalized Poisson bracket with itself is presented. Equivalence of ordinary and noncommutative gauge theories is formulated in terms of generalized phase space variables and a solution is given. We also briefly discuss gauge fixing within this formalism.
Noncommutative gauge theory for $`U(1)`$ in $`^d`$ is given with the action
$$S=\frac{1}{4}d^dxF_{\mu \nu }F^{\mu \nu }$$
(1)
in terms of the metric $`\eta _{\mu \nu }=\mathrm{diag}(1,1,\mathrm{}),`$ and
$$F^{\mu \nu }(x)=^\mu A^\nu ^\nu A^\mu i\left(A^\mu A^\nu A^\nu A^\mu \right).$$
$``$–product is defined as
$$G(x)K(x)=G(x)e^{\frac{i}{2}\theta ^{ij}\stackrel{}{_i}\stackrel{}{_j}}K(x),$$
(2)
where $`\theta ^{ij}`$ is an antisymmetric, constant matrix. The $``$–product reflects the fact that the space coordinates $`x_i`$ are noncommuting:
$$x^ix^jx^jx^i=i\theta ^{ij}.$$
(1) is invariant under the gauge transformations
$$\delta A_\mu =_\mu \lambda +i(\lambda A_\mu A_\mu \lambda ),$$
because we have the equalities
$`{\displaystyle d^{d1}xG(x)\left(K(x)L(x)\right)}`$ $`=`$ $`{\displaystyle d^{d1}xG(x)\left(K(x)L(x)\right)}`$
$`=`$ $`{\displaystyle d^{d1}x\left(L(x)G(x)\right)K(x)},`$
$`=`$ $`{\displaystyle d^{d1}x\left(G(x)K(x)\right)L(x)},`$
which follow from the assumption that all of the fields which we deal with are vanishing at infinity.
The time coordinate is not deformed and noncommutativity is due to the $``$–product. Hence, the fields are ordinary ones and the canonical momenta are defined as usual:
$`P_0={\displaystyle \frac{S}{\dot{A}^0}}`$ $`=`$ $`0,`$ (3)
$`P_i={\displaystyle \frac{S}{\dot{A}^i}}`$ $`=`$ $`_0A_i_iA_0i(A_0A_iA_iA_0).`$ (4)
The canonical hamiltonian reads
$$H=d^{d1}x\left(\frac{1}{2}P_i^2+\frac{1}{4}F_{ij}F^{ij}A_0\mathrm{\Phi }(x)\right),$$
(5)
where
$$\mathrm{\Phi }(x)=_iP^i+i[P_i,A^i],$$
(6)
written in terms of the Moyal bracket
$$[G(x),K(x)]G(x)K(x)K(x)G(x).$$
The primary constraints (3) should be constant in time:
$$\{P_0(x),H\}=0,$$
(7)
where the basic Poisson brackets are
$$\{P_\mu (x),A_\nu (y)\}=\eta _{\mu \nu }\delta (xy).$$
The condition (7) leads to the secondary constraints
$$\mathrm{\Phi }(x)=0.$$
(8)
Obviously, $`P_0`$ gives a vanishing Poisson bracket with $`\mathrm{\Phi }(x).`$ To classify the constraints $`\mathrm{\Phi }(x)`$ let us introduce a bosonic parameter $`\lambda (x)`$ and deal with
$`\mathrm{\Phi }_\lambda `$ $``$ $`{\displaystyle d^{d1}x\mathrm{\Phi }(x)\lambda (x)}`$
$`=`$ $`{\displaystyle d^{d1}x\mathrm{\Phi }(x)\lambda (x)}`$
$`=`$ $`{\displaystyle d^{d1}xP_i\left(_i\lambda (x)+i[P_i(x),\lambda (x)]\right)}.`$
The last equality follows from the fact that the Moyal bracket in (6) possesses only odd powers of $`\theta ^{ij}.`$ Poisson bracket of the integrated constraints is calculated:
$$\{\mathrm{\Phi }_\lambda ,\mathrm{\Phi }_\kappa \}=i\mathrm{\Phi }_{[\lambda ,\kappa ]},$$
(9)
where on the right hand side parameter is the Moyal bracket of the ones appearing on the left. Moreover, $`\mathrm{\Phi }(x)`$ do not lead to new constraints:
$$\{H_0,\mathrm{\Phi }_\lambda \}=0,$$
where $`H_0=H|_{\mathrm{\Phi }=0}.`$ Hence the whole set of constraints is given by (3) and (6) which are first class.
For the following discussions the primary constraints (3) are not essential. Hence, we do not deal with them any more by fixing the gauge as $`A_0=0`$ and setting $`P_0(x)=0.`$
To perform the BFV-BRST analysis, we enlarge the phase space with the anticommuting ghost fileds $`C(x)`$ and their canonical conjugates $`P_C(x)`$ and use the generalized Poisson bracket
$$\{G,K\}=d^{d1}x\left(\frac{_rG}{𝒫(x)}\frac{_lK}{𝒬(x)}(1)^{\eta (G)\eta (K)}\frac{_rK}{𝒫}\frac{_lG}{𝒬}\right).$$
(10)
Here the subscripts $`r`$ and $`l`$ indicate right and left derivatives, $`\eta (G)`$ Grassmann parity of $`G`$ and $`𝒬,𝒫`$ denote the generalized phase space coordinates and momenta including the original ones and the ghosts. We attribute also ghost numbers
$$\mathrm{gh}(C)=\mathrm{gh}(P_C)=1,\mathrm{gh}(A_i)=\mathrm{gh}(P_i)=0.$$
The BFV-BRST charge is
$`\mathrm{\Omega }`$ $`=`$ $`{\displaystyle d^{d1}x\left(\mathrm{\Phi }(x)C(x)P_C(x)\left(C(x)C(x)\right)\right)}`$ (11)
$`=`$ $`{\displaystyle d^{d1}x\left(\mathrm{\Phi }(x)C(x)\frac{1}{2}(P_C(x)C(x)+C(x)P_C)C(x)\right)}.`$
To deal with fermionic fields, we generalize the Moyal bracket as
$$[G,K]GK()^{\eta (G)\eta (K)}KG.$$
Obviously, the generalized Moyal brackets satisfy the generalized Jacobi identity. Thus,
$$d^{d1}x(CC)[P_C,C]=\frac{1}{3}d^{d1}x\left([[C,C],P_C]+[[C,P_C],C]+[[P_C,C],C]\right)C=0,$$
which is the unique nontrivial term to conclude that the BFV-BRST charge satisfies
$$\{\mathrm{\Omega },\mathrm{\Omega }\}=0.$$
(12)
By introducing the rigid, fermionic parameter $`ϵ`$ possessing ghost number $`1,`$ the phase space fields transform as
$`\delta _ϵA_i`$ $`=`$ $`ϵ\left(_iC+iCA_iiA_iC\right),`$ (13)
$`\delta _ϵC`$ $`=`$ $`iϵCC,`$ (14)
$`\delta _ϵP_i`$ $`=`$ $`iϵ\left(P_iCCP_i\right),`$ (15)
$`\delta _ϵP_C`$ $`=`$ $`ϵ\left(\mathrm{\Phi }+iP_CC+iCP_C\right),`$ (16)
under the BFV-BRST charge (11).
Now, we would like to discuss the equivalence of noncommutative and ordinary gauge theories. Ordinary gauge theory action for the gauge group $`U(1)`$ is given in terms of the field strength $`f_{\mu \nu }=_\mu a_\nu _\nu a_\mu ,`$ as
$$S_o=d^dxf_{\mu \nu }f^{\mu \nu }.$$
(17)
The definition of canonical momenta $`p_\mu ,`$ yields the primary constraints $`p_0(x)=0.`$ These lead to the secondary constraints $`_ip_i=0,`$ which are first class. We set strongly $`p_0=0`$ by fixing the gauge as $`a_0=0.`$ Now, as usual we enlarge the phase space with the anticommuting ghost field $`c(x)`$ and its canonical conjugate $`p_c(x),`$ possessing ghost numbers $`1`$ and $`1`$ to write the BFV-BRST charge as
$$\mathrm{\Omega }^o=d^{d1}x_ip^ic.$$
(18)
Thus the BFV-BRST transformation of the fields is given as
$`\delta _ϵ^oa_i=ϵ_ic`$ , $`\delta _ϵ^oc=0,`$ (19)
$`\delta _ϵ^op_i=0`$ , $`\delta _ϵ^op_c=ϵ_ip_i.`$ (20)
Let us denote the phase space variables in a unified notation as
$$Q_z(A_i,P_i,C,P_C),q_z(a_i,p_i,c,p_c).$$
(21)
Then, we can formulate equivalence of the noncommutative and ordinary gauge theories as
$$Q_z(q)+\delta _ϵQ_z(q)=Q_z(q+\delta _ϵ^oq).$$
(22)
Moreover, there are the conditions
$$Q_z(q)|_{\theta =0}=q_z.$$
(23)
To first order in $`\theta ^{kl},`$ there is a solution of (22)–(23):
$`A_i(a)`$ $`=`$ $`a_i\theta ^{kl}(a_k_la_i{\displaystyle \frac{1}{2}}a_k_ia_l),`$ (24)
$`P_i(a,p)`$ $`=`$ $`p_i\theta ^{kl}a_k_lp_i,`$ (25)
$`C(a,c)`$ $`=`$ $`c+{\displaystyle \frac{1}{2}}\theta ^{kl}_kca_l`$ (26)
$`P_C(a,p,p_c)`$ $`=`$ $`p_c+p_c(_jp^j)^1\theta ^{kl}_kp_if_{il}+\theta ^{kl}_kp_ca_l.`$ (27)
As expected, the solutions $`A_i(a)`$ (24) and $`C(a,c)`$ (26) can be derived by using the solution for $`A_i`$ and the gauge parameter $`\lambda `$ given in . At first glance (27) can be thought of being inconsistent with the constraint $`_ip^i=0`$ of the ordinary $`U(1)`$ gauge theory. However, this is due to the fact that noncommutative and ordinary gauge groups can not be isomorphic to each other: when one deals with the noncommutative gauge theory the constraint $`\mathrm{\Phi }=0`$ leads to $`\theta ^{ij}=0`$ for $`_ip^i=0`$ and arbitrary gauge field $`a_i.`$ There is another way of explaining this fact: observe that the BRST transformations of $`P_C`$ and $`p_c`$ (16), (20) are proportional to the related constraints, so that they are aware of the structure of the gauge group. Thus, as far as we deal with nonvanishing $`\theta ^{ij}`$ we assume that $`_ip^i0.`$
For performing gauge fixing one enlarges the phase space with the bosonic and fermionic canonical conjugate pairs $`(\mathrm{\Lambda },\mathrm{\Pi })`$ and $`(\overline{C},\overline{P}_C)`$ satisfying the generalized Poisson bracket relations
$`\{\mathrm{\Lambda }(x),\mathrm{\Pi }(y)\}=\delta (xy)`$ ; $`\{\overline{C}(x),\overline{P}_C(y)\}=\delta (xy).`$ (28)
$`\mathrm{\Lambda },\mathrm{\Pi }`$ possess zero ghost number and
$$\mathrm{gh}(\overline{C})=\mathrm{gh}(\overline{P}_C)=1.$$
Gauge fixed action which can be used in the related path integral can be given in terms of the gauge fixed hamiltonian
$$H=H_0+\{\mathrm{\Psi },\mathrm{\Omega }^{}\},$$
where $`\mathrm{\Psi }`$ is the gauge fixing fermion possessing ghost number $`1`$ and
$$\mathrm{\Omega }^{}=\mathrm{\Omega }+d^{d1}x\mathrm{\Pi }\overline{P}_C.$$
The easist choice for the gauge fixing fermion $`\mathrm{\Psi }`$ is
$$\mathrm{\Psi }=d^{d1}\left(P_C\xi +\overline{C}\mathrm{\Lambda }\right),$$
(29)
where $`\xi `$ is a function of the original fileds and defined to possess a nonvanishing Poisson bracket with $`\mathrm{\Phi }(x).`$ The gauge fixed hamiltonian in this gauge becomes
$$H=H_0+d^{d1}x\left(C\{\mathrm{\Phi },\xi \}\overline{C}+\mathrm{\Lambda }\left(\mathrm{\Phi }+[C,P_C]\right)+\overline{P}_CP_C+\mathrm{\Pi }\xi \right).$$
(30)
By choosing a gauge condition $`\xi (x)`$ we can perform perturbative calculations within the Hamiltonian approach to study noncommutative gauge theory and its relations with ordinary gauge theory by making use the solutions (24)–(27).
By using the constrained Hamiltonian structure presented, one can also proceed to perform operator quantization by expanding the generalized phase space variables in terms of normal modes. |
warning/0001/cond-mat0001019.html | ar5iv | text | # Analytic Lyapunov exponents in a classical nonlinear field equation
## Abstract
It is shown that the nonlinear wave equation $`_t^2\phi _x^2\phi \mu _0_x(_x\phi )^3=0`$, which is the continuum limit of the Fermi-Pasta-Ulam (FPU) $`\beta `$ model, has a positive Lyapunov exponent $`\lambda _1`$, whose analytic energy dependence is given. The result (a first example for field equations) is achieved by evaluating the lattice-spacing dependence of $`\lambda _1`$ for the FPU model within the framework of a Riemannian description of Hamiltonian chaos. We also discuss a difficulty of the statistical mechanical treatment of this classical field system, which is absent in the dynamical description.
The numerical study that E.Fermi, J.Pasta and S.Ulam made at Los Alamos, on a chain of anharmonically coupled oscillators, represents a milestone in the development of the modern theory of nonlinear dynamics . The surprising outcomes of this first computer experiment in the history of physics have been very seminal. Among the other attempts to explain the apparent lack of thermalization of the energy initially stored in one of the linear modes of the chain, Zabusky and Kruskal remarked that a continuum limit version of the Fermi-Pasta-Ulam (FPU) model leads to the Korteweg-de-Vries (KdV) equation, when the $`\alpha `$ model is considered, and to a modified KdV equation when the $`\beta `$ model is considered. These are nonlinear, integrable partial differential equations where a special class of solitary waves – that Zabusky and Kruskal called solitons – exists and can to some extent explain the FPU recurrences .
In the present work we tackle the FPU-$`\beta `$ model, described by the Hamiltonian
$$H(p,q)=\underset{i=1}{\overset{N}{}}\left[\frac{1}{2}p_i^2+\frac{1}{2}(q_{i+1}q_i)^2+\frac{\mu }{4}(q_{i+1}q_i)^4\right],$$
(1)
and we consider another legitimate continuum limit of this system, described by the Hamiltonian
$$H(\pi ,\varphi )=_0^L𝑑x\left[\frac{1}{2\nu }\pi (x)^2+\frac{\eta }{2}(\varphi )^2+\frac{\mu _0}{4}(\varphi )^4\right],$$
(2)
and leading to the field equation (setting $`\eta =\nu =1`$)
$$\frac{^2\varphi }{x^2}\frac{^2\varphi }{t^2}+\mu _0\frac{}{x}\left(\frac{\varphi }{x}\right)^3=0.$$
(3)
Also this continuum limit of the FPU model might appear integrable on the basis of the following reasoning . Let us consider the Legendre transform: $`\chi =_x\varphi `$, $`\tau =_t\varphi `$ and $`x=_\chi \psi `$, $`t=_\tau \psi `$ with the relation $`\psi +\varphi =x\chi +t\tau `$, whence $`_x^2\varphi =D_\tau ^2\psi `$, $`_t^2\varphi =D_\chi ^2\psi `$, where $`1/D=_\chi ^2\psi _\tau ^2\psi (_{\chi \tau }^2\psi )^2`$. Substituting into Eq.(3) we can linearize it to
$$\frac{^2\psi }{\chi ^2}=(1+3\mu _0\chi ^2)\frac{^2\psi }{\tau ^2}$$
(4)
which can be solved by variable separation by setting $`\psi =F(\chi )G(\tau )`$, where $`G^{\prime \prime }\pm c^2G=0`$, and $`F^{\prime \prime }\pm c^2(1+3\mu _0\chi ^2)F=0`$. In principle, by inverting this Legendre transform, Eq.(3) could be analytically solved. However, as we shall discuss, the method sketched above can only lead at most to local and not to global invertibility, both in time and space. Actually we are going to show that the field equation (3) for the FPU $`\beta `$ model in the continuum limit is chaotic (the Lyapunov characteristic exponent of the solutions is positive). The pattern of the Lyapunov exponent that we shall derive is shown in Fig.1. The cross-over in the energy dependence of the largest Lyapunov exponent has been attributed in Refs. to a (smooth) transition between weak and strong chaos . The persistence of such a cross-over also in the continuum limit is a remarkable fact. Loosely speaking, at high energy, in the strongly chaotic regime, the spatio-temporal behavior of the field $`\varphi (x,t)`$ should look like a fully developed “turbulent” field, whereas this should not be the case at low energy, where a weakly “turbulent” dynamics would set in. Our calculation shows that, no matter whether weak or strong, chaos is present in the system at any energy, and since the system is Hamiltonian, we can believe that a standard statistical mechanical description is allowed. This raises an important problem that we now discuss.
For the lattice system corresponding to (2) the standard canonical partition function $`Z=_{i=1}^Nd\pi _id\varphi _i\mathrm{exp}\left[\beta H(\{\pi _i\},\{\varphi _i\})\right]`$ in the limit of lattice spacing $`a0`$ becomes
$$𝒵=𝒟\pi (x)𝒟\varphi (x)e^{\beta _0^L𝑑x\left[\frac{1}{2}\pi ^2+\frac{1}{2}(\varphi )^2+\frac{\mu _0}{4}(\varphi )^4\right]}.$$
(5)
It is interesting at this point to raise a problem that comes out when some expected dynamical properties of the field $`\varphi (x,t)`$ are compared to their statistical counterparts worked out by means of $`𝒵`$. By means of an orthogonal change of coordinates, the Hamiltonian (1) is rewritten in the form $`H(P_k,Q_k)=_kE_k=_k[\frac{1}{2}(P_k^2+\omega _k^2Q_k^2)+\mu _{k_1,k_2,k_3}C(k,k_1,k_2,k_3)Q_kQ_{k_1}Q_{k_2}Q_{k_3}]`$ where the summations run from $`k=1`$ to $`k=k_{max}`$ for a lattice, and from $`k_0`$ to $`k_{max}=\mathrm{}`$ in the continuum limit. In the new variables the partition function becomes $`\stackrel{~}{Z}=_{k=1}^{k_{max}}dP_kdQ_k\mathrm{exp}\left[\beta H(\{P_k\},\{Q_k\})\right]`$ which, in the limit $`k_{max}\mathrm{}`$, gives $`\stackrel{~}{𝒵}=𝒟P(k)𝒟Q(k)\mathrm{exp}\left[\beta _{k_0}^{\mathrm{}}𝑑kH(P(k),Q(k))\right]`$, where $`k_0=\frac{2\pi }{L}>0`$ for a finite support. The generalized equipartition theorem states that $`Q_kH/Q_k=const`$ independently of $`k`$, and from Ref. we know that for the FPU $`\beta `$ model such an average is $`Q_kH/Q_kk^2Q_k^2(1+\alpha )`$, where $`\alpha `$ is a constant, and therefore $`Q_k^2k^2(1+\alpha )^1`$, which, being independent of $`k_{max}`$, still holds true in the continuum limit, i.e $`|\widehat{\varphi }(k)|^2k^2(1+\alpha )^1`$. Whence an ultraviolet catastrophe, i.e. the divergence of average total energy when $`k_{max}\mathrm{}`$ also on a finite support $`[0,L]`$. However such a difficulty is absent in the dynamical equation (3), which, being derived from a Hamiltonian, must keep the energy constant, therefore finite if it is finite at time $`t=0`$.
The point is that statistical mechanics and dynamics yield very different large-$`k`$ spectral properties. In fact, already on the basis of the purely dimensional analysis discussed in Refs., we can easily find that in order to bound the total energy (on a finite support $`[0,L]`$) to a finite value, the ultraviolet asymptotic power spectrum of $`\varphi (x)`$ must be bounded above by $`|\widehat{\varphi }(k)|^2k^3`$, which means that the high spatial-frequency modes cannot have the same energy of the low spatial-frequency modes.
The rough estimate of the ultraviolet spectrum of the continuum model can be improved as follows. Any smooth initial condition $`\varphi (x,0)𝒞^{\mathrm{}}()`$ of the equation of motion (3) has to remain smooth at any further time $`t`$, i.e. $`\varphi (x,t)𝒞^{\mathrm{}}()`$. Now, the analytic continuation $`\mathrm{\Phi }(z,t)`$ of $`\varphi (x,t)`$ will be of class $`𝒞^\omega `$ on a strip whose width $`\delta `$ will be determined at any given time $`t`$ by the distance of the closest singularity to the real axis. If $`z_n^s=x_n^s+iy_n^s`$ are the locations of the singularities of $`\mathrm{\Phi }(z,t)`$, to the leading order one finds $`\widehat{\mathrm{\Phi }}(k)k^\alpha _nc_ne^{ikz_n^s}`$ for $`k\mathrm{}`$, where, apart from the power-law prefactor $`k^\alpha `$, the sum is carried over the $`Imz>0`$ half plane and $`c_n`$ are suitable coefficients. The factors $`e^{ikz_n^s}=e^{ikx_n^s}e^{ky_n^s}`$ efficiently single out the closest singularity to the real axis to give $`|\widehat{\mathrm{\Phi }}(k)|^2Ae^{\delta k}`$ at $`k\mathrm{}`$ which provides a better ultraviolet estimate of $`|\widehat{\varphi }(k)|^2`$ stemming from the constraint of differentiability of $`\varphi (x)`$. Needless to say, this ultraviolet exponential fall off of $`|\widehat{\varphi }(k)|^2`$ guarantees the finiteness of the total energy for any finite support $`[0,L]`$. It is not unreasonable to think that a relationship might exist between the minimum width of the spatio-temporal cell where the Legendre transform of Eq.(3) - sketched at the beginning of this paper - is invertible, that is where Eq.(3) is locally integrable, and the width of the $`(k,\omega (k))`$ cell where the power spectrum $`|\widehat{\varphi }(k)|^2`$ exponentially falls off. There is apparently no way of reproducing the ultraviolet exponential decay of $`|\widehat{\varphi }(k)|^2`$ \- which is naturally brought about by the dynamical equation (3) - within the statistical mechanical framework. The field $`\varphi (x,t)`$ can be chaotic in space and time only down to some small scale which naturally provides a cutoff that, inserted into $`\stackrel{~}{𝒵}`$, would remove the mentioned divergences. Let us mention that the physical meaning of a rigorous treatment of another classical nonlinear field equation (complex Ginzburg-Landau) is coherent with the theoretical scenario depicted above.
To come now to our main result, chaoticity in the continuum limit, we calculate the Lyapunov exponent by extending to such a limit a method to tackle Hamiltonian chaos that has already given excellent analytic predictions of the largest Lyapounov exponent in the thermodynamic limit of the FPU $`\beta `$-model on a lattice. This method exploits the mathematical identification of an Hamiltonian flow with the geodesic flow on a suitable Riemannian “mechanical manifold” consisting of an enlarged configuration spacetime ($`\{q^0t,q^1,\mathrm{},q^N\}`$ plus one real coordinate $`q^{N+1}`$), whose arc-length $`ds^2=2V(𝐪)(dq^0)^2+a_{ij}dq^idq^j+2dq^0dq^{N+1}`$ defines the so-called Eisenhart metric $`g__E`$ ; $`V`$ is the potential. In the geometrical framework, the (in)stability of the trajectories is the (in)stability of the geodesics, and it is completely determined by the curvature properties of the underlying manifold according to the Jacobi equation for the geodesic deviation. This equation, written for Eisenhart metric, entails the usual tangent dynamics equation $`\ddot{\xi }^i+(^2V/q_iq^j)\xi ^j=0`$, which is used to measure Lyapunov exponents in standard Hamiltonian systems. Having recognized its geometric origin, it can be transformed, under certain hypotheses , into an effective scalar stability equation that, independently of the knowledge of dynamical trajectories, provides a measure of their average degree of instability. This effective stability equation is in the form of a stochastic oscillator equation $`\ddot{\psi }+[\kappa _0+\sigma _\kappa \eta (t)]\psi =0`$ , where $`\kappa _0=K_R/N`$, $`\sigma _\kappa ^2=(K_RK_R)^2/N`$, and $`K_R\mathrm{\Delta }V=_{i=1}^N^2V/q_i^2`$ is the Ricci curvature of the mechanical manifold, computed with $`g_E`$; $`\eta (t)`$ is a gaussian $`\delta `$-correlated random process of unit variance. The exponential growth rate $`\lambda `$ of $`(\dot{\psi }^2+\psi ^2)`$ is computed exactly to provide the following estimate of the largest Lyapunov exponent:
$$\lambda =\frac{\mathrm{\Lambda }}{2}\frac{2\kappa _0}{3\mathrm{\Lambda }},\mathrm{\Lambda }=\left[2\sigma _\kappa ^2\tau +\sqrt{\frac{4\kappa _0}{3}^3+4\sigma _\kappa ^4\tau ^2}\right]^{\frac{1}{3}}$$
(6)
where $`\tau =\pi \sqrt{\kappa _0}/(2[\kappa _0(\kappa _0+\sigma _\kappa )]^{\frac{1}{2}}+\pi \sigma _\kappa )`$ is a characteristic time scale worked out on the basis of geometrical arguments. In the limit $`\sigma _\kappa /\kappa _01`$ one finds $`\lambda \sigma _\kappa ^2`$ .
In this geometric picture chaos is mainly originated by the parametric instability activated by the fluctuating curvature “felt” by the geodesics.
The quantities $`\kappa _0`$ and $`\sigma _\kappa ^2`$ can be exactly computed for the FPU-$`\beta `$ model as microcanonical averages by taking advantage of the analytically known canonical partition function and by using the conversion formulas relating canonical and microcanonical averages. The final analytic expressions are given in Ref., and among them we report those needed here
$`{\displaystyle \frac{1}{N}}K_R^\mu (\theta )`$ $`=`$ $`2+{\displaystyle \frac{3}{\theta }}\mathrm{\Delta }(\theta )`$ (7)
$`{\displaystyle \frac{1}{N}}\delta ^2K_R^\mu (\theta )`$ $`=`$ $`{\displaystyle \frac{1}{N}}{\displaystyle \frac{9}{\theta ^2}}\left[22\theta \mathrm{\Delta }(\theta )\mathrm{\Delta }^2(\theta )\right]`$ (8)
$``$ $`{\displaystyle \frac{\beta ^2}{c_v(\theta )}}\left({\displaystyle \frac{K_R/N^G}{\beta }}\right)^2`$ (9)
$`ϵ(\theta )`$ $`=`$ $`{\displaystyle \frac{1}{8\mu }}\left[{\displaystyle \frac{3}{\theta ^2}}+{\displaystyle \frac{1}{\theta }}\mathrm{\Delta }(\theta )\right],`$ (10)
where $`\mathrm{\Delta }(\theta )=\frac{D_{3/2}(\theta )}{D_{1/2}(\theta )}`$, with $`D_{3/2}`$ and $`D_{1/2}`$ parabolic cylinder functions, the parameter $`\theta `$ is a function of the inverse temperature $`\beta `$ through $`\theta =(\beta /2\mu )^{1/2}`$, and $`ϵ(\theta )`$ is the energy per degree of freedom of the system. The microcanonical average of curvature fluctuations involves a correction term to their canonical average which involves the specific heat $`c_v(\theta )`$ computed, as usual, as the second derivative of the free energy. These quantities enter the formula (6) to give the analytic values of Lyapunov exponents at different energies. Now, it is useful to work out simple analytic expressions of $`\lambda _1(ϵ)`$ at low and high energy densities. In order to obtain them, notice that from Eq.(10) one gets $`\mu ϵ=f(\theta )`$, a function of $`\theta `$; this function is invertible, because in the absence of phase transitions $`ϵ(T)`$ ( $`T`$ is the temperature ) is always one-to-one. Therefore we can write $`\theta =f^1(\mu ϵ)`$ and consequently $`\kappa _0`$ and $`\sigma _\kappa ^2`$ can be expressed as functions of $`\mu ϵ`$. Making use of the asymptotic expressions of the parabolic cylinder functions, this inversion is easily obtained in the two opposite limits $`ϵ0`$ and $`ϵ\mathrm{}`$. Also simple analytic expressions for $`\kappa _0`$, $`\sigma _\kappa ^2`$ and $`\tau `$ can be obtained in these two limits. At $`ϵ0`$, and $`\theta 1`$, we get
$`\kappa _0`$ $`=`$ $`2+6\mu ϵ{\displaystyle \frac{27}{2}}\mu ^2ϵ^2+𝒪(ϵ^3)`$ (11)
$`\sigma _\kappa ^2`$ $`=`$ $`36\mu ^2ϵ^2+𝒪(ϵ^3).`$ (12)
Let us remark that there is some arbitrariness in the derivation of $`\tau `$ in Ref.; however from Eq.(6) we see that the only relevant constraint on $`\tau `$ is that $`\sigma _\kappa ^2\tau 0`$ for $`ϵ0`$, so that $`\mathrm{\Lambda }`$ – and $`\lambda _1`$ with it – can be expanded in series of the small quantity $`\sigma _\kappa ^2\tau `$ yielding
$$\lambda _1(ϵ)=\frac{\sigma _\kappa ^2\tau }{2\kappa _0}+\mathrm{}=9\mu ^2ϵ^2\tau +\mathrm{}$$
(13)
which, with the explicit computation of $`\tau `$ gives
$$\lambda _1(ϵ)=\frac{9}{4}\frac{\pi }{\sqrt{2}}\mu ^2ϵ^2+𝒪(ϵ^3)$$
(14)
which is in strikingly good agreement with the low energy “experimental” results for $`\lambda _1`$. Similar developments can be worked out in the limit $`ϵ\mathrm{}`$ (and so $`\theta 1`$) giving
$$\kappa _0\frac{4\pi }{\mathrm{\Gamma }^2\left(\frac{1}{4}\right)}\sqrt{24\mu ϵ},\sigma _\kappa ^216\mu ϵ\left(3\frac{32\pi ^2}{\mathrm{\Gamma }^4\left(\frac{1}{4}\right)}\right)$$
(15)
by computing again $`\tau (ϵ)`$ we find $`\tau (ϵ)c(\mu ϵ)^{1/4}`$ yielding
$$\lambda _1(ϵ)c^{}(\mu ϵ)^{1/4}$$
(16)
where $`c`$ and $`c^{}`$ are constants. Also this scaling law is in perfect agreement with numerical results. The comparison between the full analytic prediction of $`\lambda _1(ϵ)`$, obtained by substituting Eqs.(8) and (10) into Eqs.(6), shows a perfect agreement with the outcomes of a standard numerical computation of the largest Lyapunov exponent at different values of $`ϵ`$. Let us remark that the analytic prediction of $`\lambda _1(ϵ)`$ is obtained in the $`N\mathrm{}`$ limit, which is implicit in the computation of the microcanonical averages of the Ricci curvature and of its fluctuations. Now, as the analytic expression in (6) is in excellent agreement with the numerical data obtained for discrete (particle) systems at different values of $`N`$, there is no reason to doubt that such an agreement will hold true also when $`N`$ is varied while keeping the total length $`L`$ of the system fixed, i.e. in the continuum limit. The non-trivial difference between this limit and thermodynamic limit in Ref. is that $`N\mathrm{}`$ while the energy remains finite in the former case, whereas both $`N`$ and energy diverge in the latter case. Thus let us now consider the Hamiltonian (2) discretized on a lattice of length $`L`$ and spacing $`a`$ (with $`L=Na`$)
$`H(\{\pi _i\},\{\varphi _i\})`$ $`=`$ (17)
$`{\displaystyle \underset{i=1}{\overset{N}{}}}a[{\displaystyle \frac{\pi _i^2}{2\nu }}`$ $`+`$ $`{\displaystyle \frac{\eta }{2}}\left({\displaystyle \frac{\varphi _{i+1}\varphi _i}{a}}\right)^2+{\displaystyle \frac{\mu _0}{4}}\left({\displaystyle \frac{\varphi _{i+1}\varphi _i}{a}}\right)^4],`$ (18)
where the dimensional constants $`\nu ,\eta `$ have been introduced. We have thus obtained a discrete (lattice) system, similar to that described by the Hamiltonian (1); in order to make a direct comparison between the two Hamiltonians of Eqs.(18) and (1), we must now absorbe the dimensional parameter $`a`$ into the constants of the Hamiltonian. By means of the following rescaling of the coordinates $`\varphi _i`$ and of the time $`t`$: $`\xi _i=\varphi _i/\alpha `$, with $`\alpha =\sqrt{a/\eta }`$, $`t^{}=t/\gamma `$ and $`\gamma =\sqrt{\eta /\nu }`$, and denoting by $`\{\zeta _i\}`$ the momenta conjugated to $`\{\varphi _i\}`$, $`H`$ of Eq.(18) is put in the form
$$H_a(\{\zeta _i\},\{\xi _i\})=\underset{i=1}{\overset{N}{}}\left[\frac{\zeta _i^2}{2}+\frac{1}{2}(\xi _{i+1}\xi _i)^2+\frac{\mu _0}{a\eta ^2}\frac{1}{4}(\xi _{i+1}\xi _i)^4\right],$$
(19)
which is just the form of $`H`$ in Eq.(1), so that we can apply the geometric treatment of dynamical chaos also to the systems described by this family of Hamiltonians $`H_a`$. The precise meaning of this comparison is the following. The continuum Hamiltonian (2) has $`H`$ of Eq. (18) as its discretized version, where the parameters $`\eta ,\nu ,\mu _0`$ are equal to the values of the continuum model. After the recasting of Eq.(18) into the form (19) we can easily derive the value of the coupling constant $`\mu `$ of the discrete FPU model (1) so that a correspondence can be made between the discrete model and the lattice version (19) of the continuum model for any value of the lattice spacing $`a`$. Evidently the relation between the coupling constants is $`\mu =\mu _0/a\eta ^2`$, and this tells that in order to represent the solutions of a set of systems (18) by means of those of a set of discrete systems (1) with a finer and finer sampling of the spatial support of length $`L`$, i.e. letting $`a0`$, the coupling constant $`\mu `$ must be larger and larger. To the lattice-discretized system (19) we can now apply all the above equations for the geometric description of chaos, provided that we replace $`\mu `$ with $`\mu _0/a\eta ^2`$. In passing to the continuum limit we have to let $`a0`$ while the total energy of the system $`E_{tot}`$ is kept fixed. Hence the energy density $`ϵ`$ has to diminish as $`E_{tot}/N`$, but $`N=L/a`$ and then $`ϵ=E_{tot}a/L`$. Hence we obtain
$$\mu ϵ=\frac{\mu _0E_{tot}a}{a\eta ^2L}=cost.$$
(20)
From Eq.(20) it then follows that the equations for $`\kappa _0`$ and $`\sigma _\mathrm{\Omega }^2`$ (8)-(10) are stable with respect to the limit to the continuum and so is the largest Lyapunov exponent. Therefore the field equation (3) is chaotic and its Lyapunov exponent has the pattern shown by Fig.1. This constitutes a first example of analytic calculation of $`\lambda _1(ϵ)`$ for a field equation. Implicit in our computation is the possibility of approximating, as accurately as needed, a partial differential equation by means of a finite (truncated) system of ordinary differential equations, for which Lyapunov exponents are well defined. In conclusion we have shown that the continuum limit of the Fermi-Pasta-Ulam $`\beta `$ model is chaotic. We have discussed the interesting problematics raised by this result on the statistical description of classical field theory models that might be relevant also to the Wick-rotated quantum field theories.
It is a pleasure to thank M. Casartelli for useful comments. |
warning/0001/nucl-th0001056.html | ar5iv | text | # Adiabatic Selfconsistent Collective Coordinate Method for Large Amplitude Collective Motion in Superconducting Nuclei
## I Introduction
Large amplitude collective motions (LACM), such as fissions, shape transitions, anharmonic vibrations and low energy heavy ion reactions, are often encountered in the studies of nuclear structures and reactions. To go beyond the phenomenological models assuming some macroscopic or collective degrees of freedom motivated by the experimental facts and intuitions, many attempts have been made to construct theories that are able to describe the LACM on a microscopic basis of the nuclear many-body Hamiltonian. Especially, theories based on the time-dependent Hartree-Fock (TDHF) approximation have been investigated extensively . The TDHF is a general framework for describing low energy nuclear dynamics accompanying evolution of the nuclear selfconsistent mean-field. A LACM corresponds to a specific solution of the TDHF equation of motion. Since such a solution forms only a subset of the whole TDHF states (Slater determinants), it is often called a collective path, a collective subspace, or a collective submanifold. The collective coordinates are then a set of small number of variables that parameterize the collective subspace, and the collective Hamiltonian is a function governing the time evolution of the collective coordinates. One of the main purposes of the LACM theories is to provide a scheme to determine the collective subspace and the collective Hamiltonian on the basis of microscopic many-body Hamiltonian. Although the studies of LACM theories are the vast field of research with many recent developments to different directions, realistic applications to nuclear structure problems are rather limited. In this paper, we would like to propose a new practical method to calculate the collective subspace.
The adiabatic approximation has been often utilized for formulating the theory of collective subspace. Indeed, some class of LACM, such as nuclear fissions, can be regarded as a slow motion, thus justifying the adiabatic approximation. The adiabatic TDHF (ATDHF) theory is one of the most well known adiabatic theories and has been applied in some cases to realistic description of heavy ion reactions . The ATDHF theory, however, accompanied a problem of non-uniqueness of the solution . Later, efforts to settle the non-uniqueness problem were made from different view points. The works of Ref. emphasize the importance of the canonical variable condition and the analyticity as a function of collective coordinate for finding a unique solution. The proposed procedure relying on the Taylor expansion method has not been applied to realistic calculations. Another work pointed out that the collective subspace can be uniquely determined by using the next order equation of the ATDHF theory. It is clarified also that the adiabatic collective path of LACM becomes the valley line of the potential function in the multi-dimensional space associated with the TDHF states . Further, the adiabatic collective path can be defined by equations for a local harmonic mode at each point of the collective path. These developments are summarized in a consistent way in the formalism of Ref. . Note however that the adiabatic theory of Ref. relys on a multi-dimensional classical phase space representation of the TDHF determinantal states . A realistic application of this theory has not been done yet except for the one to a light nucleus . Furthermore a problem of particle number conservation arises when applied to superconducting nuclei .
Theories without the adiabatic approximation have been also developed within the TDHF framework. The early works of this direction are called local harmonic approximations . Later, a set of general equations that can determine the collective subspace and the collective Hamiltonian were found and formulated in a consistent form known as the selfconsistent collective coordinate method (SCC or SCCM) . The theory is purely based on the TDHF with no further approximation. The method also provides a concrete and practical scheme to solve the basic equations using a power series expansion with respect to the boson-like variables defined as a linear combination of the collective coordinates and momenta. The pairing correlation in superconducting nuclei is easily incorporated within the SCCM by adopting the time-dependent Hartree-Fock-Bogoliubov (TDHFB) equation in place of the TDHF, and the conservation law of the particle number is consistently introduced in the basic framework of the SCCM . Thanks to these features the SCCM has been applied for many realistic descriptions of anharmonic vibrations in medium and heavy nuclei . However, the expansion method may not be suitable for the large amplitude motions of adiabatic nature, for which change of the nuclear mean-field is so large that the power series expansion of the collective coordinates may not be justified.
In the present paper, we try to combine merits of two approaches mentioned above, i.e. the SCCM and the adiabatic thoery in order to formulate a theory that provides a consistent and practical method easily applicable to realistic descriptions of the adiabatic LACM in superconducting nuclei. We achieve this aim by introducing an adiabatic approximation to the general framework of SCCM. Here we treat superconducting nuclei since the pairing correlations play essential roles in many cases, like spontaneous fission, tunneling between superdeformed and normal deformed configurations, and coupling between coexisting states with different nuclear shape (shape coexistence phenomena). Although the use of the superconducting mean field requires us to respect the particle number conservation, the SCCM allows a simple and consistent treatment of the conservation law. We also avoid the non-uniqueness problem by utilizing the principles similar to that of Refs. . Furthermore, we shall show that the equations of the adiabatic SCCM thus formulated can be transformed to another set of equations that have a similar structure as the local harmonic approach to the adiabatic theories . Therefore, the present formalism is not only an extension of the SCCM, but also succeeds some aspects of the recent adiabatic theories such as Ref..
In addition to the general formulation (Sect.II), we present a practical scheme to solve the basic equations given in the local harmonic form for general classes of the many-body nuclear Hamiltonian (Sect.III and Appendix). These equations are given in terms of the matrix elements of the many-body Hamiltonian expressed by the quasiparticle operators, thus enabling ones to develop a straightforward coding of a numerical program to solve the equations. In this way, we provide a complete procedure to extract the collective subspace and the collective Hamiltonian. We also discuss a possible prescription to extend the formalism to the cases of the multi-dimensional collective motions (Sect.IV). Conclusions are outlined in Sect.V.
## II Basic Equations
### A The SCC method for superconducting nuclei
In this subsection, we recapitulate the basic equations of the SCC method in a way suitable for treating the superconducting nuclei.
We introduce the TDHFB approximation to describe LACM in superconducting many-fermion systems. Here the time-dependent many-body state vector $`|\varphi (t)`$ is constrained to a generalized Slater determinant, which is chosen as a variational wave function. Time evolution of $`|\varphi (t)`$ is then determined by the time-dependent variational principle
$$\delta \varphi (t)\left|i\frac{}{t}\widehat{H}\right|\varphi (t)=0,$$
(1)
where the variation is given by $`\delta |\varphi (t)=a_\alpha ^{}a_\beta ^{}|\varphi (t)`$ in terms of the quasiparticle operators $`\{a_\alpha ^{},a_\alpha \}`$ which satisfy the vacuum condition $`a_\alpha |\varphi (t)=0`$.
We assume that the LACM can be described in terms of the collective variables, i.e. the collective coordinate and momentum $`\{q,p\}`$ that are variables parameterizing the TDHFB state vector.<sup>*</sup><sup>*</sup>* We focus our discussion on a case of single collective coordinate. A multi-dimensional case is discussed in Sect.IV. The whole space of the TDHFB state vectors can be parameterized by $`M\times (M1)`$ variables ($`M`$ being the number of the single particle states) as shown by the generalized Thouless theorem . A set of the TDHFB state vector $`|\varphi (q,p)`$ forms the collective subspace in which the LACM can be properly described. One of the main problems we concern is how to determine the collective subspace on the basis of the TDHFB equations of motion. At the same time, we need to determine the collective Hamiltonian $`(q,p)`$ that governs the equation of motion for the collective variables $`\{q,p\}`$. This is a general purpose of theories of LACM.
When we apply the LACM theories to nuclei in the superconducting phase, a special attention has to be paid to the particle number conservation. Since the TDHFB state vector is not an eigenstate of the particle number operator $`\widehat{N}`$, one would like to formulate the LACM theory so that the particle number expectation value is conserved during the course of collective motion. This is a problem which is specific to the TDHFB, but not to TDHF for which the state vector is a number eigenstate.
It is well known that the expectation value of a conserved observable is kept constant during the time-evolution of $`|\varphi (t)`$ governed by the TDHF(B) equations of motion. In the case of the pairing problem, the TDHFB state vector violates spontaneously the symmetry with respect to the gauge rotation $`e^{i\phi \widehat{N}}`$, but a rotational motion related to the gauge rotation (often called the pairing rotation) emerges automatically to restore the gauge symmetry. Therefore, the LACM of superconducting nuclei, described by the TDHFB theory, necessarily accompany the pairing rotation, for which we introduce the collective coordinate, $`\phi `$, the gauge angle, and the conjugate collective momenta, $`N`$, which represents the particle number . Thus, we are obliged to consider a collective subspace that is parameterized by the set of four collective variables $`\{q,p,\phi ,N\}`$. For simplicity, here we assume a single kind of particles. Extension to systems with many kinds (e.g., protons and neutrons in nuclei) is straightforward.
Let us now present the basic equations of the SCCM that determine the collective subspace $`|\varphi (q,p,\phi ,N)`$ and the collective Hamiltonian $`(q,p,\phi ,N)`$. As discussed above, the variable $`\phi `$ is introduced to represent the gauge angle. This requirement is easily satisfied if one uses the following parameterization
$$|\varphi (q,p,\phi ,N)=e^{i\phi \widehat{N}}|\varphi (q,p,N),$$
(2)
where $`\widehat{N}`$ is the number operator of particles. Here $`|\varphi (q,p,N)`$ represents an intrinsic state that rotates in the gauge space.
Basic equations of the SCCM consists of a canonical variable condition and invariance principle of the time-dependent Schrödinger equation (TDHFB equation in our case). The canonical variable condition is, in general, given by
$`\varphi (q,p,\phi ,N)\left|i{\displaystyle \frac{}{q}}\right|\varphi (q,p,\phi ,N)=p+{\displaystyle \frac{S}{q}},`$ (4)
$`\varphi (q,p,\phi ,N)\left|{\displaystyle \frac{}{ip}}\right|\varphi (q,p,\phi ,N)={\displaystyle \frac{S}{p}},`$ (5)
$`\varphi (q,p,\phi ,N)\left|i{\displaystyle \frac{}{\phi }}\right|\varphi (q,p,\phi ,N)=N+{\displaystyle \frac{S}{\phi }},`$ (6)
$`\varphi (q,p,\phi ,N)\left|{\displaystyle \frac{}{iN}}\right|\varphi (q,p,\phi ,N)={\displaystyle \frac{S}{N}},`$ (7)
for the collective subspace parameterized by two sets of coordinates $`(q,\phi )`$ and momenta $`(p,N)`$. Although $`S`$ is an arbitrary function of $`\{q,p,\phi ,N\}`$, we choose $`S=0`$ which is appropriate for the adiabatic approximation. Then the canonical variable condition can be rewritten as equations for the state $`|\varphi (q,p,N)`$,
$`\varphi (q,p,N)\left|i{\displaystyle \frac{}{q}}\right|\varphi (q,p,N)`$ $`=p,`$ (9)
$`\varphi (q,p,N)\left|{\displaystyle \frac{}{ip}}\right|\varphi (q,p,N)`$ $`=0,`$ (10)
$`\varphi (q,p,N)\left|\widehat{N}\right|\varphi (q,p,N)`$ $`=N,`$ (11)
$`\varphi (q,p,N)\left|{\displaystyle \frac{}{iN}}\right|\varphi (q,p,N)`$ $`=0.`$ (12)
The third equation requires that the collective variable $`N`$ is identical to the expectation value of the number operator. In other words, the particle number expectation value does not depend on the collective variables $`(q,p)`$ for the LACM under consideration. This is nothing but the condition of particle number conservation.
The collective Hamiltonian is defined as value of the total energy on the collective subspace, given by
$``$ $`=\varphi (q,p,\phi ,N)\left|\widehat{H}\right|\varphi (q,p,\phi ,N)`$ (14)
$`=\varphi (q,p,N)\left|\widehat{H}\right|\varphi (q,p,N).`$ (15)
Since the Hamiltonian $`\widehat{H}`$ commutes with the number operator $`\widehat{N}`$, the collective Hamiltonian does not depend on the gauge angle $`\phi `$. Therefore, $`\phi `$ becomes cyclic as we expect.
The invariance principle of the TDHFB equation plays a central role to determine the collective subspace, which requires that the TDHFB state vector $`|\varphi (q(t),p(t),\phi (t),N(t))`$ evolving in time within the collective subspace should obey the full TDHFB equation, Eq.(1). This is equivalent to a condition that the collective subspace is an invariant subspace of the TDHFB equations of motion. Inserting Eq.(2) into the time-dependent variational principle, Eq.(1), one obtains
$$\delta \varphi (q,p,N)\left|\widehat{H}\frac{dq}{dt}\stackrel{̊}{P}+\frac{dp}{dt}\stackrel{̊}{Q}+\frac{dN}{dt}\stackrel{̊}{\mathrm{\Theta }}\frac{d\phi }{dt}\widehat{N}\right|\varphi (q,p,N)=0,$$
(16)
where the infinitesimal generators defined by
$`\stackrel{̊}{P}|\varphi (q,p,N)`$ $`=i{\displaystyle \frac{}{q}}|\varphi (q,p,N),`$ (18)
$`\stackrel{̊}{Q}|\varphi (q,p,N)`$ $`={\displaystyle \frac{1}{i}}{\displaystyle \frac{}{p}}|\varphi (q,p,N),`$ (19)
$`\stackrel{̊}{\mathrm{\Theta }}|\varphi (q,p,N)`$ $`={\displaystyle \frac{1}{i}}{\displaystyle \frac{}{N}}|\varphi (q,p,N)`$ (20)
are used. These operators are one-body operators which can be written as linear combinations of bilinear products $`\{a_\alpha ^{}a_\beta ^{},a_\beta a_\alpha ,a_\alpha ^{}a_\beta \}`$ of the quasiparticle operators defined with respect to $`|\varphi (q,p,N)`$. Because of the canonical variable conditions, these infinitesimal generators satisfy the following commutation relations
$`\varphi (q,p,N)\left|[\stackrel{̊}{Q},\stackrel{̊}{P}]\right|\varphi (q,p,N)=i,`$ (22)
$`\varphi (q,p,N)\left|[\stackrel{̊}{\mathrm{\Theta }},\widehat{N}]\right|\varphi (q,p,N)=i,`$ (23)
and commutators of other combinations of $`\stackrel{̊}{Q},\stackrel{̊}{P},\stackrel{̊}{\mathrm{\Theta }},\widehat{N}`$ give zero expectation value. By taking the variation as $`\delta |\varphi (q,p,N)=\{\stackrel{̊}{P},\stackrel{̊}{Q},\stackrel{̊}{\mathrm{\Theta }},\widehat{N}\}|\varphi (q,p,N)`$, Eq.(16) produces the canonical equations of motion for the collective variables
$`{\displaystyle \frac{dq}{dt}}`$ $`=`$ $`{\displaystyle \frac{}{p}}=i\varphi (q,p,N)\left|[\widehat{H},\stackrel{̊}{Q}]\right|\varphi (q,p,N),`$ (25)
$`{\displaystyle \frac{dp}{dt}}`$ $`=`$ $`{\displaystyle \frac{}{q}}=i\varphi (q,p,N)\left|[\widehat{H},\stackrel{̊}{P}]\right|\varphi (q,p,N),`$ (26)
$`{\displaystyle \frac{d\phi }{dt}}`$ $`=`$ $`{\displaystyle \frac{}{N}}=i\varphi (q,p,N)\left|[\widehat{H},\stackrel{̊}{\mathrm{\Theta }}]\right|\varphi (q,p,N),`$ (27)
$`{\displaystyle \frac{dN}{dt}}`$ $`=`$ $`{\displaystyle \frac{}{\phi }}=0.`$ (28)
Using Eq.(II A), Eq.(16) then reduces to an equation of collective subspace
$$\delta \varphi (q,p,N)\left|\widehat{H}\frac{}{p}\stackrel{̊}{P}\frac{}{q}\stackrel{̊}{Q}\frac{}{N}\widehat{N}\right|\varphi (q,p,N)=0.$$
(29)
If we take a variation $`\delta _{}`$ that is orthogonal to the infinitesimal generators $`\{\stackrel{̊}{P},\stackrel{̊}{Q},\stackrel{̊}{\mathrm{\Theta }},\widehat{N}\}`$, one can immediately show $`\delta _{}\varphi (q,p,N)\left|\widehat{H}\right|\varphi (q,p,N)=0`$, which implies that the the energy expectation value is stationary on the collective subspace for all the variations except for the tangent directions along the collective subspace. In other words, the collective mode is decoupled from the other modes of excitation.
We remark here that the above basic equations of the SCCM are invariant under point transformations of the collective coordinate
$`qq^{}=q^{}(q),`$ (31)
$`pp^{}=p\times \left(q^{}/q\right)^1.`$ (32)
The basic principles, i.e. the canonical variable condition, Eq.(II A), and the invariance principle of the TDHFB equation, Eq.(16), are not affected by the general canonical transformations of collective variables $`\{q,p,\phi ,N\}\{q^{},p^{},\phi ^{},N^{}\}`$. By taking the parameterization, Eq.(2), and the specific choice of $`S=0`$ in Eq.(II A), the allowed canonical transformations are restricted to the point transformations .
### B Adiabatic approximation
Assuming that the LACM described by the collective variables $`\{q,p\}`$ is slow motion, we here introduce the adiabatic approximation to the SCCM. Namely we shall expand the basic equations with respect to the collective momentum $`p`$, which is appropriate for small value of momentum. Since the particle number variable $`N`$ is a momentum variable in the present formulation, we also expand the basic equations with respect to $`n=NN_0`$, when we consider a system with particle number $`N_0`$.
Let us first consider the expansion of the TDHFB state vector $`|\varphi (q,p,N)`$ in the collective subspace. The origin of the expansion is the state $`|\varphi (q)|\varphi (q,p,N)|_{p=0,N=N_0}`$. We can assume that this is a time-even state, i.e., $`𝒯|\varphi (q)=|\varphi (q)`$ under the time-reversal operation $`𝒯`$ (Here we consider the system of even numbers of particles). Thanks to the generalized Thouless theorem, the state vector $`|\varphi (q,p,N)`$ is expressed as
$$|\varphi (q,p,N)=e^{i\widehat{G}(q,p,n)}|\varphi (q)$$
(33)
by using the unitary transformation $`e^{i\widehat{G}(q,p,n)}`$. Here the hermitian operator $`\widehat{G}`$ is given by
$$\widehat{G}(q,p,n)=\underset{\alpha >\beta }{}\left(G_{\alpha \beta }(q,p,n)a_\alpha ^{}a_\beta ^{}+G_{\alpha \beta }^{}(q,p,n)a_\beta a_\alpha \right)=\widehat{G}(q,p,n)^{}.$$
(34)
Here and hereafter, the quasiparticle operators $`\{a_\alpha ^{},a_\alpha \}`$ are always defined locally at each value of $`q`$ and satisfy the condition $`a_\alpha |\varphi (q)=0`$. We now expand the operator $`\widehat{G}(q,p,n)`$ in powers of $`p`$ and $`n`$ and keep only the lowest order term. Namely,
$`\widehat{G}(q,p,n)`$ $`=`$ $`p\widehat{Q}(q)+n\widehat{\mathrm{\Theta }}(q),`$ (36)
$`\widehat{Q}(q)`$ $`=`$ $`{\displaystyle \underset{\alpha >\beta }{}}\left(Q_{\alpha \beta }(q)a_\alpha ^{}a_\beta ^{}+Q_{\alpha \beta }^{}(q)a_\beta a_\alpha \right)=\widehat{Q}(q)^{},`$ (37)
$`\widehat{\mathrm{\Theta }}(q)`$ $`=`$ $`{\displaystyle \underset{\alpha >\beta }{}}\left(\mathrm{\Theta }_{\alpha \beta }(q)a_\alpha ^{}a_\beta ^{}+\mathrm{\Theta }_{\alpha \beta }^{}(q)a_\beta a_\alpha \right)=\widehat{\mathrm{\Theta }}(q)^{}.`$ (38)
If we require that time-reversal of $`|\varphi (q,p,N)`$ causes sign inversion of the collective momentum $`p`$, i.e. $`𝒯|\varphi (q,p,N)=|\varphi (q,p,N)`$, the operators $`\widehat{Q}(q)`$ and $`\widehat{\mathrm{\Theta }}(q)`$ are time-even ($`𝒯\widehat{Q}(q)𝒯^1=\widehat{Q}(q)`$) and time-odd ($`𝒯\widehat{\mathrm{\Theta }}(q)𝒯^1=\widehat{\mathrm{\Theta }}(q)`$), respectively. If we put $`n=0`$ (i.e. $`N=N_0`$), the parameterization Eq.(33) together with Eq.(II B) reduces to $`|\varphi (q,p)=e^{ip\widehat{Q}(q)}|\varphi (q)`$ , which is the same form as the one introduced by Villars and often used in the ATDHF theories .
The collective Hamiltonian is expanded as
$`(q,p,N)`$ $`=`$ $`V(q)+{\displaystyle \frac{1}{2}}B(q)p^2+\lambda (q)n,`$ (40)
$`V(q)`$ $`=`$ $`(q,p,N)|_{p=0,N=N_0}=\varphi (q)\left|\widehat{H}\right|\varphi (q),`$ (41)
$`B(q)`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{^2(q,p,N)}{p^2}}|_{p=0,N=N_0}=\varphi (q)\left|[[\widehat{H},\widehat{Q}(q)],\widehat{Q}(q)]\right|\varphi (q),`$ (42)
$`\lambda (q)`$ $`=`$ $`{\displaystyle \frac{(q,p,N)}{N}}|_{p=0,N=N_0}=\varphi (q)\left|[\widehat{H},i\widehat{\mathrm{\Theta }}(q)]\right|\varphi (q),`$ (43)
where we kept the collective momentum $`p`$ up to the second order, while up to the first order in $`n`$. The collective Hamiltonian for the system with $`N=N_0`$ particles ( $`n=0`$ ) is given by
$$(q,p,N_0)=V(q)+\frac{1}{2}B(q)p^2$$
(44)
as the sum of the collective potential $`V(q)`$ and the collective kinetic energy (the second term).
We next expand the infinitesimal generators. It is convenient for this purpose to define the unitary transformation $`\stackrel{̊}{P}^{}=e^{i\widehat{G}}\stackrel{̊}{P}e^{i\widehat{G}},\stackrel{̊}{Q}^{}=e^{i\widehat{G}}\stackrel{̊}{Q}e^{i\widehat{G}},\stackrel{̊}{\mathrm{\Theta }}^{}=e^{i\widehat{G}}\stackrel{̊}{\mathrm{\Theta }}e^{i\widehat{G}}`$ of the infinitesimal generators $`\stackrel{̊}{P},\stackrel{̊}{Q},\stackrel{̊}{\mathrm{\Theta }}`$. They are expanded as
$`\stackrel{̊}{P}^{}`$ $`=`$ $`\widehat{P}(q)+e^{i\widehat{G}}i{\displaystyle \frac{}{q}}e^{i\widehat{G}}=\widehat{P}(q)p{\displaystyle \frac{\widehat{Q}}{q}}n{\displaystyle \frac{\widehat{\mathrm{\Theta }}}{q}}+\mathrm{},`$ (45)
$`\stackrel{̊}{Q}^{}`$ $`=`$ $`e^{i\widehat{G}}{\displaystyle \frac{}{ip}}e^{i\widehat{G}}=\widehat{Q}(q)+{\displaystyle \frac{i}{2}}[\widehat{Q},p\widehat{Q}+n\widehat{\mathrm{\Theta }}]+\mathrm{},`$ (46)
$`\stackrel{̊}{\mathrm{\Theta }}^{}`$ $`=`$ $`e^{i\widehat{G}}{\displaystyle \frac{}{iN}}e^{i\widehat{G}}=\widehat{\mathrm{\Theta }}(q)+{\displaystyle \frac{i}{2}}[\widehat{\mathrm{\Theta }},p\widehat{Q}+n\widehat{\mathrm{\Theta }}]+\mathrm{},`$ (47)
with use of the general expansion formula
$$e^{i\widehat{G}}e^{i\widehat{G}}=i\widehat{G}+\frac{1}{2!}[i\widehat{G},i\widehat{G}]+\frac{1}{3!}[[i\widehat{G},i\widehat{G}],i\widehat{G}]+\mathrm{}$$
(48)
The operator $`\widehat{P}(q)`$ is the infinitesimal generator with respect to $`|\varphi (q)`$ defined by
$$\widehat{P}(q)|\varphi (q)=i\frac{}{q}|\varphi (q).$$
(49)
Similarly, we introduce the unitary transformation of the number operator and expand it as
$$\stackrel{̊}{N}^{}e^{i\widehat{G}}\widehat{N}e^{i\widehat{G}}=\widehat{N}+i[\widehat{N},p\widehat{Q}+n\widehat{\mathrm{\Theta }}]+\mathrm{}$$
(50)
Substituting these operators in the canonical variable condition, Eq.(II A), we have
$`\varphi (q)\left|\stackrel{̊}{P}^{}(q,p,N)\right|\varphi (q)`$ $`=`$ $`p,`$ (52)
$`\varphi (q)\left|\stackrel{̊}{Q}^{}(q,p,N)\right|\varphi (q)`$ $`=`$ $`0,`$ (53)
$`\varphi (q)\left|\stackrel{̊}{\mathrm{\Theta }}^{}(q,p,N)\right|\varphi (q)`$ $`=`$ $`0,`$ (54)
$`\varphi (q)\left|\stackrel{̊}{N}^{}(q,p,N)\right|\varphi (q)`$ $`=`$ $`N.`$ (55)
Now we expand these equations with respect to momentum $`p`$ and $`n`$.
The zeroth order canonical variable conditions:
$`\varphi (q)\left|\widehat{P}(q)\right|\varphi (q)=\varphi (q)\left|i{\displaystyle \frac{}{q}}\right|\varphi (q)=0,`$ (56)
$`\varphi (q)\left|\widehat{Q}(q)\right|\varphi (q)=0,`$ (57)
$`\varphi (q)\left|\widehat{\mathrm{\Theta }}(q)\right|\varphi (q)=0,`$ (58)
$`\varphi (q)\left|\widehat{N}\right|\varphi (q)=N_0.`$ (59)
Eqs.(57) and (58) are automatically fulfilled by the definition, Eq.(II B), of the operators $`\widehat{Q}(q),\widehat{\mathrm{\Theta }}(q)`$. Eq.(56) can be satisfied if the $`q`$-dependent phase of $`|\varphi (q)`$ is properly chosen. Eq.(59) is nothing but the constraint on $`|\varphi (q)`$ for the conservation of average particle number.
The first order canonical variable conditions:
$`\varphi (q)\left|{\displaystyle \frac{\widehat{Q}(q)}{q}}\right|\varphi (q)=1,`$ (60)
$`\varphi (q)\left|[\widehat{Q}(q),\widehat{\mathrm{\Theta }}(q)]\right|\varphi (q)=0,`$ (61)
$`\varphi (q)\left|[\widehat{Q}(q),\widehat{N}]\right|\varphi (q)=0.`$ (62)
One finds
$$\varphi (q)\left|[\widehat{Q}(q),\widehat{P}(q)]\right|\varphi (q)=i,$$
(63)
which can be derived by differentiating Eq.(57) with respect to $`q`$ and using Eq.(60). One can also derive from Eq.(59)
$$\varphi (q)\left|[\widehat{P}(q),\widehat{N}]\right|\varphi (q)=0.$$
(64)
These equations give constraints on the infinitesimal generators $`\widehat{Q}(q),\widehat{P}(q)`$ concerning the normalization, Eq.(63), and the orthogonal condition to the particle number operator, Eq.(64).
Next we expand the equation of collective subspace, Eq.(29), to obtain a complete set of the basic equations for the adiabatic approximation. After rewriting Eq.(29) as
$$\delta \varphi (q)\left|e^{i\widehat{G}}\widehat{H}e^{i\widehat{G}}\frac{}{p}\stackrel{̊}{P}^{}\frac{}{q}\stackrel{̊}{Q}^{}\frac{}{N}\stackrel{̊}{N}^{}\right|\varphi (q)=0,$$
(65)
one can expand each term with respect to $`p`$ and $`n`$ with use of the equations listed above.
The zeroth order equation of collective subspace:
$$\delta \varphi (q)\left|\widehat{H}\lambda (q)\widehat{N}\frac{V}{q}\widehat{Q}(q)\right|\varphi (q)=0.$$
(66)
The first order equation of collective subspace:
$$\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},\widehat{Q}(q)]\frac{1}{i}B(q)\widehat{P}(q)\right|\varphi (q)=0.$$
(67)
These equations are similar to the equations of path in the Villars’ ATDHF theory except that the present paper deals with the superconducting Hartree-Fock-Bogoliubov (HFB) state, and that the Hamiltonian accompanies the $`q`$-dependent chemical potential term $`\lambda (q)\widehat{N}`$. As we mentioned in Sect.I, the ATDHF theory has the problem that the solution satisfying these two equations is not uniquely determined . Although an additional validity condition was introduced to further constrain the solutions , the procedure of Ref. does not fully solve the problem since the method does not work around the HF minima.
The non-uniqueness problem has been investigated in recent developments of the adiabatic theories, and in our opinion they are classified into two different approaches. The first one represented by Ref. claims that the solution is uniquely determined if an RPA boundary condition is specified at the HF minimum and if the analyticity of the collective path as a function of $`q`$ is imposed together with the canonical variable condition. The solution, however, needs to be constructed in an analytic way or by means of a Taylor expansion method with respect to the collective coordinate $`q`$. We do not adopt this approach since we wish to construct a method applicable to systems under large excursion from the HFB minimum. We rather follow the other approach represented by Refs. . These theories require an additional condition that the equation of collective subspace (corresponding to the decoupling condition in Ref.) should be satisfied up to the next order of the adiabatic expansion. In the present formulation, this second order condition is expressed as follows.
The second order equation of collective subspace:
$$\delta \varphi (q)\left|\frac{1}{2}[[\widehat{H}\lambda (q)\widehat{N},\widehat{Q}(q)],\widehat{Q}(q)]B(q)\mathrm{\Delta }\widehat{Q}(q)\right|\varphi (q)=0,$$
(68)
where
$`\mathrm{\Delta }\widehat{Q}(q)={\displaystyle \frac{\widehat{Q}}{q}}+\mathrm{\Gamma }(q)\widehat{Q}(q),`$ (69)
$`\mathrm{\Gamma }(q)={\displaystyle \frac{1}{2B(q)}}{\displaystyle \frac{B}{q}}.`$ (70)
This equation is equivalent in its mathematical form to the one given by Ref. if the chemical potential term $`\lambda (q)\widehat{N}`$ is neglected. The last term $`B(q)\mathrm{\Delta }\widehat{Q}(q)`$, often called a curvature term, was simply neglected in the original version of local harmonic approximation . In the next subsection, instead of neglecting this curvature term, we shall rewrite $`\mathrm{\Delta }\widehat{Q}(q)`$ and change Eq.(68) into a workable form.
It is worth noting here the invariance of the adiabatic equations against the coordinate transformation. The collective momentum $`p`$ undergoes the linear homogeneous transformation under the point transformation, Eq.(II A). Therefore, different orders of the expansion with respect to the power of $`p`$ are not mixed up under the point transformation. The invariance property of the basic equations of SCCM is thus inherited to each equation of the adiabatic approximation listed above. One can also confirm this property by seeing that the quantities appearing in the equations transform as
$`\widehat{Q}(q)\widehat{Q}^{}(q^{})=\widehat{Q}(q(q^{}))\left({\displaystyle \frac{q^{}}{q}}\right),`$ (72)
$`\widehat{P}(q)\widehat{P}^{}(q^{})=\widehat{P}(q(q^{}))\left({\displaystyle \frac{q^{}}{q}}\right)^1,`$ (73)
$`{\displaystyle \frac{V}{q}}{\displaystyle \frac{V^{}}{q^{}}}={\displaystyle \frac{V}{q}}\left({\displaystyle \frac{q^{}}{q}}\right)^1,`$ (74)
$`BB^{}(q^{})=B(q(q^{}))\left({\displaystyle \frac{q^{}}{q}}\right)^2,`$ (75)
$`\mathrm{\Delta }\widehat{Q}(q)\mathrm{\Delta }\widehat{Q}^{}(q^{})=\mathrm{\Delta }\widehat{Q}(q).`$ (76)
## III Local Harmonic Approximation to Collective Subspace
### A Local Harmonic Equations
In this section we give an approximate but concrete procedure to construct a solution of the adiabatic SCC method. To this end, we first derive, from the adiabatic equations, another set of equations of collective subspace which can be solved in a way similar to the RPA equation.
We first take a derivative of the zeroth order equation, Eq.(66), with respect to $`q`$, which leads to
$`\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},{\displaystyle \frac{1}{i}}\widehat{P}(q)]C(q)\widehat{Q}(q){\displaystyle \frac{V}{q}}\mathrm{\Delta }\widehat{Q}(q){\displaystyle \frac{\lambda }{q}}\widehat{N}\right|\varphi (q)=0,`$ (77)
$`C(q)={\displaystyle \frac{^2V}{q^2}}\mathrm{\Gamma }(q){\displaystyle \frac{V}{q}},`$ (78)
where $`\mathrm{\Delta }\widehat{Q}(q)`$ and $`\mathrm{\Gamma }(q)`$ are given by Eqs.(69) and (70), respectively. Using Eqs.(68), we eliminate $`\mathrm{\Delta }\widehat{Q}(q)`$ and rewrite Eq.(77) as
$$\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},\frac{1}{i}\widehat{P}(q)]C(q)\widehat{Q}(q)\frac{1}{2B(q)}[[\widehat{H}\lambda (q)\widehat{N},\frac{V}{q}\widehat{Q}(q)],\widehat{Q}(q)]\frac{\lambda }{q}\widehat{N}\right|\varphi (q)=0.$$
(79)
Furthermore, due to Eq.(66), we find
$$\frac{V}{q}\widehat{Q}=(\widehat{H}\lambda \widehat{N})_A,$$
(80)
where $`(\widehat{H}\lambda \widehat{N})_A`$ means $`a^{}a^{}`$ and $`aa`$ part of the operator $`\widehat{H}\lambda \widehat{N}`$ containing two-quasiparticle creation and annihilation in the normal-ordered expression.
We thus replace Eqs.(66)-(68) by the equivalent set,
$$\delta \varphi (q)\left|\widehat{H}_M(q)\right|\varphi (q)=0,$$
(81)
$$\delta \varphi (q)\left|[\widehat{H}_M(q),\widehat{Q}(q)]\frac{1}{i}B(q)\widehat{P}(q)\right|\varphi (q)=0,$$
(82)
$$\delta \varphi (q)\left|[\widehat{H}_M(q),\frac{1}{i}\widehat{P}(q)]C(q)\widehat{Q}(q)\frac{1}{2B(q)}[[\widehat{H}_M(q),(\widehat{H}\lambda (q)\widehat{N})_A],\widehat{Q}(q)]\frac{\lambda }{q}\widehat{N}\right|\varphi (q)=0.$$
(83)
In Eq.(82) and (83), $`\widehat{H}\lambda \widehat{N}`$ has been replaced by
$$\widehat{H}_M(q)=\widehat{H}\lambda (q)\widehat{N}\frac{V}{q}\widehat{Q}(q),$$
(84)
since the last term has no influence. The operator $`\widehat{H}_M(q)`$ may be regarded as the Hamiltonian in the moving frame. The second and third terms can be identified with generalized cranking terms associated with the pairing rotation and the LACM, respectively.
Equations (82) and (83) are linear equations with respect to the one-body operators $`\widehat{Q}(q)`$ and $`\widehat{P}(q)`$. They have essentially the same structure as the standard RPA equations except for the last two terms in Eq.(83). The quantity $`C(q)`$ is the local stiffness parameter defined as the second (covariant) derivative of the collective potential $`V(q)`$. The infinitesimal generators $`\widehat{Q}(q)`$ and $`\widehat{P}(q)`$ are thus closely related to the harmonic normal modes locally defined for $`|\varphi (q)`$ and the moving frame Hamiltonian $`\widehat{H}_M(q)`$. These equations may be called local harmonic equations.
It was shown in Ref. that the zeroth, first and second order equations of ATDHF give a valley line of a potential energy surface in a multi-dimensional configuration space associated with the TDHF states. Similarly, the local harmonic equations we have obtained, Eqs.(81)-(83), define the valley of the multi-dimensional potential energy surface. The solution of these equations will be uniquely determined if a suitable boundary condition is specified. These features are similar to the formulation of Ref. where the valley equation of the potential energy surface is derived from the decoupling condition.
We remark again that the local harmonic equations in the present paper differ from the ones of Rowe-Bassermann and Marumori with respect to the the third and the fourth terms of Eq.(83), which arise from the curvature term (derivative of the generator) and the particle number constraint, respectively. It is important to keep the curvature term in order to maintain the relation between the collective subspace and the valley of the potential surface. We also note that the present formalism is invariant with respect to the point transformation of the collective coordinate, as is the formulation of Ref..
### B Matrix Formulation of Local Harmonic Equations
Let us now give a procedure to find the operators $`\widehat{Q}(q),\widehat{P}(q)`$ that satisfy the local harmonic equations, (82) and (83) , for a given state $`|\varphi (q)`$. Since they are linear equations with respect to these operators, it can be solved in an analogous way to the standard RPA. To show this, we first express the operator $`\widehat{P}(q)`$ and $`\widehat{N}`$ in terms of the quasiparticle operators:
$$\widehat{P}(q)=i\underset{\alpha >\beta }{}\left(P_{\alpha \beta }(q)a_\alpha ^{}a_\beta ^{}P_{\alpha \beta }^{}(q)a_\beta a_\alpha \right)=\widehat{P}(q)^{},$$
(85)
$$\widehat{N}=\underset{\alpha >\beta }{}\left(N_{\alpha \beta }(q)a_\alpha ^{}a_\beta ^{}+N_{\alpha \beta }^{}(q)a_\beta a_\alpha \right).$$
(86)
Note that the $`a^{}a`$ and c-number parts are neglected here since they do not change the state vector $`|\varphi (q)`$ except for the phase. The Hamiltonian $`\widehat{H}`$ is also expressed in terms of the same quasiparticle operators. Assuming that the matrix elements $`Q_{\alpha \beta },P_{\alpha \beta }`$ are real, the local harmonic equations are written as the following matrix equations.
$`(𝑨𝑩)𝑸B(q)𝑷=0,`$ (88)
$`(𝑨+𝑩)𝑷C(q)𝑸{\displaystyle \frac{1}{B(q)}}𝑫𝑸\lambda ^{}𝑵=0,`$ (89)
$`𝑷^T𝑵=0,`$ (90)
$`2𝑸^T𝑷=1,`$ (91)
$`\lambda ^{}={\displaystyle \frac{\lambda }{q}}.`$ (92)
Here all quantities are functions of $`q`$, and $`𝑸=(\mathrm{},Q_{\alpha \beta },\mathrm{})^T`$, $`𝑷=(\mathrm{},P_{\alpha \beta },\mathrm{})^T`$, and $`𝑵=(\mathrm{},N_{\alpha \beta },\mathrm{})^T`$, are the vector representation of the matrix elements with $`\alpha >\beta `$. $`𝑨`$ and $`𝑩`$ are the matrices whose elements are given by
$`(𝑨)_{\alpha \beta ,\gamma \delta }`$ $`=`$ $`\delta _{\alpha \gamma }\delta _{\beta \delta }(e_\alpha +e_\beta )+v_{\alpha \beta ,\gamma \delta }^{22},`$ (94)
$`(𝑩)_{\alpha \beta ,\gamma \delta }`$ $`=`$ $`v_{\alpha \beta \gamma \delta }^{40},`$ (95)
in terms of the matrix elements of the moving frame Hamiltonian
$`\widehat{H}_M(q)`$ $`=`$ $`{\displaystyle \underset{\alpha }{}}e_\alpha a_\alpha ^{}a_\alpha `$ (100)
$`+{\displaystyle \frac{1}{4}}{\displaystyle \underset{\alpha \beta \gamma \delta }{}}v_{\alpha \beta ,\gamma \delta }^{22}a_\alpha ^{}a_\beta ^{}a_\delta a_\gamma `$
$`+{\displaystyle \frac{1}{4!}}{\displaystyle \underset{\alpha \beta \gamma \delta }{}}\left(v_{\alpha \beta \gamma \delta }^{40}a_\alpha ^{}a_\beta ^{}a_\gamma ^{}a_\delta ^{}+v_{\alpha \beta \gamma \delta }^{04}a_\delta a_\gamma a_\beta a_\alpha \right)`$
$`+{\displaystyle \frac{1}{3!}}{\displaystyle \underset{\alpha \beta \gamma \delta }{}}\left(v_{\alpha \beta \gamma ,\delta }^{31}a_\alpha ^{}a_\beta ^{}a_\gamma ^{}a_\delta +v_{\delta ,\alpha \beta \gamma }^{13}a_\delta ^{}a_\gamma a_\beta a_\alpha \right).`$
Here, due to Eq.(66), $`a^{}a^{}`$ and $`aa`$ parts of $`\widehat{H}_M(q)`$ vanish, and $`a^{}a`$ part of $`\widehat{H}_M(q)`$ is diagonalized. The matrix elements of the residual interactions in Eqs.(100)-(100) are antisymmetrized with respect to the quasiparticle indices. The matrices $`𝑨`$ and $`𝑩`$ have the same structures as the ones often defined in the quasiparticle RPA formalism . The matrix $`𝑫`$ is defined by
$$(𝑫)_{\alpha \beta ,\gamma \delta }=\frac{1}{2}\varphi (q)\left|[[[\widehat{H}_M(q),(\widehat{H}\lambda (q)\widehat{N})_A],a_\alpha ^{}a_\beta ^{}+a_\beta a_\alpha ],a_\gamma a_\delta ]\right|\varphi (q).$$
(101)
These matrix elements are expressed also in terms of the Hamiltonian matrix elements as
$`(𝑫)_{\alpha \beta ,\gamma \delta }`$ $`=(d_{\alpha \beta ,\gamma \delta }^{22}d_{\alpha \beta \gamma \delta }^{40}+d_{\alpha \gamma }^{11}\delta _{\beta \delta }d_{\beta \gamma }^{11}\delta _{\alpha \delta }d_{\alpha \delta }^{11}\delta _{\beta \gamma }+d_{\beta \delta }^{11}\delta _{\alpha \gamma })/2,`$ (103)
$`d_{\alpha \beta ,\gamma \delta }^{22}`$ $`={\displaystyle \underset{ϵ}{}}(v_{\alpha \beta ϵ,\gamma }^{31}h_{\delta ϵ}v_{\alpha \beta ϵ,\delta }^{31}h_{\gamma ϵ}v_{\alpha ,ϵ\gamma \delta }^{13}h_{\beta ϵ}+v_{\beta ,ϵ\gamma \delta }^{13}h_{\alpha ϵ}),`$ (104)
$`d_{\alpha \beta \gamma \delta }^{40}`$ $`={\displaystyle \underset{ϵ}{}}(v_{\alpha \beta \gamma ,ϵ}^{31}h_{ϵ\delta }v_{\beta \gamma \delta ,ϵ}^{31}h_{ϵ\alpha }+v_{\gamma \delta \alpha ,ϵ}^{31}h_{ϵ\beta }v_{\delta \alpha \beta ,ϵ}^{31}h_{ϵ\gamma }),`$ (105)
$`d_{\alpha \beta }^{11}`$ $`={\displaystyle \underset{\gamma >\delta }{}}(v_{\alpha ,\beta \gamma \delta }^{13}h_{\gamma \delta }v_{\gamma \delta \alpha ,\beta }^{31}h_{\gamma \delta }),`$ (106)
where $`h_{\alpha \beta }`$ is the matrix elements of $`(\widehat{H}\lambda \widehat{N})_A`$ defined by
$$(\widehat{H}\lambda \widehat{N})_A=\underset{\alpha >\beta }{}h_{\alpha \beta }(a_\alpha ^{}a_\beta ^{}+a_\beta a_\alpha ).$$
(107)
Note that $`𝑫`$ contains the matrix elements of the type $`v^{13}`$ and $`v^{31}`$. These terms of the Hamiltonian do not contribute to the standard RPA equations.
The solution of the matrix equations is obtained as follows. From Eq.(III B), one obtains
$`𝑸`$ $`=`$ $`\lambda ^{}B(q)\left((𝑨+𝑩)(𝑨𝑩)𝑫\mathrm{\Omega }\right)^1𝑵,`$ (109)
$`𝑷`$ $`=`$ $`\lambda ^{}(𝑨𝑩)\left((𝑨+𝑩)(𝑨𝑩)𝑫\mathrm{\Omega }\right)^1𝑵,`$ (110)
with
$$\mathrm{\Omega }=B(q)C(q).$$
(111)
The condition that the collective mode is orthogonal to the number operator, Eq.(90), gives the following equation
$$S(\mathrm{\Omega })𝑵^T(𝑨𝑩)\left((𝑨+𝑩)(𝑨𝑩)𝑫\mathrm{\Omega }\right)^1𝑵=0.$$
(112)
The quantity $`\mathrm{\Omega }=B(q)C(q)`$ represents the square of the frequency $`\omega =\sqrt{BC}`$ of the local harmonic mode, which is not necessarily positive. This equation can be regarded as a dispersion equation to determine $`\mathrm{\Omega }=\omega ^2`$ as a zero point of $`S(\mathrm{\Omega })`$. The normalization condition, Eq.(91), then gives a constraint on the value of $`\lambda ^2B(q)`$. The value of the mass parameter $`B(q)`$ is arbitrary, being related to the invariance under the point transformation Eq.(II A). A choice of the coordinate, $`q`$, specifies a value of the mass parameter, $`B(q)`$. In practice, the coordinate is often scaled so as to make the mass parameter unity.
When the residual interactions are the separable forces such as the monopole pairing and the quadruple-quadrupole forces, the local harmonic equations reduce to a simpler form. The dispersion equation for the separable interaction does not require a matrix inversion as in Eq.(112). The details are discussed in Appendix.
Ref. has discussed a problem of spurious (Nambu-Goldstone) modes for local harmonic approaches, and stated that the RPA equation at non-equilibrium points must be extended in order to guarantee separation of the spurious modes. However, no practical way of solving the equation was given because the equation has parameters which we do not have a method to calculate. In our present formulation, the RPA equation is indeed extended to assure the number conservation.
### C Construction of Collective Subspace
Let us finally give algorithms to construct the collective subspace $`|\varphi (q)`$ as a function of the collective coordinate $`q`$. Note that the local harmonic equations, Eqs.(81)-(83), are regarded as local equations in a sense that the equations can be solved independently for different values of $`q`$. At the HFB ground state, $`|\varphi _0`$, defined by the HFB equation
$$\delta \varphi _0\left|\widehat{H}\lambda _0\widehat{N}\right|\varphi _0=0,$$
(113)
we find $`V/q=0`$. Therefore, $`|\varphi _0`$ is always a state on the collective subspace because Eq.(81) is automatically satisfied. Eqs. (82) and (83) reduce to the standard RPA equations at $`|\varphi _0`$ since the last two terms in Eq.(83) vanishes. The operators $`\widehat{Q},\widehat{P}`$ are then determined as one of the normal modes of the RPA equation.
For non-equilibrium states, in general, Eq.(81) and the other two equations, (82) and (83), are coupled. We may solve the coupled equations in an iterative way. As discussed in Sect. III B, one can find the operators $`\widehat{Q}(q)^{(n)},\widehat{P}(q)^{(n)}`$ by solving Eqs.(82) and (83) for a given trial state $`|\varphi (q)^{(n)}`$ ($`n`$ denoting the iteration step). This defines the moving frame Hamiltonian $`\widehat{H}_M(q)^{(n+1)}=\widehat{H}\lambda (q)^{(n+1)}\widehat{N}\left(\frac{V}{q}\right)^{(n)}\widehat{Q}(q)^{(n)}`$, which can be used to construct a trial state $`|\varphi (q)^{(n+1)}`$ for the next iteration. If the iteration converges, one obtains a state $`|\varphi (q)`$ on which Eqs. (81)-(83) are simultaneously satisfied. Repeating the same procedure for different values of $`q`$, one finally obtains the collective subspace $`|\varphi (q)`$ and the collective Hamiltonian as a function of $`q`$.
We remark here that the operator $`\widehat{P}(q)`$ thus determined does not guarantee Eq.(49), although the other equations are satisfied. In this sense, the local harmonic solution is an approximate solution. The exact solution satisfying all the basic equations in Sect.II B may not exist in realistic situations. Only when the system is “exactly decoupled” , the above procedure gives the exact solution.
It is possible to choose another algorithm which satisfies Eq.(49) at the sacrifice of errors in Eq. (81). Let $`|\varphi (q_0)`$ be a solution that satisfies the basic equations at $`q=q_0`$. The infinitesimal generators $`\widehat{Q}(q_0),\widehat{P}(q_0)`$ are determined by solving Eqs.(82) and (83). Then one can generate the state $`|\varphi (q_0+\delta q)`$ for an infinitesimal shift of the collective coordinate as
$$|\varphi (q_0+\delta q)=e^{i\delta q\widehat{P}(q_0)}|\varphi (q_0).$$
(114)
Repeating this procedure, one can construct a collective subspace. This solution should coincide with the one solved by the previous method if the system is exactly decoupled. Difference between the two gives a quality of decoupling for the collective subspace in the adiabatic approximation. The second procedure can be used to provide an initial guess, $`|\varphi (q)^{(0)}`$, for the iteration of the first method.
## IV Extension to multi-dimensional collective subspace
In this section we extend the adiabatic SCC method to a case of multi-dimensional collective subspace described by $`D`$ collective coordinates and conjugate momenta $`\{q^i,p_i;i=1,\mathrm{},D\}`$.
One can easily derive the basic equations of the adiabatic SCC method in parallel to the derivation given in Sections II by noting first that Eqs.(33,II B) are now extended to
$`|\varphi (q,p,N)=e^{i\widehat{G}(q,p,n)}|\varphi (q),`$ (115)
$`\widehat{G}=p_i\widehat{Q}^i(q)+n\widehat{\mathrm{\Theta }}(q),`$ (116)
where the operator $`\widehat{Q}^i(q)`$ have now $`D`$ components having coordinate label $`i`$. It is implied here and hereafter that the same coordinate index ($`i`$ in the above expression) appearing in the super- and subscripts means to take the summation over it. The infinitesimal generator $`\widehat{P}_i(q)`$ have also $`D`$ components each of which is related to the derivative $`i\frac{}{q^i}|\varphi (q)`$. In the following, the coordinate dependence is often omitted. For instance, $`B^{ij}(q)`$ and $`\widehat{Q}^i(q)`$ will be simply denoted by $`B^{ij}`$ and $`\widehat{Q}^i`$, respectively.
The adiabatic collective Hamiltonian is expressed as
$`(q,p,N)`$ $`=`$ $`V(q)+{\displaystyle \frac{1}{2}}B^{ij}(q)p_ip_j+\lambda (q)n.`$ (117)
The zeroth and the first order equations of the collective subspace are derived as
$`\delta \varphi (q)\left|\widehat{H}\lambda (q)\widehat{N}{\displaystyle \frac{V}{q^i}}\widehat{Q}^i\right|\varphi (q)=0,`$ (118)
$`\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},\widehat{Q}^i]{\displaystyle \frac{1}{i}}B^{ij}\widehat{P}_j\right|\varphi (q)=0,`$ (119)
while the second order equation becomes
$$\delta \varphi (q)\left|\frac{1}{2}[\widehat{H}\lambda (q)\widehat{N},\widehat{Q}^i,\widehat{Q}^j]+\frac{1}{6}[\frac{V}{q^k}\widehat{Q}^k,\widehat{Q}^i,\widehat{Q}^j]\frac{1}{2}(B^{ik}\widehat{Q}_{;k}^j+B^{jk}\widehat{Q}_{;k}^i)\right|\varphi (q)=0$$
(120)
with
$`\widehat{Q}_{;j}^i={\displaystyle \frac{\widehat{Q}^i}{q^j}}+\mathrm{\Gamma }_{kj}^i\widehat{Q}^k,`$ (121)
$`\mathrm{\Gamma }_{kj}^i={\displaystyle \frac{1}{2}}B^{il}\left({\displaystyle \frac{B_{lk}}{q^j}}+{\displaystyle \frac{B_{lj}}{q^k}}{\displaystyle \frac{B_{kj}}{q^l}}\right),`$ (122)
where $`B_{ij}`$ is the inverse matrix of $`B^{ij}`$ and the bracket including three operators defined by
$$[A,B,C]=\frac{1}{2}([[A,B],C]+[[A,C],B]).$$
(123)
Expanding the canonical variable condition with respect to $`p_i`$ and $`n`$, the following equations are derived;
$`\varphi (q)\left|\widehat{P}_i\right|\varphi (q)`$ $`=0,`$ (124)
$`\varphi (q)\left|\widehat{N}\right|\varphi (q)`$ $`=N_0,`$ (125)
and
$`\varphi (q)\left|[\widehat{Q}^i,\widehat{P}_j]\right|\varphi (q)=i\delta _{ij},`$ (126)
$`\varphi (q)\left|[\widehat{Q}^i,\widehat{N}]\right|\varphi (q)=0,`$ (127)
$`\varphi (q)\left|[\widehat{P}_i,\widehat{N}]\right|\varphi (q)=0.`$ (128)
These basic equations are invariant under the point transformation of the collective variables
$`q^iq^i=q^i(q),`$ (130)
$`p_ip_i^{}=p_j\times \left(q^j/q^i\right).`$ (131)
We have adopted the vector-tensor notation to manifest the transformation properties under the point transformation. Quantities which have a coordinate index in the subscript and in the superscript have the transformation properties of the covariant and contravariant vectors, respectively. For example,
$`\widehat{Q}^i\widehat{Q}^i=\widehat{Q}^j\times \left(q^i/q^j\right),`$ (132)
$`\widehat{P}_i\widehat{P}_i^{}=\widehat{P}_j\times \left(q^j/q^i\right).`$ (133)
The mass tensor $`B^{ij}`$ is the contravariant tensor of second rank. The operator $`\widehat{Q}_{;j}^i`$ defined by Eq.(121) is the covariant derivative of $`\widehat{Q}^i`$, and $`\mathrm{\Gamma }_{kj}^i`$ is the Christoffel symbol where the mass tensor $`B_{ij}`$ plays the role of metric tensor.
Let us now derive local harmonic equations of collective subspace. Taking the $`q`$-derivative, the zeroth order equation (118) leads to
$`\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},{\displaystyle \frac{1}{i}}\widehat{P}_i]C_{ij}(q)\widehat{Q}^j{\displaystyle \frac{V}{q^j}}\widehat{Q}_{;i}^j{\displaystyle \frac{\lambda }{q^i}}\widehat{N}\right|\varphi (q)=0,`$ (134)
$`C_{ij}(q)={\displaystyle \frac{^2V}{q^iq^j}}\mathrm{\Gamma }_{ij}^k{\displaystyle \frac{V}{q^k}}.`$ (135)
As we have done for the $`D=1`$ case, we would like to eliminate the covariant derivative $`\widehat{Q}_{;i}^j`$ in Eq.(134) in order to give a feasible form of the local harmonic equation. This was done for the $`D=1`$ case with help of the second order equation of the collective subspace. The corresponding equations (120) give $`D(D+1)/2`$ constraints, while number of unknown parameters, $`\widehat{Q}_{;i}^j`$, is $`D^2`$. In fact, Eq.(120) is equivalent to
$$\delta \varphi (q)\left|\frac{1}{2}[[\widehat{H}\lambda (q)\widehat{N},\widehat{Q}^j],\widehat{Q}^i]+\frac{1}{6}[[\frac{V}{q^k}\widehat{Q}^k,\widehat{Q}^j],\widehat{Q}^i](B^{ik}\widehat{Q}_{;k}^j+\widehat{R}^{ij})\right|\varphi (q)=0,$$
(136)
where $`\widehat{R}^{ij}`$ are arbitrary one-body operators which are antisymmetric for exchange of indices $`i`$ and $`j`$. If we choose $`\widehat{R}^{ij}=0`$, we can eliminate the derivative term $`\frac{V}{q^j}\widehat{Q}_{;i}^j`$. Then, Eq.(134) leads to
$$\delta \varphi (q)\left|[\widehat{H}\lambda (q)\widehat{N},\frac{1}{i}\widehat{P}_i]C_{ij}\widehat{Q}^j\frac{1}{2}[[\widehat{H}\lambda (q)\widehat{N},(\widehat{H}\lambda (q)\widehat{N})_A],B_{ij}\widehat{Q}^j]\frac{\lambda }{q^i}\widehat{N}\right|\varphi (q)=0.$$
(137)
This equation is an analog of Eq.(83) and is linear with respect to the infinitesimal generators $`\widehat{Q}^i,\widehat{P}_i`$. We can numerically solve Eqs.(118), (119) and (137) on the same line as in Sect.III B and III C.
It should be remarked that the local harmonic equation Eq.(137) for $`D>1`$ are derived from equations, (118) and (120), but with an additional condition $`\widehat{R}^{ij}=0`$ in Eq.(136). This condition is introduced to obtain the local harmonic equations parallel to the one-dimensional case.
## V Conclusions
We have formulated the adiabatic approximation to the general framework of the selfconsistent collective coordinate method in order to describe large amplitude collective motions in superconducting nuclei. The formalism, based on the TDHFB equations of motion, guarantees the conservation of particle number in a transparent way. We have shown that the equations of collective subspace are reduced to local linear equations for the infinitesimal generators, which can be solved with use of quasiparticle representation of the Hamiltonian matrix elements. This provides a complete procedure to determine the states $`e^{ip\widehat{Q}(q)}|\varphi (q)`$ in the collective subspace and the collective Hamiltonian $`(q,p)=V(q)+\frac{1}{2}B(q)p^2`$ as a function of the collective coordinate $`q`$ and momentum $`p`$. A possible extension to the case of the multi-dimensional collective coordinates is also discussed.
We emphasize that the equations given in this paper are solvable by means of the matrix method similar to the standard RPA. We hope that the present adiabatic theory is useful to solve number of open questions in the realistic studies of nuclear large amplitude collective motion.
## Solution for the Separable Interactions
In this appendix, we give solutions of the local harmonic equations of collective subspace for the case where the two-body interaction is given by the separable forces. We assume that the Hamiltonian is given by
$$\widehat{H}=\widehat{h}_0\frac{\kappa }{2}\widehat{F}^{}\widehat{F},$$
(138)
where $`\widehat{h}_0(=\widehat{h}_0^{})`$ and $`\widehat{F}`$ are one-body operators. Equivalently, one may write
$`\widehat{H}=\widehat{h}_0{\displaystyle \frac{\kappa }{2}}\widehat{F}^{(+)}\widehat{F}^{(+)}+{\displaystyle \frac{\kappa }{2}}\widehat{F}^{()}\widehat{F}^{()},`$ (139)
$`\widehat{F}^{(\pm )}(\widehat{F}\pm \widehat{F}^{})/2=\pm \widehat{F}^{(\pm )}.`$ (140)
For the separable forces, it is customary to neglect the Fock term of the forces. This approximation is easily and consistently implemented in the SCCM by assuming that the equation of motion for the time-dependent mean-field state $`|\varphi (t)`$ is now given by the time-dependent Hartree-Bogoliubov equation without the Fock terms,
$`\delta \varphi (t)\left|i{\displaystyle \frac{}{t}}\widehat{h}(t)\right|\varphi (t)=0,`$ (141)
$`\widehat{h}(t)=\widehat{h}_0\kappa \widehat{F}^{(+)}\varphi (t)\left|\widehat{F}^{(+)}\right|\varphi (t)+\kappa \widehat{F}^{()}\varphi (t)\left|\widehat{F}^{()}\right|\varphi (t).`$ (142)
The local harmonic equations (81)-(83) then become
$$\delta \varphi (q)\left|\widehat{h}_M(q)\right|\varphi (q)=0,$$
(143)
$$\delta \varphi (q)\left|[\widehat{h}_M(q),\widehat{Q}(q)]f_Q^{()}\widehat{F}^{()}\frac{1}{i}B(q)\widehat{P}(q)\right|\varphi (q)=0,$$
(144)
$`\delta \varphi (q)|[\widehat{h}_M(q),{\displaystyle \frac{1}{i}}B(q)\widehat{P}(q)]f_P^{(+)}\widehat{F}^{(+)}B(q)C(q)\widehat{Q}(q)f_R^{(+)}\widehat{F}^{(+)}`$ (145)
$`[\widehat{F}^{()},(\widehat{h}(q)\lambda (q)\widehat{N})_A]f_Q^{()}f_N\widehat{N}|\varphi (q)=0,`$ (146)
where $`\widehat{h}_M(q)`$ is the mean-field Hamiltonian in the moving frame defined by
$`\widehat{h}_M(q)=\widehat{h}(q){\displaystyle \frac{V}{q}}\widehat{Q}(q)\lambda (q)\widehat{N},`$ (147)
$`\widehat{h}(q)=\widehat{h}_0\kappa \widehat{F}^{(+)}\varphi (q)\left|\widehat{F}^{(+)}\right|\varphi (q),`$ (148)
and definitions of other symbols are
$`f_Q^{()}=\kappa \varphi (q)\left|[\widehat{F}^{()},\widehat{Q}(q)]\right|\varphi (q),`$ (150)
$`f_P^{(+)}=\kappa \varphi (q)\left|[\widehat{F}^{(+)},{\displaystyle \frac{1}{i}}B(q)\widehat{P}(q)]\right|\varphi (q),`$ (151)
$`f_R^{(+)}=\kappa \varphi (q)\left|[[\widehat{F}^{(+)},(\widehat{h}(q)\lambda (q)\widehat{N})_A],\widehat{Q}(q)]\right|\varphi (q)/2,`$ (152)
$`f_N=B(q){\displaystyle \frac{\lambda }{q}}.`$ (153)
We express all operators in the above equations in terms of the quasiparticle operators $`\{a_\alpha ^{},a_\alpha \}`$ defined for $`\widehat{h}_M(q)`$ and $`|\varphi (q)`$. For example,
$`\widehat{h}_M(q)={\displaystyle \underset{\alpha }{}}e_\alpha a_\alpha ^{}a_\alpha ,`$ (154)
$`\widehat{F}^{(+)}={\displaystyle \underset{\alpha >\beta }{}}F_{\alpha \beta }^{(+)}(a_\alpha ^{}a_\beta ^{}+a_\beta a_\alpha )+{\displaystyle \underset{\alpha \beta }{}}F_{B,\alpha \beta }^{(+)}a_\alpha ^{}a_\beta ,`$ (156)
$`\widehat{F}^{()}={\displaystyle \underset{\alpha >\beta }{}}F_{\alpha \beta }^{()}(a_\alpha ^{}a_\beta ^{}a_\beta a_\alpha )+{\displaystyle \underset{\alpha \beta }{}}F_{B,\alpha \beta }^{()}a_\alpha ^{}a_\beta .`$
We have assumed that all matrix elements are real. Equations (144,145) are then reduced to linear equations for the matrix elements $`Q_{\alpha \beta },P_{\alpha \beta }`$ of the infinitesimal generators $`\widehat{Q}(q),\widehat{P}(q)`$. They are easily solved to give the expression
$`Q_{\alpha \beta }={\displaystyle \frac{e_\alpha +e_\beta }{(e_\alpha +e_\beta )^2\mathrm{\Omega }}}F_{\alpha \beta }^{()}f_Q^{()}+{\displaystyle \frac{1}{(e_\alpha +e_\beta )^2\mathrm{\Omega }}}\left(F_{\alpha \beta }^{(+)}f_{PR}^{(+)}+R_{\alpha \beta }^{()}f_Q^{()}+N_{\alpha \beta }f_N\right),`$ (157)
$`BP_{\alpha \beta }={\displaystyle \frac{e_\alpha +e_\beta }{(e_\alpha +e_\beta )^2\mathrm{\Omega }}}\left(F_{\alpha \beta }^{(+)}f_{PR}^{(+)}+R_{\alpha \beta }^{()}f_Q^{()}+N_{\alpha \beta }f_N\right)+{\displaystyle \frac{\mathrm{\Omega }}{(e_\alpha +e_\beta )^2\mathrm{\Omega }}}F_{\alpha \beta }^{()}f_Q^{()},`$ (158)
$`f_{PR}^{(+)}=f_P^{(+)}+f_R^{(+)},`$ (159)
where we introduced the one-body operator
$$\widehat{R}(q)^{(\pm )}[\widehat{F}_B^{(\pm )}(q),(\widehat{h}(q)\lambda (q)\widehat{N})_A]=\underset{\alpha >\beta }{}R_{\alpha \beta }^{(\pm )}(a_\alpha ^{}a_\beta ^{}a_\beta a_\alpha ),$$
(160)
with $`\widehat{F}_B^{(\pm )}(q)`$ being the last terms of $`\widehat{F}^{(\pm )}`$ in Eqs.(156,156).
Inserting this expression for the definition of $`f_{PR}^{(+)},f_Q^{()}`$, we obtain the equations for unknown quantities $`f_{PR}^{(+)},f_Q^{()},f_N`$. Similarly, the condition of orthogonality to the number operator Eq.(90) gives another equation for $`f_{PR}^{(+)},f_Q^{()},f_N`$. They are summarized as a linear homogeneous equation which can be written in a $`3\times 3`$ matrix form
$$\left(\begin{array}{ccc}& & \\ & S_{xx^{}}(\mathrm{\Omega })& \end{array}\right)\left(\begin{array}{c}f_{PR}^{(+)}\\ f_Q^{()}\\ f_N\end{array}\right)=0,$$
(161)
where
$`S_{11}=2S_{F^{(+)}F^{(+)}}^{(1)}+S_{R^{(+)}F^{(+)}}^{(2)}{\displaystyle \frac{1}{\kappa }},`$ (163)
$`S_{12}=2\mathrm{\Omega }S_{F^{(+)}F^{()}}^{(2)}+2S_{F^{(+)}R^{()}}^{(1)}+S_{R^{(+)}F^{()}}^{(1)}+S_{R^{(+)}R^{()}}^{(2)},`$ (164)
$`S_{13}=2S_{F^{(+)}N}^{(1)}+S_{R^{(+)}N}^{(2)},`$ (165)
$`S_{21}=2S_{F^{()}F^{(+)}}^{(2)},`$ (166)
$`S_{22}=2S_{F^{()}F^{()}}^{(1)}+2S_{F^{()}R^{()}}^{(2)}{\displaystyle \frac{1}{\kappa }},`$ (167)
$`S_{23}=2S_{F^{()}N}^{(2)},`$ (168)
$`S_{31}=S_{NF^{(+)}}^{(1)},`$ (169)
$`S_{32}=\mathrm{\Omega }S_{NF^{()}}^{(2)}+S_{NR^{()}}^{(1)},`$ (170)
$`S_{33}=S_{NN}^{(1)}.`$ (171)
The functions $`S_{XY}^{(1)}`$ with the symbols $`X,Y`$ denoting $`(X,Y)=(F^{(+)},F^{(+)}),`$ $`(F^{(+)},R^{()}),`$ $`(F^{(+)},N),`$ $`(R^{(+)},F^{()}),`$ $`(F^{()},F^{()}),`$ $`(N,N),`$ $`(N,F^{(+)}),`$ $`(N,R^{()})`$ are given by
$$S_{XY}^{(1)}=\underset{\alpha >\beta }{}\frac{e_\alpha +e_\beta }{(e_\alpha +e_\beta )^2\mathrm{\Omega }}X_{\alpha \beta }Y_{\alpha \beta },$$
(172)
while the functions $`S_{XY}^{(2)}`$ with $`(X,Y)=(F^{(+)},F^{()}),`$ $`(R^{(+)},F^{(+)}),`$ $`(R^{(+)},R^{()}),`$ $`(R^{(+)},N),`$ $`(F^{()},F^{(+)}),`$ $`(F^{()},R^{()}),`$ $`(F^{()},N),`$ $`(N,F^{()})`$ are given by
$$S_{XY}^{(2)}=\underset{\alpha >\beta }{}\frac{1}{(e_\alpha +e_\beta )^2\mathrm{\Omega }}X_{\alpha \beta }Y_{\alpha \beta }.$$
(173)
The value of $`\mathrm{\Omega }`$ is determined by finding the zero point of the dispersion equation
$$\mathrm{det}\{S_{xx^{}}(\mathrm{\Omega })\}=0.$$
(174)
Normalizations of $`f_{PR}^{(+)},f_Q^{()},f_N`$ are fixed by the condition Eq.(91). It is straightforward to extend the above procedure to the case where the two-body interaction is given by a sum of the separable forces. |
warning/0001/nlin0001032.html | ar5iv | text | # Statistics of pressure and of pressure-velocity correlations in isotropic turbulence.
## 1 Introduction
Scaling in turbulent flows is one of the most challenging open issue in fluid-dynamics . Typical problems concern both the understanding of the ideal case of isotropic and homogeneous turbulence in the limit of high Reynolds numbers or more realistic and applied situations with anisotropic and inhomogeneous statistics (for recent examples see ). In 1941, Kolmogorov, used a clever application of dimensional analysis to predict that the scaling of velocity increments in the inertial range should have a power law behaviour depending only on the averaged energy dissipation in the flow, $`ϵ`$. Namely, for structure function of order $`p`$, i.e. the p-th moment of a velocity difference across a distance $`R`$ we have:
$$S_q(R)=\left(v(x+R)v(x)\right)^qϵ^{q/3}R^{q/3}$$
(1)
with $`\eta rL_0`$ where $`\eta `$ is the dissipative scale and $`L_0`$ is the typical external scale where forcing acts. Let us notice that in (1) we have explicitly neglected any tensorial structure in the velocity field such as to stress the typical dimensional character of the Kolmogorov theory, i.e. the scaling properties must be the same for any observable which has the same physical dimension and which is built in terms of local field increments, $`\delta _rv(x)=v(x+r)v(x)`$.
Kolmogorov theory, as previously summarized, is quantitatively wrong. Experiments and numerical simulations show a quantitative disagreement with the dimensional prediction $`p/3`$ for the scaling exponents. For example, the longitudinal velocity structure functions:
$$S_q^v\left(r\right)=\left|\left(v_i\left(𝐱+𝐑\right)v_i\left(𝐱\right)\right)\widehat{R}_i\right|^qR^{\zeta _v(p)}$$
(2)
show a power law behaviour with a set of exponents $`\zeta _v(p)`$ non linear in $`p`$.
The failure of the dimensional estimate $`p/3`$ goes under the name of anomalous scaling.
Many problems naturally arises as a consequence of the failure of the main Kolmogorov prediction. The main open problem is to find an analytical way to calculate from first principle the anomalous exponents, a problem which is still out of control but for the case of anomalous exponents characterizing the statistics of passive quantities advected by gaussian velocity fields .
Another interesting question, opened by the failure of Kolmogorov dimensional prediction, consists in the possibility that local observable with the same physical dimensions but with different tensorial structures have different scaling properties. About this point, there are some experimental and numerical evidences regarding different possible anomalous behaviour of longitudinal and transversal structure functions even in isotropic turbulence. Somehow related to this issue is also the apparent different anomalous scaling between the coarse grained averages of dissipative quantities like enstrophy and energy dissipation . On the other hand, on the basis of a SO(3) decomposition of velocity correlation functions, other authors claim that the supposed different scaling of quantities like transversal and longitudinal structure functions can only be due to spurious sub-leading non-isotropic effects, i.e. in isotropic high-Reynolds numbers all components of the same tensorial observable should have the same -maybe anomalous- scaling behaviour. A first numerical support to this claim has been presented in the analysis of a channel flow simulation in .
Even more complex is the situation when multi-point pressure correlations is involved . Dimensionally speaking pressure is just a velocity squared, and Kolmogorov-like argument can be easily generalized to the case of pressure structure functions, $`F_q(r)`$. Indeed, a simple application of dimensional analysis leads to :
$$F_q(r)\left|P(x)P(x+r)\right|^qϵ^{2q}r^{2q}$$
(3)
where as usual, all distances are supposed to belong to the inertial range of scales. Of course, intermittency will also affect pressure scaling. By following the straightforward hypothesis that pressure can be treated as a velocity squared one would be tempted to assign the same intermittency exponents of the velocity field to the pressure scaling, i.e. to replace (3) with
$$F_q(r)r^{\zeta _v(2q)}$$
(4)
This prediction is just a simple consequences of the assumptions that all velocity correlations have the same scaling behaviour supposed that all distances involved are in the inertial range and that the statistics is locally isotropic. Such a dimensional ansatz has been questioned on the basis of a phenomenological argument in , some numerical support to the latter argument have been recently presented in .
In this paper we will mainly present some numerical evidences that indeed the dimensional ansatz (4) is correct. It is well known that this must be the case at least for for $`q=2`$ in (4). In this case, there exist an exact relation which connect the scaling of the second order pressure structure function with a linear integral combination of fourth-order velocity structure functions. The problem is if the exact result can be simply extrapolated to other pressure-dependent observable and, in the case, how strong finite-Reynolds effects can be. Indeed, one may argue that pressure feels strongly non-local effects, being just the inversion of the Poisson problem $`\mathrm{\Delta }P=_i_jv_iv_j`$, and therefore the assumptions of independence from large scales and/or from boundary conditions may not be satisfied even at very high Reynolds numbers. Indeed, to our knowledge, neither experimental studies nor numerical simulations have ever been able to make a firm quantitative statement about pressure scaling properties . In this paper we show that it is possible to find a set of velocity-pressure observable which have indeed a quite good scaling behaviour in agreement with the dimensional ansatz (4) also at moderate Reynolds numbers.
Scaling in turbulence is particularly difficult to test in both experiments and numerical simulations. Experiments reaches high Reynolds numbers by paying the price to have a very limited set of information on the whole velocity fields, typically only a long time series of one velocity components in a few spatial points. Moreover, in most cases, there is not a precise control of the degree of isotropy and homogeneity in the flow. On the other hand, numerical simulations have a perfectly controlled lay out, the velocity field is exactly known at any point, but the maximum reachable Reynolds number is still order of magnitude smaller then in typical experiments .
Nevertheless, numerical simulations, if exploited in a clever way, are the only tool where complex measurements can be performed. Therefore, questions like the dependency of scaling properties from the tensorial nature of the observable can, up to know, be investigated only in numerical data base.
In this paper, we present a detailed analysis of pressure scaling and pressure-velocity correlations scaling in a set of moderate Reynolds number simulations.
Starting from the analysis of the energy transfer in real space we propose a set of pressure-velocity observable which show better scaling properties then the usual pressure structure functions. We presents quantitative evidences that indeed, while pressure structure functions are strongly affected from Reynolds numbers effects, the pressure-velocity correlation functions we investigated have a fairly good scaling behaviour, even at modest Reynolds numbers, in agreement with the hypothesis that pressure ”behaves” like a velocity squared. In order to understand whether the bad scaling behaviour detected in the pure-pressure structure functions is due to spurious anisotropic sub-leading effects we also present some results on the SO(3) decomposition of the pressure field.
The paper is organized as follows. In section 2 we summarize the known analitycal result which connect the second order pressure structure function to the integral linear combination of fourth-order velocity correlations and the experimental and numerical attempts to test the relation. In section 3 we introduce the set of pressure-velocity correlations which should have a better scaling properties on the basis of a simple argument based on the energy transfer of Navier-Stokes equations in the real space. In section 4 we present the analysis of our numerical data base. In section 5 we briefly comment on the analysis of non isotropic fluctuations. Conclusions follow in section 6.
## 2 Pressure structure functions
Under the assumptions of local isotropy, local homogeneity, incompressibility, and by use of Navier-Stokes equation, one can relate the second order pressure structure functions, $`F_2(r)`$, to some fourth-order velocity structure functions . Namely:
$`F_2(r){\displaystyle \frac{1}{3}}D_{1111}(r)+{\displaystyle \frac{4}{3}}r^2{\displaystyle _r^{\mathrm{}}}y^3\left[D_{1111}(y)+D_{\beta \beta \beta \beta }(y)6D_{11\gamma \gamma }\right]𝑑y`$
$`+{\displaystyle \frac{4}{3}}{\displaystyle _r^{\mathrm{}}}y^1\left[D_{\beta \beta \beta \beta }(y)3D_{11\gamma \gamma }\right]𝑑y`$ (5)
where the fourth-order structure function is
$$D_{ijkl}(\stackrel{}{r})(u_iu_i^{})(u_ju_j^{})(u_ku_k^{})(u_lu_l^{})$$
and where for simplicity we have used primed variables to express velocities at the position $`\stackrel{}{x^{}}=\stackrel{}{x}+\stackrel{}{r}`$ and where $`i,j,k,l`$ is $`1`$ if the velocity component is parallel to the separation vector, $`\stackrel{}{r}`$, and $`2,3`$ otherwise. Subscripts $`\beta ,\gamma `$ denote either $`2`$ or $`3`$. Of course, (5) implies that whenever the fourth-order structure functions entering in the above expressions are all dominated by the inertial-range intermittent scaling behaviour, $`D_{i,j,k,l}(r)r^{\zeta _v(4)}`$, then also the second order pressure structure functions should scale with the exponent $`\zeta _p(2)=\zeta _v(4)`$. Relation (5) have been carefully tested in numerical simulations without any appreciable deviations . Nevertheless, the overall scaling behaviour of the pressure structure function is very poor. Similarly, the analysis of experimental data does not show any power law behaviour for the pressure structure functions even if the Reynolds number was extremely high ($`Re_\lambda 10000`$). In the latter case, authors tried to explain the difference between pressure scaling quality and velocity scaling quality by invoking a possible different scaling for the different velocity correlations entering in the RHS of (5), leading to the final prediction that pressure structure functions is made in terms of different power law contributions with slightly different exponents. The resulting superposition of power laws would be the responsible of the poor observed scaling behaviour. This statement would anyhow contradict the theoretical prediction made in terms of the SO(3) decomposition which forbids different component of the same tensorial observable to scale differently in a isotropic ensemble.
Another interesting remark consists in the strong cancellation among the different contribution of (5) observed in numerical simulations : the LHS of (5) is more then an order of magnitude smaller than the single different contributions entering in the RHS. One, cannot exclude apriori the possibility that there exist an almost perfect cancellations of all leading scaling terms of all contribution appearing in the RHS of (5), even if such a perfect cancellation would call for some unknown physical interpretation.
More probably, the cancellation is not perfect but strong enough to hide completely the pressure scaling at the available experimental and numerical Reynolds numbers.
On the other hand, the possibility that pressure-increments behave as velocity-increments, $`\delta P\delta v`$, instead than as a velocity-increment squared has been recently proposed . This would violate the exact results previously reported and therefore cannot be correct unless strong anisotropic effects are present at all scales.
In order to better assess the pressure statistical properties we present in the following, some results for pressure-dependent observable. This observable do not posses the strong cancellation properties showed by structure function.
## 3 Pressure-velocity correlation
Let us start by looking at the energy balance inside any volume $`V`$ of the flow. From the Navier-Stokes eqs we obviously have:
$$_tE_V+_Vv_i(𝐱)v_j(𝐱)_jv_i(𝐱)d𝐱+_Vv_i(𝐱)_iP(𝐱)d𝐱=\nu _Vv_i(𝐱)\mathrm{\Delta }v_i(𝐱)𝑑𝐱$$
(6)
where with $`E_V=1/2_V𝑑𝐱v_iv_i`$ we denote the total energy in the sub-volume $`V`$. Let us notice that the two terms in the LHS of (6) can be obviously written as the fluxes across the boundaries of $`V`$ by using Gauss theorem:
$$_Vv_i(𝐱)v_j(𝐱)_jv_i(𝐱)d𝐱_{\mathrm{\Omega }_V}𝑑\mathrm{\Sigma }n_jv_j(𝐱)v^2(𝐱)\mathrm{\Phi }_{\mathrm{\Omega }_V}(𝐯v^2)$$
(7)
$$_Vv_i(𝐱)_iP(𝐱)d𝐱_{\mathrm{\Omega }_V}𝑑\mathrm{\Sigma }n_iv_i(𝐱)P(𝐱)\mathrm{\Phi }_{\mathrm{\Omega }_V}(𝐯P)$$
(8)
where with $`n_i`$ we denote the versor perpendicular to the infinitesimal surface on the boundaries of $`V`$. Rewritten in this way, relation (6) is just a simple restatement of the conservation of energy: the total energy change inside a volume is given by the flux across the volume surface and by the energy dissipation inside the volume. Let us now use this simple fact in order to extract some useful guess about scaling properties of velocity-pressure correlations. Let us consider a very particular class of volume $`V`$, i.e. a cylinder with an infinitesimal squared basis of surface $`ϵ`$ and with a finite axis in the direction of $`𝐑`$. In the limit when the basis becomes smaller and smaller, the flux across the lateral sides goes to zero because contributions from two opposite walls are equal but with different sign. The only contributions to the total flux come from the two infinitesimal basis and can be written as
$$\mathrm{\Phi }_{\mathrm{\Omega }_V}(𝐯v^2)=ϵ(v_i(𝐱)v^2(𝐱)v_i(𝐱+𝐑)v^2(𝐱+𝐑))\widehat{R}_i$$
(9)
for the flux involving the velocity correlation and as
$$\mathrm{\Phi }_{\mathrm{\Omega }_V}(𝐯P)=ϵ(v_i(𝐱)P(𝐱)v_i(𝐱+𝐑)P(𝐱+𝐑))R_i$$
(10)
for the flux involving pressure-velocity correlations. In both cases we have exploited the fact that the two infinitesimal basis are centered in $`𝐱`$ and in $`𝐱+𝐑`$, i.e. their versor is oriented along $`\widehat{R}`$. Similarly the two volume integral giving the time variation of the total energy and the energy dissipation becomes two linear integral times the infinitesimal basis area $`ϵ`$: $`_t(ϵ_𝐑𝑑sv^2(𝐱))`$ and $`(ϵ_𝐑𝑑sv_i(𝐱)\mathrm{\Delta }v_i(𝐱))`$ respectively, where with $`ds`$ we parametrized the segment going from $`𝐱`$ to $`𝐱+𝐑`$.
Let us now assume that all the four observable entering in the energy balance have the same statistical behaviour. This is somehow a “local Kolmogorov refined hypothesis”: we link the scaling of the local energy dissipation to the scaling of some particular third order velocity correlation and to the scaling of a velocity-pressure correlation. The claim is therefore that the particular structure functions emerging from our flux analysis should scale exactly like the coarse grained energy dissipation, i.e. should have an anomalous scaling like the usual longitudinal structure functions which satisfy the original Kolmogorov Refined Hypothesis. In the next section we present some numerical data in support of this claim.
## 4 Numerical Analysis
The data set we are going to analyze has been obtained from a direct numerical integration of NS eqs using a pseudo-spectral method with dealiasing on a grid of $`128^3`$ points. The forcing was implemented isotropically on all wavevectors with $`|k|<1`$ such as to enforce the $`k^{5/3}`$ spectrum at small wavectors . We have analyzed about 100 configurations stored each eddy-turn over time. The simulation has a $`Re_\lambda =70`$.
Let us denote with:
$`S_q^{vv^2}(R)`$ $`=`$ $`\left|v_i(𝐱)v^2(𝐱)v_i(𝐱+𝐑)v^2(𝐱+𝐑)\widehat{R}_i\right|^{q/3}`$
$`S_q^{vP}(R)`$ $`=`$ $`\left|\left(v_i(𝐱)P(𝐱)v_i(𝐱+𝐑)P(𝐱+𝐑)\right)\widehat{R}_i\right|^{q/3}`$ (11)
the two different structure functions which can be made in terms of the two flux quantities defined in the previous section. Let us notice that in (11) both $`S_q^{vv^2}(R)`$ and $`S_q^{vP}(R)`$ have been defined as the $`q/3`$ power of the original fluxes such as to have the same dimensions of $`S_q^v(R)`$.
Let us start by showing in Fig.1 the strong cancellation effects present in the pressure structure functions $`F_q(R)`$ with respect to the velocity longitudinal structure functions with the same physical dimensions $`S_{2q}^v(R)`$. In Fig. 1 we show the log-log plot of $`F_1(R)`$ and of $`S_2(R)`$, as one can see the overall amplitude of pressure fluctuations is about one order of magnitude smaller than the velocity fluctuations. This is just to confirm that pressure by itself is a much weaker signal than the usual velocity correlations.
As one can see in Fig. 1 the scaling is quite poor, as one can expect in any DNS. As usual, in order to extract quantitative statement about scaling exponents one has to exploit the Extended Self Similarity (ESS) property enjoyed by homogeneous and isotropic turbulent flows . ESS consists in looking for relative scaling of two different observable. Usually, one takes two structure functions of two different orders, i.e. in the case of longitudinal structure functions $`S_q^v(R)\left[S_q^{}^v(R)\right]^{\zeta _v(q)/\zeta _v(q^{})}`$.
Let us now define the same relative scaling for the two generalized structure functions defined in (11).
$$S_q^{vv^2}(R)\left[S_q^{}^{vv^2}(R)\right]^{\frac{\zeta _{vv^2}(q)}{\zeta _{vv^2}(q^{})}},S_q^{vP}(R)\left[S_q^{}^{vP}(R)\right]^{\frac{\zeta _{vP}(q)}{\zeta _{vP}(q^{})}}$$
(12)
In Figs. 2,3,4 we show the ESS plot for the two generalized structure functions and for the usual longitudinal structure functions respectively, with $`q=1,q^{}=2`$. As one can see the all ESS plots showed a scaling behaviour consistent with the usual homogeneous and isotropic high Reynolds value which give for the relative exponents: $`\zeta _v(2)/\zeta _v(1)=1.92\pm 0.02`$ . Similar agreements are found for higher order moments (not showed).
The scaling ansatz assumed for the generalized structure functions seems therefore quite well satisfied. These findings support the fact that pressure does not behave abnormally as far as its “dimensional” scaling properties are concerned. Indeed, pressure-velocity correlations behaves exactly like velocity-velocity correlations once pressure is counted as a “velocity squared”. Nevertheless, The only realistic way to perform a quantitative statement about scaling exponents is to study the logarithmic local slopes of (12). Only when logarithmic local slopes show a fairly constant behaviour one can really speak about scaling. In Fig. 5 we show the logarithmic local slopes of (12) for $`q=1,q^{}=2`$ together with the corresponding quantities measure for the longitudinal velocity structure functions. In order to show that the hypothesis that pressure-increments behave as a linear velocity-increment, $`\delta P\delta v`$, is definitely ruled out by our data we also show in Fig. 5 the logarithmic local slope of the ESS applied to pressure structure-functions for $`q=1,q^{}=2`$. As one can see, while the three slopes measured on the flux structure functions and on the longitudinal structure functions agree perfectly with the high-Reynolds numbers measurements, the pure-pressure structure functions is definitely much poorer. In Fig. 6 we plot the same of Fig. 5 but for a different choice of moments, $`q=2,q^{}=4`$, for both fluxes and longitudinal structure functions and with $`q=1,q^{}=2`$ for the pure pressure structure functions. In this way we are comparing quantities with exactly the same dimensional properties. Again, while the flux-made structure functions, $`S_q^{vP}(R)`$, $`S_q^{vv^2}(R)`$ and the longitudinal structure functions, $`S_q^v(R)`$ have the same local slope the pure-pressure result obtained on the ESS of $`F_q(R)`$ shows a poorer and different scaling.
A few comments are now in order. On one hand, we see from Figs. 5 and 6 that the simple local-refined Kolmogorov hypothesis derived in the previous section is correct, i.e. fluxes (11) have the same scaling properties of the usual longitudinal structure function in homogeneous and isotropic turbulence, confirming that these observable with the same physical dimensions and built in terms of local field increments scale in the same way. On the other hand, (see Fig. 6) pure-pressure structure functions seem to violate the previous statement despite of the fact that in this case there even exist an exact result (5) supporting it. Why pure-pressure correlations show this strong deviation from the straightforward dimensional estimate? One possible explanation is connected to the -possible- lack of isotropy in the statistics. Any isotropically forced DNS is affected by possible non-isotropic fluctuations induced by the discretization of the numerical grid. In the following section we have analyzed non-isotropic effects on both velocity and pressure fluctuations.
## 5 Anisotropic effects
The exact relation which connect the second order pressure structure function with a linear integral combination of fourth order velocity structure function (5) is correct only in the isotropic and homogeneous case. In order to test the degree of isotropy of our simulation we have proceeded in a systematic decomposition in terms of the irreducible representations of the SO(3) symmetry group . The SO(3) decomposition is particularly simple to apply to scalar observable, i.e. observable whit all vectorial indexes contracted, like pressure structure functions or longitudinal structure functions on the kind analyzed in this work. In these case, the SO(3) decomposition is nothing but a decomposition in spherical harmonics. For example, the longitudinal structure functions, $`S_q^v(𝐑)=<(v_i(𝐱+𝐑)v_i(𝐱))\widehat{R}_i)^q>`$, can be decomposed as:
$$S_q^v(𝐑)\underset{jm}{}S_q^{jm}(R)Y_{jm}(\widehat{𝐑})$$
(13)
where now, we have explicitly considered the possibility that the undecomposed structure functions depend on the whole vector $`𝐑`$ and not only on its magnitude as in the previous sections when isotropy was assumed. The coefficient of the decomposition, $`S_q^{jm}(R)`$ depend only on the magnitude of $`R`$ and on the two “quantum” numbers $`j,m`$ which labels the properties under rotations of the $`Y_{jm}`$ eigenfunction. Obviously, in the case of perfect isotropy we would have only one projection alive, i.e. the projection on $`Y_{00}`$. The SO(3) decomposition here summarized as been already used in some experimental and numerical data analysis to properly disentangle the anisotropic effects from the isotropic ones . In our case, the relative amplitudes of $`S_q^{00}(|R|)`$ with respect to the anisotropic fluctuations $`S_q^{jm}(|R|)`$ with $`j>0,jmj`$, gives a direct quantitative estimate of the degree of anisotropic fluctuations for any scale $`|R|`$.
In Fig. 7 we show the log-log plot of the undecomposed second order longitudinal structure functions and of its projection on the fully-isotropic eigenfunction $`Y_{00}`$. As it is possible to see, despite of the isotropic forcing used in the simulation, the finite-size effects introduced by our computational grid are quite important at large scales: the projection, $`S_2^{00}(|R|)`$, shows a definitely better scaling than the undecomposed structure function already as a functions of the real separation $`R`$, i.e. without using ESS. This dramatic effect was already observed in a similar application to the decomposition of velocity fluctuations inside a channel . Fig. 7 definitely show that the SO(3) decomposition can help in cleaning scaling properties also in “quasi-isotropic” simulations.
On the other hand, the situation is quite different when the same decomposition is applied to the pressure structure functions:
$$F_q^v(𝐑)\underset{jm}{}F_q^{jm}(R)Y_{jm}(\widehat{𝐑})$$
(14)
Let us notice that pressure is a quasi-isotropic observable also for strong anisotropic velocity configuration. Indeed, being the solutions of a Poisson problem, pressure is always an average of velocity fluctuations on all spatial directions. This simple considerations is perfectly verified on our numerical simulation. In Fig. 8 we show the undecomposed second-order pressure structure function together with its projection on the fully isotropic harmonics, $`F_q^{00}(R)`$. One can hardly detect any differences, suggesting tha anisotropic fluctuations cannot be the responsible for the poor scaling observed in the previous sections for the pure-pressure structure functions.
The only possibility to reconcile the exact result (5) with the poor scaling agreement between pressure structure functions and velocity structure functions is, in our opinion, to invoke strong Reynolds effects. Indeed, the resolution of the Poisson problem certainly introduces strong non-local effects on the statistics. Non-locality may also translate in strong long-range effects in the Fourier space as far as the importance of boundary conditions and forcing on the inertial range properties are concerned. If this is correct, there are not reason to expect good scaling properties for pure-pressure correlations, unless Reynolds number is high enough to recover also in laboratory experiments an almost-“infinite” inertial range extension.
## 6 Conclusions
We have analyzed some pressure and pressure-velocity correlations in a Direct Numerical Simulations at moderate Reynolds numbers. We have derived on the basis of a simple analysis of energy transfer properties across any sub-volume in the real space what we call a “local”-Refined Kolmogorov Hypothesis. We have identified a set of pressure-velocity correlations which should have a good scaling behaviour because connected via the local-RKH to the scaling of the energy dissipation coarse grained on inertial range scales.
We have showed that our scaling hypothesis is well verified, while pure pressure correlations feels strong Reynolds effects. According to our analysis pressure scaling is perfectly determined by the dimensional assumption that pressure behaves as a “velocity squared”, unless the finite-Reynolds effects are overwhelming. We do not find any sign which could support the fact that pressure differences behave as velocity differences as proposed in .
We have also applied the SO(3) decompositions to the pressure structure functions in order to show that poor scaling properties showed by pure-pressure structure functions are not connected to anisotropic fluctuations.
We acknowledge some help by I. Mazzitelli in the SO(3) analysis. LB and FT have been partially supported by INFM (PRA-TURBO) and by the EU contract FMRX CT98-0175, |
warning/0001/math0001108.html | ar5iv | text | # Manifolds with boundary and of bounded geometry
## 1 Introduction
Manifolds of bounded geometry arise naturally when one deals with non-compact Riemannian manifolds, and are studied extensively in the literature. So far, the focus was on manifolds without boundary.
One main source of examples are coverings of compact manifolds, which are particularly important in the context of $`L^2`$-cohomology and other $`L^2`$-invariants. These invariants are studied frequently also for manifolds with boundary. Therefore, it is natural to look at more general manifolds with boundary and bounded geometry.
There are mainly two ways to define manifolds of bounded geometry: either one uses bounds on the curvature (and its covariant derivatives) —this is the coordinate-free description— or one uses geodesic charts and bounds on the metric tensor and its derivatives in these coordinates —the coordinate approach. A proof of the equivalence of these two definitions for manifolds without boundary can be found in Eichhorn , using Jacobi fields. Related but different results are obtained in using synchronous frames. The case of manifolds with boundary causes additional technical difficulties and seems not to be covered in the literature. Therefore we give a proof here, using synchronous frames.
Dealing with manifolds with boundary, in addition to the usual requirements in the interior we must impose boundary regularity conditions. These involve the second fundamental form (in the coordinate-free description) or special charts for the boundary (in the coordinate description).
In the last section, we show that the functions given by a change of geodesic coordinates and their derivatives admit uniform bounds on manifolds of bounded geometry. And we provide one technical tool, namely a nice atlas with subordinate nice partition of unity. This was introduced and used by Shubin \[10, 1.2 and 1.3\] if the boundary is empty.
This paper grew out of part of the Dissertation of the author, and the results obtained here are used in . I thank my advisor, Prof. Wolfgang Lück, for his constant support and encouragement. I also thank the referees for valuable comments and suggestions concerning the exposition of the paper.
## 2 Coordinate-free versus coordinate-wise curvature bounds
###### 2.1 Definition.
On a Riemannian manifold $`(M^m,g)`$ with boundary $`M`$, $`R`$ denotes the curvature tensor of $`M`$, $`l`$ the second fundamental form of $`M`$, and $`\overline{R}`$ the curvature tensor of $`M`$ (with its induced metric). The (Levi-Civita)-covariant derivative of $`M`$ is denoted with $``$, the one of $`M`$ with $`\overline{}`$. We use $`\nu `$ for the unit inward normal vector field at $`M`$.
If not stated otherwise, a manifold $`M`$ will always have dimension $`m`$.
Given an open subset $`UM`$ and a chart $`x=(x_1,\mathrm{},x_m):U^m`$, we consider the corresponding derivations $`\frac{}{x_i}`$ as derivations on $`U`$, or as elements in the tangent bundle $`TM`$. We abbreviate $`_i:=\frac{}{x_i}`$. We let $`g_{ij}:=g(_i,_j)`$ be the metric tensor in the given coordinates and $`g^{ij}`$ be the coefficients of the inverse matrix.
We use the notation of multi-indices throughout: Let $`\alpha =(\alpha _1,\mathrm{},\alpha _m),\beta =(\beta _1,\mathrm{},\beta _m)`$ be multi-indices (with $`\alpha _i,\beta _i\{0\}`$). Then
$$D^\alpha :=D_x^\alpha :=\frac{^{\alpha _1}}{x_1^{\alpha _1}}\mathrm{}\frac{^{\alpha _m}}{x_m^{\alpha _m}};$$
and we set $`\beta \alpha `$ if and only if $`\beta _i\alpha _i`$ for $`i=1,\mathrm{},n`$. Define $`\left|\alpha \right|:=_{i=1}^m\alpha _i`$.
For $`VM`$ and $`r>0`$ set $`U_r(V):=\{xM|d(x,V)<r\}`$. For $`pM`$ set $`B(p,r):=U_r(\{p\})`$. If $`pM`$, $`(B(p,r)M)`$ means the corresponding set for $`M`$ with the induced Riemannian metric.
We use the normal geodesic flow $`K:M\times [0,\mathrm{})M:(x^{},t)\mathrm{exp}_x^{}^M(t\nu _x^{})`$. For $`pM`$ set $`Z(p,r_1,r_2):=K((B(p,r_1)M)\times [0,r_2))M`$.
Set $`N(s):=K(M\times [0,s])`$ if $`s0`$.
###### 2.2 Definition.
Suppose $`M`$ is a manifold with boundary $`M`$ (possibly empty). It is of (coordinate-free defined) bounded geometry if the following holds:
* Normal collar: there exists $`r_C>0`$ so that the *geodesic collar*
$$M\times [0,r_C)M:(t,x)\mathrm{exp}_x(t\nu _x)$$
is a diffeomorphism onto its image ($`\nu _x`$ is the unit inward normal vector).
* The injectivity radius $`r_{inj}(M)`$ of $`M`$ is positive.
* Injectivity radius of $`M`$: There is $`r_i>0`$ so that if $`rr_i`$ then for $`xMN(r)`$ the exponential map is a diffeomorphism on $`B(0,r)T_xM`$. Hence, if we identify $`T_xM`$ with $`^m`$ via an orthonormal frame we have *Gaussian coordinates* $`^mB(0,r)\stackrel{\mathrm{exp}_x^M}{}M`$ around every point in $`MN(r)`$.
* Curvature bounds: For every $`k`$ there is $`C_k>0`$ so that $`\left|^iR\right|C_k`$ and $`\left|\overline{}^il\right|C_k`$ for $`0ik`$.
The injectivity radius and curvature bounds are what one is used to for manifolds without boundary (compare e.g. \[4, Section 3\]). The embedding of the boundary is described by the second fundamental form. Because the injectivity radius does not make sense near the boundary, we replace it by the geodesic collar.
To give the coordinate-wise definition of bounded geometry, we have to explain which charts we want to use:
###### 2.3 Definition.
Let $`M`$ be a Riemannian manifold with boundary $`M`$. Fix $`x^{}M`$ and an orthonormal basis of $`T_x^{}M`$ to identify $`T_x^{}M`$ with $`^{m1}`$. For $`r_1,r_2>0`$ sufficiently small (such that the following map is injective) define normal collar coordinates
$$\kappa _x^{}:\underset{^{m1}}{\underset{}{B(0,r_1)}}\times [0,r_2)M:(v,t)\mathrm{exp}_{\mathrm{exp}_x^{}^M(v)}^M(t\nu ).$$
(We compose the exponential maps of $`M`$ and of $`M`$, and $`\nu `$ is the inward unit normal vector field). The tuple $`(r_1,r_2)`$ is called the *width of the normal collar chart* $`\kappa _x^{}`$.
We adopt the convention that the boundary defining coordinate is the last (i.e. $`m^{\text{th}}`$) coordinate.
For $`xMM`$ and $`r_3>0`$ sufficiently small the exponential map yields Gaussian coordinates (identifying $`T_xM`$ with $`^m`$ via an orthonormal base)
$$\kappa _x:B(0,r_3)M:v\mathrm{exp}_x^M(v).$$
We call $`r_3`$ the *radius of the Gaussian chart* $`\kappa _x`$.
We use the common name normal coordinates for normal collar coordinates as well as Gaussian coordinates.
###### 2.4 Definition.
A Riemannian manifold $`M`$ with boundary $`M`$ has (coordinate-wise defined) bounded geometry if and only if (N), (IC), (I) of Definition 2.2 hold and (instead of (B))
* There exist $`0<R_1r_{inj}(M)`$, $`0<R_2r_C`$ and $`0<R_3r_i`$ and constants $`C_K>0`$ (for each $`K`$) such that whenever we have normal boundary coordinates of width $`(r_1,r_2)`$ with $`r_1R_1`$ and $`r_2R_2`$, or Gaussian coordinates of radius $`r_3R_3`$ then in these coordinates
$$\left|D^\alpha g_{ij}\right|C_K\text{and}\left|D^\alpha g^{ij}\right|C_K\left|\alpha \right|K.$$
The numbers $`R_1`$, $`R_2`$, $`R_3`$ and $`C_K`$ are called the bounded geometry constants of $`M`$.
The main result of the paper is the following:
###### 2.5 Theorem.
Let $`(M^m,g)`$ be a Riemannian manifold with boundary $`M`$. To given $`C>0`$, $`k`$, and dimension $`m`$ there are $`R_1,R_2,R_3>0`$ and $`D>0`$ such that the following holds:
* If $`xMM`$, $`0<r_3R_3`$ and $`\kappa _x:B(0,r_3)(MM)`$ is a Gaussian chart, and if $`\left|^iR\right|C`$ for $`i=0,\mathrm{},k`$ on the image of $`\kappa _x`$ then in these coordinates
$$\left|D^\alpha g_{ij}\right|D\text{and}\left|D^\alpha g^{ij}\right|D\text{whenever }\left|\alpha \right|k.$$
* If on the other hand
$$\left|D^\alpha g_{ij}\right|C\text{and}\left|D^\alpha g^{ij}\right|C\text{for }\left|\alpha \right|k+2$$
then on the image of $`\kappa _x`$ we have
$$\left|^iR\right|D\text{for }i=0,\mathrm{},k.$$
* If $`x^{}M`$, $`0<r_1R_1`$, $`0<r_2R_2`$ and $`\kappa _x^{}:B(0,r_1)\times [0,r_2)M`$ is a normal boundary chart, and if $`\left|^iR\right|C`$ and $`\left|\overline{}^il\right|C`$ for $`i=0,\mathrm{},k`$ on the image of $`\kappa _x^{}`$, then in these coordinates we get
$$\left|D^\alpha g_{ij}\right|D\text{and}\left|D^\alpha g^{ij}\right|D\text{whenever }\left|\alpha \right|k.$$
* If, on the other hand,
$$\left|D^\alpha g_{ij}\right|C\text{and}\left|D^\alpha g^{ij}\right|C\text{for }\left|\alpha \right|k+2$$
then, on the image of $`\kappa _x^{}`$,
$$\left|^iR\right|D\text{ and }\left|\overline{}^il\right|D\text{for }i=0,\mathrm{},k.$$
* $`M`$ has (coordinate-wise defined) bounded geometry if and only if it has (coordinate-free defined) bounded geometry. In particular, we can drop the prefix in notation. The bounded geometry constants of Definition 2.4 can be chosen to depend only on $`r_i`$, $`r_C`$, $`r_{inj}(M)`$ and $`C_k`$ of Definition 2.2.
Observe that (c) follows from (a1)-(b2). Moreover, (a2) and (b2) are immediate consequences of the formulas for $`R`$ and $`l`$ (and their covariant derivatives) in local coordinates in terms of $`g_{ij}`$, $`g^{ij}`$ and their partial derivatives (compare 2.54, 3.16 and 5.1 of — note that our charts near the boundary are adapted to the embedding $`MM`$). The statement (a1) about internal points is already included in \[5, Theorem A and Proposition 2.3\]. It remains to establish (b1). Since in the course of this proof we have to set up most of the notation necessary for the synchronous-frame-proof of (a1), we include a complete proof also of (a1).
The proof is done in four steps. First, we give the argument for $`k=0`$, using the Rauch comparison theorem. Secondly, we prove (a1). In the third step, we establish bounds on the curvature tensor of the boundary. Last, we derive (b1).
Step 1: Proof of Theorem 2.5(a1) and 2.5(b1) for $`k=0`$
###### 2.6 Proposition.
Suppose we are in the situation of Theorem 2.5(a1) or (b1) and $`k=0`$. Suppose $`(x_i):=\kappa ^1:UM^m`$ is the normal coordinate system. There are $`R_1,R_2,R_3>0`$ and $`C_1,C_2>0`$ (depending only on $`C`$ and $`m`$) such that if for width or radius we have $`(r_1,r_2)(R_1,R_2)`$ or $`r_3R_3`$, respectively, then
$$C_1\left|\lambda _i\frac{}{x_i}\right|_{TM}C_2,\text{if }\lambda _i^2=1,$$
(2.7)
where $`\left|v\right|_{TM}:=\sqrt{g(v,v)}`$ for $`vTM`$.
Moreover, $`g_{ij}`$ and $`g^{ij}`$ are bounded with a bound depending only on $`C`$ and the dimension.
The numbers $`R_1`$, $`R_2`$ and $`R_3`$ of Theorem 2.5 are determined by Proposition 2.6 and (IC), (I), (N).
###### Proof.
The last statement is a reformulation of Inequality (2.7). To prove (2.7), we apply Warner’s generalization of the Rauch comparison theorem \[11, 4.3\]. We compare with two complete manifolds of constant sectional curvature $`C`$ and $`C`$, respectively. To compare with normal collar coordinates, choose a hypersurface in this manifold so that all the eigenvalues of its second fundamental form at one (comparison) point are equal to $`C`$ in the first case and to $`C`$ in the second case. Inequality (2.7) for vectors orthogonal to $`:=x_i_i`$ (in Gaussian coordinates), or orthogonal to $`_m`$ (in normal boundary coordinates) is just the statement of the comparison theorem, with $`C_1`$ and $`C_2`$ depending only on the manifold we compare with (i.e. on $`C`$ and on $`m`$). Here, $`r_1`$, $`r_2`$ and $`r_3`$ must be sufficiently small (again depending only on the manifolds we compare with). The comparison theorem says nothing about $``$ or about $`_m`$, respectively. But for these vectors Euclidean length and length in $`TM`$ as well as the orthogonal complements coincide by the following Proposition 2.8. Therefore, the inequality is true in general. ∎
In the proof of Proposition 2.6 we used the Gauss lemma:
###### 2.8 Proposition.
Let $`(M,g)`$ be a Riemannian manifold and $`\mathrm{exp}:B(0,R)M`$ a Gaussian chart. Pull the metric $`g`$ back to $`B(0,R)`$. Then $`g(,)=r^2`$, ($`=_ix_i_i`$), and $`g(,v)=0`$ if and only if $`v`$ is a tangent vector to a sphere with center the origin $`0`$.
Let $`K:M\times [0,r_C)M`$ be the geodesic collar and pull $`g`$ back to $`M\times [0,r_C)`$. Then $`g(_m,_m)=1`$ and $`g(_m,v)=0`$ if and only if $`v`$ is tangent to a translate $`M\times \{t\}`$.
###### Proof.
Compare \[6, 2.93\] — the proof there works also for the collar. ∎
Step 2: Proof of 2.5(a1).
Suppose we are in the situation of 2.5(a1) with $`pMM`$ and Gaussian coordinates $`x=(x_1,\mathrm{},x_m)=\kappa _p^1:B(p,r_3)^m`$. We will state a (differential) equation for $`g_{ij}`$ in terms of the curvature tensor, so that a bound on partial derivatives of the components of the curvature tensor will give corresponding bounds for the metric. Partial and covariant derivative are related by the Christoffel symbols, so we will compute them, too.
Choose an orthonormal base $`\{s_i\}`$ for $`T_pM`$. Using parallel transport along geodesics emanating from $`p`$, construct a synchronous orthonormal frame $`\{s_i(x)\}`$ of the tangent space restricted to $`B(p,r_3)`$. Let $`\{\theta ^i\}`$ be the frame of 1-forms dual to $`\{s_i\}`$ (therefore orthonormal). The connection forms $`\theta _j^i`$ for this frame are defined by
$$s_j=\underset{i}{}\theta _j^is_i,$$
with associated Christoffel symbols $`\mathrm{\Gamma }_{jk}^i`$ and curvature tensor $`R_{jkl}^i`$ given by
$$\theta _j^i=\underset{k}{}\mathrm{\Gamma }_{jk}^idx_k;d\theta _j^i\underset{k}{}\theta _k^i\theta _j^k=\underset{k,l}{}R_{jkl}^idx_kdx_l.$$
We can express the curvature entirely in terms of $`s_i`$ and $`\theta ^i`$, which defines $`K_{jkl}^i`$:
$$d\theta _j^i\underset{k}{}\theta _k^i\theta _j^k=\underset{k,l}{}K_{jkl}^i\theta ^k\theta ^l.$$
Define functions $`a_j^i`$ and $`b_j^i`$ via the equations
$$\theta ^i=\underset{j}{}a_j^idx_j;dx_i=\underset{j}{}b_j^i\theta ^j.$$
(2.9)
$$\begin{array}{cc}\hfill \text{Then}R_{jkl}^i& =\underset{\alpha ,\beta }{}K_{j\alpha \beta }^ia_k^\alpha a_l^\beta \text{and}g_{ij}=\underset{\alpha }{}a_i^\alpha a_j^\alpha ;g^{ij}=\underset{\alpha }{}b_\alpha ^ib_\alpha ^j.\hfill \end{array}$$
(2.10)
As matrix, $`(g_{ij})`$ is the product of $`(a_j^i)`$ and its adjoint, and accordingly for $`(g^{ij})`$ and $`(b_j^i)`$. Hence
###### 2.11 Lemma.
There are bounds on $`a_j^i`$ and $`b_j^i`$ corresponding to the bounds on $`g_{ij}`$ and $`g^{ij}`$ given by Proposition 2.6.
The Christoffel symbols $`\stackrel{~}{\mathrm{\Gamma }}_{jk}^i`$ of the covariant differentials of $`_i`$ are given by
$$__k_j=\underset{i}{}\stackrel{~}{\mathrm{\Gamma }}_{jk}^is_i.$$
Dualizing (2.9) we see that $`_j=_\alpha a_j^\alpha s_\alpha `$, hence
$$\stackrel{~}{\mathrm{\Gamma }}_{jk}^i=_ka_j^i+\underset{\alpha }{}a_j^\alpha \mathrm{\Gamma }_{\alpha k}^i.$$
(2.12)
Atiyah, Bott, and Patodi \[2, a6 and a10\] derive the following equations (note that our definition of $`R_{jkl}^i`$ takes care of the problems described in ), where $`=_ix_i_i`$:
$$\mathrm{\Gamma }_{jk}^i+\mathrm{\Gamma }_{jk}^i=\underset{l}{}2x_lR_{jkl}^ii,j,k;$$
(2.13)
$$(^2+)a_l^i=2\underset{j,k}{}R_{jkl}^ix_jx_ki,l.$$
(2.14)
Set $`f_x(t):=t\mathrm{\Gamma }_{jk}^i(tx)`$. Let denote differentiation with respect to $`t`$. Then
$$f_x^{}(t)=\mathrm{\Gamma }_{jk}^i(tx)+t\underset{l}{}x_l_l\mathrm{\Gamma }_{jk}^i(tx)\stackrel{(\text{2.13})}{=}\underset{l}{}2tx_lR_{jkl}^i(tx).$$
$$\mathrm{\Gamma }_{jk}^i(x)=f_x(1)=_0^1\underset{l}{}2\tau x_lR_{jkl}^i(\tau x)d\tau \text{and}$$
$$D_x^\alpha \mathrm{\Gamma }_{jk}^i(x)=_0^1\tau ^{\left|\alpha \right|}\left(D_x^\alpha (x\underset{l}{}x_lR_{jkl}^i(x))\right)(\tau x)𝑑\tau .$$
(2.15)
Set $`f_{il}(t,x):=a_l^i(tx)`$. Then $`tf_{ij}^{}{}_{}{}^{}(t,x)=a_j^i(tx)`$ and $`t^2f_{il}^{\prime \prime }(t,x)+tf_{il}^{}(t,x)=^2a_l^i(tx)`$. By (2.14)
$$t^2f_{il}^{\prime \prime }+2tf_{il}^{}=2t^2\underset{k,j}{}R_{jkl}^i(tx)x_jx_k.$$
With $`w_{il}(t,x):=t^2f_{il}^{}(t,x)`$ we get $`w_{il}^{}=t^2f_{il}^{\prime \prime }+2tf_{il}^{}`$. Since $`w_{il}(0)=0`$,
$$t^2f_{il}^{}(t,x)=2_0^t\tau ^2\underset{j,k}{}R_{jkl}^i(\tau x)x_jx_kd\tau \stackrel{\tau =tu}{}$$
$$f_{il}^{}(t,x)=2t_0^1u^2\underset{j,k,\alpha ,\beta }{}K_{j\alpha \beta }^i(tux)a_k^\alpha (tux)a_l^\beta (tux)x_jx_kdu.$$
(2.16)
Now we are in the position to explain how the bounds on $`R`$ and its covariant derivatives up to order $`k`$ give rise to bounds on $`g_{ij}`$, $`g^{ij}`$ and their partial derivatives up to order $`k`$. Because of (2.10) we can consider $`a_j^i`$ and $`b_j^i`$ instead of the metric tensor. Moreover, the case $`k=0`$ is done by Proposition 2.6.
###### 2.17 Lemma.
Let $`A`$, $`B`$ be matrix valued functions which are inverse to each other. Then
$$\frac{}{x_i}B=\frac{}{x_i}(A^1)=A^1(\frac{}{x_i}A)A^1=B(\frac{}{x_i}A)B.$$
Iterated application of this and of the product rule yields
$$D_x^\alpha B=P_\alpha (B,D_x^\beta A;\beta \alpha ),$$
where $`P_\alpha `$ is a fixed polynomial in non-commuting variables. Bounds for the partial derivatives of $`A`$ up to order $`k`$ and on $`B`$ yield bounds for the partial derivatives of $`B`$.
Lemma 2.17 applies to the matrices $`A=(a_j^i)`$ and $`B=(b_j^i)`$. Moreover, by Proposition 2.6, we have a bound for $`(b_{ij})`$. Hence it remains to find bounds for the derivatives of $`(a_j^i)`$.
###### 2.18 Lemma.
For $`\alpha =(\alpha _1,\mathrm{},\alpha _n)`$ there is a polynomial $`P_{\alpha ,ijkl}`$ (only depending on $`\alpha ,i,j,k,l`$) in partial derivatives up to order $`(\left|\alpha \right|1)`$ of $`K_{}^{}`$, $`\mathrm{\Gamma }_{}^{}`$, and $`a_{}^{}`$ such that as functions on the set $`B(p,r_3)`$
$$(__1)^{\alpha _1}\mathrm{}(__n)^{\alpha _n}R(s_i,s_j,_k,_l)=D_x^\alpha K_{jkl}^i+P_{\alpha ,ijkl}.$$
###### Proof.
This follows from the formula for covariant differentials in coordinates. Note that for $`\left|\alpha \right|=1`$ only $`\mathrm{\Gamma }_{jk}^i`$ shows up (since $`K_{jkl}^i`$ is defined entirely in terms of $`s_i`$). But if we iterate the covariant differentials, we have to take into account that we contracted $`R`$ with $`_i`$ and not with $`s_i`$. This yields (via $`_i`$) $`\stackrel{~}{\mathrm{\Gamma }}_{jk}^i`$ and, since we iterate the covariant differentials, their partial derivatives up to order $`\left|\alpha \right|2`$. Since $`\stackrel{~}{\mathrm{\Gamma }}_{jk}^i=_ka_j^i+_\alpha a_j^\alpha \mathrm{\Gamma }_{\alpha k}^i`$, the result follows. ∎
Now we proceed by induction on the order of derivatives $`\left|\alpha \right|`$. For $`\left|\alpha \right|=0`$ observe that by assumption we have a bound on the curvature. Since $`\{s_i\}`$ is orthonormal this gives bounds on $`K_{jkl}^i`$. By Proposition 2.6 the same is true for $`a_j^i`$.
Assume by induction that for $`r0`$ we have found bounds on the partial derivatives up to order $`r`$ of $`K_{jkl}^i`$ and $`a_j^i`$ and on the derivatives up to order $`(r1)`$ of $`\mathrm{\Gamma }_{jk}^i`$. The assumptions of the Theorem give bounds on $`\left|R\right|,\mathrm{},\left|^{r+1}R\right|`$.
From equation (2.10), relating $`K_{jkl}^i`$ and $`R_{jkl}^i`$, we get bounds on the partial derivatives up to order $`r`$ of $`R_{jkl}^i`$. Then Equation (2.15) yields bounds for the derivatives of order $`r`$ of $`\mathrm{\Gamma }_{jk}^i`$. Lemma 2.18 and the bound on $`^{r+1}R`$ yield bounds on $`(r+1)`$-order partial derivatives of $`K_{jkl}^i`$ (since by Proposition 2.6 the length of $`_i`$ is controlled). In all instances the new bounds are given in terms of the old ones.
It remains to deal with the derivatives of order $`(r+1)`$ of $`a_j^i`$. Remember Equation (2.16) for $`f_{il}(t,x)=a_l^i(tx)`$:
$$f_{il}^{}(t,x)=2t_0^1u^2\underset{j,k,\alpha ,\beta }{}\left(K_{j\alpha \beta }^ia_k^\alpha a_l^\beta \right)(tux)x_jx_kdu.$$
Let $`\alpha `$ be a multi-index with $`\left|\alpha \right|=r+1`$. We differentiate the equation with respect to $`x`$ to get an equation for $`D_x^\alpha f_{il}(t,x)=t^{\left|\alpha \right|}(D^\alpha a_l^i)(tx)`$. This yields
$$\begin{array}{c}(D_x^\alpha f_{il})^{}(t,x)=2t_0^1(u^2\underset{j,k,\beta ,\gamma }{}K_{j\beta \gamma }^i(tux)\hfill \\ \hfill ((D_x^\alpha f_{\beta k})f_{\gamma l}++(D^\alpha f_{\gamma l})f_{\beta k})(tu,x)x_jx_kdu)2t_0^1P_\alpha du.\end{array}$$
(2.19)
Here $`P_\alpha `$ is a polynomial in $`t`$, $`u`$, $`x`$, partial derivatives up to order $`(r+1)`$ of $`K_{}^{}`$ at $`tux`$, and partial derivatives up to order $`r`$ of $`f_{}`$ at $`(tu,x)`$. The left and right hand side of (2.19) are equal as function of $`x`$ and $`t`$. The induction hypothesis implies for $`0t1`$ with suitable $`C_1,C_2>0`$ the inequality
$$\left|(D_x^\alpha f_{ij})^{}(t,x)\right|C_1\underset{0\tau t}{sup}\{\left|D_x^\alpha f_{ij}(\tau ,x)\right|\}+C_2,$$
(2.20)
Moreover, $`D^\alpha f(0,x)=0`$ since $`\left|\alpha \right|1`$.
Let $`h(t):=C_2(\mathrm{exp}(C_1t)1)/C_1`$ be the unique solution of $`h^{}(t)=C_1h(t)+C_2`$ with $`h(0)=0`$. This is a positive monotonous increasing function, with an explicit bound $`h(t)C:=C_2(\mathrm{exp}(C_1)1)/C_1`$ for $`0t1`$.
Abbreviate $`u_j^i(t):=D_x^\alpha f_{ij}(t,x)`$. We will prove $`\left|u_j^i(t)\right|h(t)`$ and therefore
$$\left|D^\alpha a_j^i(x)\right|=\left|u_j^i(1)\right|h(1)C.$$
(2.21)
This then finishes the induction step. To show $`\left|u_j^i(t)\right|h(t)`$, let $`h_n`$ be the unique solution of
$$h_n^{}(t)=C_1h_n(t)+C_2+1/n\text{with }h_n(0)=0.$$
Then $`h_n(t)\stackrel{n\mathrm{}}{}h(t)`$ uniformly for $`0t1`$. Therefore, it suffices to show $`\left|u_j^i(t)\right|h_n(t)`$. For a contradiction, assume $`\left|u_j^i(t)\right|>h_n(t)`$ for some $`n`$ and $`t`$. Set $`t_0:=inf_{0t}\{\left|u_j^i(t)\right|>h_n(t)\}`$. Then $`\left|u_j^i(t_0)\right|=h(t_0)`$, since $`u_i^j(0)=0=h(0)`$, $`h_n`$ is monotonous, and $`\left|u_j^i(t)\right|h_n(t)`$ $`tt_0`$. Consequently, $`sup_{tt_0}\left|u_j^i(t)\right|=\left|u_j^i(t_0)\right|`$. Then (2.20) shows
$$\left|(u_j^i)^{}(t_0)\right|C_1\left|u_j^i(t_0)\right|+C_2<h_n^{}(t_0).$$
Moreover, $`d/dt\left|u_j^i(t_0)\right|\left|(u_j^i)^{}(t_0)\right|`$ (compare \[7, III.3.2\] for the difficult case $`u_j^i(t_0)=0`$$`d/dt`$ is understood to be the right derivative). It follows $`\left|u_j^i(t)\right|<h_n(t)`$ for $`t[t_0,t_0+ϵ)`$ and $`ϵ>0`$ sufficiently small. But this contradicts the choice of $`t_0`$.
Step 3: Curvature of $`M`$.
We adopt the notation of Definition 2.1.
In the following we consider $`(0,p)`$-tensors $`T`$ on $`M`$ and their restriction to $`M`$, given by the inclusion $`TMTM`$. We will use the same notation for $`T`$ and its restriction, the meaning will be clear from the context.
We compute the covariant derivatives $`\overline{}^k\overline{R}`$ using the following rules:
###### 2.22 Lemma.
Suppose $`T`$ is a $`(0,q)`$-tensor on $`M`$, $`S`$ a $`(0,p)`$-tensor on $`M`$, and $`S^_1`$ the $`(1,p1)`$-tensor on $`M`$ given by $`g(S_x^_1(v_2,\mathrm{},v_p),v_1)=S_x(v_1,\mathrm{},v_p)`$ for $`v_1,\mathrm{},v_pT_xM`$, where $`xM`$ and $`S_x`$, $`S_x^_1`$ are the values of $`S`$ and $`S^_1`$, respectively, at $`x`$. Let $`\sigma `$ be a permutation (operating on a multiple tensor product by permutation of the factors) with $`\sigma ^1(1)p`$.
Let $`c`$ denote the contraction of a $`(0,r)`$-tensor with a $`(1,s)`$-tensor which contracts the $`r`$-th entry of the $`(0,r)`$-tensor. The covariant derivative is understood to be a map $`:C^{\mathrm{}}(E)C^{\mathrm{}}(ET^{}M)`$. Then the following holds:
1. $`\overline{}T=T_ic(Tl\sigma _i,\nu )`$, where $`\sigma _i`$ are appropriate permutations.
2. $`\overline{}((TS)\sigma )=((\overline{}T)S)\sigma +(T\overline{}S)\sigma `$.
3. $`\overline{}c((TS)\sigma ,\nu )=c((T)S\sigma ^{},\nu )+c(T\overline{}S\sigma ^{\prime \prime },\nu )+_ic(c(TS\sigma ,\nu )l\sigma _i,\nu )+c(TS\sigma ,l^_1)`$, with $`\sigma ^{}`$, $`\sigma ^{\prime \prime }`$, and $`\sigma _i`$ appropriate permutations.
4. $`\overline{}c(T,(\overline{}^kl)^_1)=c(\overline{}T,(\overline{}^kl)^_1)+c(T,(\overline{}^{k+1}l)^_1)`$.
###### Proof.
Formulas 2. and 4. are well known. Let $`v_1,\mathrm{},v_p`$ and $`X`$ be vector fields on $`M`$ For 1. we compute:
$$\begin{array}{cc}& \overline{}T(v_1,\mathrm{},v_p,X)\hfill \\ & =X.T(v_1,\mathrm{},v_p)T(\overline{}_Xv_1,\mathrm{},v_p)\mathrm{}T(v_1,\mathrm{},\overline{}_Xv_p)\hfill \\ & \stackrel{\overline{}_XY=_XYl(X,Y)\nu }{=}X.T(v_1,\mathrm{},v_p)T(_Xv_1,\mathrm{})\mathrm{}\hfill \\ & +T(\nu ,v_2,\mathrm{})l(v_1,X)+\mathrm{}+T(v_1,\mathrm{},v_{p1},\nu )l(v_p,X)\hfill \\ & =T(v_1,\mathrm{},v_p,X)+\underset{i}{}c(Tl\sigma _i,\nu )(v_1,\mathrm{},v_p,X).\hfill \end{array}$$
For 3. set $`v_1:=\nu `$ and calculate:
$$\begin{array}{cc}& \overline{}_Xc(TS\sigma ,\nu )(v_2,\mathrm{},v_{p+q})\stackrel{v_1=\nu }{=}\hfill \\ & =(X.T(v_{\sigma 1},\mathrm{},v_{\sigma p}))S(v_{\sigma (p+1)},\mathrm{},v_{\sigma (p+q)})\hfill \\ & +T(v_{\sigma 1},\mathrm{},v_{\sigma p})(X.S(v_{\sigma (p+1)},\mathrm{},v_{\sigma (p+q)}))\hfill \\ & \underset{\begin{array}{c}i=1\\ \sigma i1\end{array}}{\overset{p+q}{}}TS(v_{\sigma 1},\mathrm{},\overline{}_Xv_{\sigma i},\mathrm{},v_{\sigma (p+q)})\hfill \\ & =c(T\overline{}_XS\sigma ,\nu )(v_2,\mathrm{},v_{p+q})\hfill \\ & +(X.T(v_{\sigma 1},\mathrm{},v_{\sigma p})\underset{i=1}{\overset{p}{}}T(v_{\sigma 1},\mathrm{},_Xv_{\sigma i},\mathrm{},v_{\sigma p}))S(\mathrm{})\hfill \\ & \underset{\begin{array}{c}i=1\\ \sigma (i)1\end{array}}{\overset{p}{}}\underset{=T(\mathrm{})l(X,v_{\sigma i})}{\underset{}{T(v_{\sigma 1},\mathrm{},l(X,v_{\sigma i})\nu ,\mathrm{},v_{\sigma p})}}S(\mathrm{})\hfill \\ & +T(v_{\sigma 1},\mathrm{},_X\nu ,\mathrm{},v_{\sigma p})S(\mathrm{})\hfill \\ & =c(T\overline{}_XS\sigma ,\nu )(\mathrm{})+c((_XT)S\sigma ,\nu )(\mathrm{})\hfill \\ & \underset{i}{}c(c(TS\sigma ,\nu )l\sigma _i,\nu )(\mathrm{},X)+c(TS\sigma ,_X\nu )(\mathrm{}).\hfill \end{array}$$
If $`YC^{\mathrm{}}(TM)`$ then
$$\begin{array}{cc}& 0=X.g(\nu ,Y)=g(_X\nu ,Y)+g(\nu ,_XY)\hfill \\ & g(_X\nu ,Y)=l(X,Y)=l(Y,X)\hfill \\ & 0=X.g(\nu ,\nu )=2g(_X\nu ,\nu ).\hfill \\ \hfill & _X\nu =l^_1(X)\nu =l^_1.\hfill \end{array}$$
This finishes the proof. ∎
###### 2.23 Corollary.
$`\overline{}^k\overline{R}`$ is a finite sum of tensor products and possibly iterated contractions, composed with permutations, involving (i) $`^jR`$ for $`jk`$; (ii) $`\overline{}^jl`$ for $`j<k`$; (iii) $`(\overline{}^jl)^_1`$ for $`j<k1`$; and (iv) $`\nu `$.
Bounds for the building blocks (i) and (ii) yield a bound for $`\overline{}^k\overline{R}`$.
###### Proof.
The first statement follows by iterated application of Lemma 2.22. The last statement follows since tensor products and contractions of tensors are bounded in terms of the bounds on the factors, and because permutations are isometric. Note that $`\left|\nu \right|=1`$ and $`\left|S^_1\right|=\left|S\right|`$ for an arbitrary tensor $`S`$. Moreover, restriction to the boundary only decreases the norm of a tensor. ∎
###### 2.24 Corollary.
If $`M`$ is a Riemannian manifold of (coordinate-free defined) bounded geometry, the same is true for its boundary.
Step 4: Proof of Theorem 2.5(b1).
Suppose we are in the situation of 2.5(b1) with $`pM`$ and normal collar coordinates $`(x_1,\mathrm{},x_m)=\kappa _p^1:U^m`$ around $`p`$. By our convention $`x_m`$ is the boundary defining coordinate, i.e. $`_m|_M=\nu `$.
First consider $`M`$ as a Riemannian $`(m1)`$-dimensional manifold of its own. Corollary 2.23 shows that bounds on the covariant derivatives $`^jR`$ and $`\overline{}^jl`$ give rise to bounds on $`\overline{}^j\overline{R}`$ ($`0jk`$). As in Step 2 (applied to $`M`$) construct the orthonormal frame $`\{s_i\}_{1im1}`$ of $`TM`$. Extend this to an orthonormal frame of $`TM|_M`$ by setting $`s_m:=\nu `$. By parallel transport along geodesics with initial speed $`\nu `$ we get a synchronous orthonormal frame of $`TM`$ on the normal collar neighborhood. Define the dual frame $`\{\theta ^i\}`$, the Christoffel symbols $`\mathrm{\Gamma }_{jk}^i`$ and $`\stackrel{~}{\mathrm{\Gamma }}_{jk}^i`$, the curvature coefficients $`R_{jkl}^i`$ and $`K_{jkl}^i`$, and $`a_j^i`$ and $`b_j^i`$ in exactly the same way as in Step 2. Note that (2.10) and (2.12) remain true.
Now we come to the differential equations which relate these quantities. By construction, $`\{s_i\}`$ is parallel to $`_m`$. This translates to
$$c(_m)\theta _j^i=0,\text{i.e.}\mathrm{\Gamma }_{jm}^i=0i,j$$
(2.25)
($`c(_m)`$ denotes contraction with $`_m`$). The Lie derivative along $`_m`$ (denoted by $`_m`$) acts on differential forms via $`_m=c(_m)d+dc(_m)`$. Hence
$$\begin{array}{cc}\hfill _m\theta _j^i& =c(_m)d\theta _j^i\stackrel{(\text{2.25})}{=}c(_m)(d\theta _j^i\underset{k=1}{\overset{m}{}}\theta _k^i\theta _j^k)\hfill \\ & =c(_m)(\underset{k,l}{}R_{jkl}^i)dx_kdx_l=2\underset{k}{}R_{jmk}^idx_k.\hfill \end{array}$$
On the other hand, $`_m\theta _j^i=_k_m(\mathrm{\Gamma }_{jk}^i)dx_k`$. Hence, applying $`D^\alpha `$ yields
$$_m(D^\alpha \mathrm{\Gamma }_{jk}^i)=2D^\alpha R_{jmk}^ii,j,k.$$
(2.26)
Additionally, we need an equation for $`a_j^i`$. We apply $`_m`$ twice to the dual frame. Since $`_m=s_m`$ we have $`c(_m)\theta ^i=\delta _{im}`$ ($`\delta `$ the Kronecker symbol). Then
$$_m\theta ^i=c(_m)d\theta ^i+dc(_m)\theta ^i=c(_m)d\theta ^i.$$
The connection is torsion free. This means
$$\begin{array}{cc}\hfill d\theta ^i& =\underset{j}{}\theta _j^i\theta ^j\hfill \\ \hfill & _m(\theta ^i)=\underset{j}{}c(_m)(\theta _j^i\theta ^j)\stackrel{(\text{2.25})}{=}\theta _m^i\hfill \\ \hfill & _m^2(\theta ^i)=_m(\theta _m^i)=2\underset{k}{}R_{mmk}^idx_k.\hfill \end{array}$$
(2.27)
The left hand side can be computed in terms of $`a_j^i`$
$$_m(\theta ^i)=\underset{j}{}_m(a_j^i)dx_j_m^2(\theta ^i)=\underset{j}{}_m^2(a_j^i)dx_j.$$
(2.28)
Equating coefficients, applying $`D^\alpha `$, and expressing $`R_{jkl}^i`$ in terms of $`K_{jkl}^i`$ yields
$$_m^2D^\alpha a_j^i=2\underset{k,l}{}D^\alpha (K_{mkl}^ia_m^ka_j^l)i,j.$$
(2.29)
For $`\left|\alpha \right|>0`$ this is (for each point in the boundary) a system of inhomogeneous linear ordinary differential equations for $`D^\alpha a_j^i`$, with coefficients given by partial derivatives of $`K_{mkl}^i`$ up to order $`\left|\alpha \right|`$ and of $`a_j^i`$ up to order $`\left|\alpha \right|1`$.
To make use of the differential equation (2.26) and (2.29) we have to determine the initial values (at $`x_m=0`$).
If $`i,j<m`$, $`a_j^i|_{x_m=0}`$ is given by application of Step 2 to $`M`$, which is possible because of Corollary 2.23. In particular, we get bounds for these functions. And by construction $`a_m^i=a_i^m=\delta _{im}`$. For the first derivative we have $`_ma_j^i=\mathrm{\Gamma }_{mj}^i`$ (this follows from (2.27) and (2.28)).
Next we compute $`\mathrm{\Gamma }_{jk}^i`$ on $`M`$. By definition $`\overline{}__is_j=__is_jl(_i,s_j)\nu `$ for $`i,j<m`$. If we define $`l_{ij}:=l(_i,s_j)`$ then for $`j,k<m`$
$$\mathrm{\Gamma }_{jk}^i|_{x_m=0}=\{\begin{array}{cc}\overline{\mathrm{\Gamma }}_{jk}^i|_{x_m=0};\hfill & i<m\hfill \\ l_{kj};\hfill & i=m.\hfill \end{array}$$
By (2.25) $`\mathrm{\Gamma }_{jm}^i=0`$ $`i,j`$. To compute $`\mathrm{\Gamma }_{mj}^i`$ we use for $`j<m`$
$$g(__j_m,s_i)=_jg(_m,s_i)g(_m,__js_i)\stackrel{_m|_M=\nu }{=}l(_j,s_i)\text{for }i<m$$
$$2g(__j_m,_m)=_jg(_m,_m)=0$$
$$__j_m=\underset{i}{}l_{ji}s_i\text{on }M.$$
It follows for $`i<m`$
$$\mathrm{\Gamma }_{mj}^i|_{x_m=0}=\{\begin{array}{cc}l_{ji};\hfill & j<m\hfill \\ 0;\hfill & j=m.\hfill \end{array}$$
The arguments given in the proof of Step 2 show that if bounds exist on the covariant derivatives up to order $`k`$ of the second fundamental form and of $`R`$ (hence by Corollary 2.23 also on $`\overline{R}`$) then the initial values of (2.26) and (2.29), namely $`D^\alpha \mathrm{\Gamma }_{jk}^i|_{x_m=0}`$, $`D^\alpha a_j^i|_{x_m=0}`$ and $`_mD^\alpha a_j^i|_{x_m=0}`$ are bounded, as long as $`\alpha _m=0`$ (if $`\alpha =(\alpha _1,\mathrm{},\alpha _m)`$). Later, we will by induction on $`\left|\alpha \right|`$ get bounds on the right had sides of (2.26) and (2.29) (using the bounds on the initial values), giving in particular bounds on $`_mD^\alpha \mathrm{\Gamma }_{jk}^i|_{x_m=0}`$ and $`_m^2D^\alpha a_j^i|_{x_m=0}`$ which are the initial values of the equation for $`D^\beta \mathrm{\Gamma }_{jk}^i`$ and $`D^\beta a_j^i`$ where $`\beta =(\alpha _1,\mathrm{},\alpha _m+1)`$. We will therefore, inductively, get the required bounds
$$\left|D^\gamma \mathrm{\Gamma }_{jk}^i|_{x_m=0}\right|C,\left|D^\gamma a_j^i|_{x_m=0}\right|C,\left|_mD^\gamma a_j^i|_{x_m=0}\right|C.$$
(2.30)
We proceed with a bootstrap argument similar to the one in Step 2. We have to find bounds on $`g_{ij}`$, $`g^{ij}`$ and their derivatives. Because of (2.9) it suffices to look at $`a_j^i`$ and $`b_j^i`$ and their derivatives. Bounds on $`a_j^i`$ and $`b_j^i`$ are given by Lemma 2.11. As in the proof of Step 2, because of Lemma 2.17 we only have to control the derivatives of $`a_j^i`$. We do this by induction on the order of these derivatives. To carry out the induction step, we also have to control the derivatives of $`K_{jkl}^i`$, $`R_{jkl}^i`$ and $`\mathrm{\Gamma }_{jk}^i`$ (and the initial values in (2.30)).
To conclude the start of the induction, Lemma 2.18 and the assumptions give bounds on $`K_{jkl}^i`$ and (since the length of $`_i`$ is bounded by Proposition 2.6) on $`R_{jkl}^i`$. Integrating Equation (2.26), we find bounds for $`\mathrm{\Gamma }_{jk}^i`$ (depending also on the given width $`R_1`$ of the normal boundary charts).
Assume now by induction that we have bounds on $`D^\alpha a_{ij}`$, $`D^\alpha \mathrm{\Gamma }_{ij}^k`$, $`D^\alpha K_{jkl}^i`$, and $`D^\alpha R_{jkl}^i`$ for $`\left|\alpha \right|r`$. By Lemma 2.18, the assumed bound on $`^{r+1}R`$ therefore gives bounds on $`^\gamma K_{jkl}^i`$ ($`\left|\gamma \right|=r+1`$). Because of these and the bounds on $`^\alpha K_{jkl}^i`$, $`^\alpha a_j^i`$ ($`\left|\alpha \right|r`$) and on the initial values (2.30) we can apply \[7, IV.4.2\] to (2.29) to obtain bounds on $`^\gamma a_j^i`$ ($`\left|\gamma \right|=r+1`$). This in turn, together with the relation (2.10) between $`R_{jkl}^i`$ and $`K_{jkl}^i`$ yields bounds on $`^\gamma R_{jkl}^i`$ ($`\left|\gamma \right|=r+1`$).
Now we integrate (2.26) to get bounds on $`^\gamma \mathrm{\Gamma }_{jk}^i`$ ($`\left|\gamma \right|=r+1`$) to finish the induction step and to conclude the proof of Theorem 2.5.
The bounds we obtain, inductively, depend only on the bounds we started with.
## 3 Technical properties of manifolds of bounded geometry
We use the notation of Definition 2.1.
###### 3.1 Lemma.
Let $`(M^n,g)`$ be a Riemannian manifold with boundary and with bounded geometry as in Definition 2.4. We find $`r_0>0`$ such that for all $`r,sr_0`$ the following holds:
1. If $`x,x^{}M`$ and $`Z(x^{},s,\frac{2R_2}{3})U_r(Z(x,\frac{R_1}{2},\frac{2R_2}{3}))\mathrm{}`$ then $`Z(x^{},s,\frac{2R_2}{3})Z(x,\frac{9R_1}{10},\frac{2R_2}{3})`$.
2. We find $`0<D_1(r)<D_2(r)`$ for $`r0`$ such that
$$\begin{array}{cc}& D_1(r)\mathrm{vol}(B(x,r))D_2(r)xMN(\frac{R_2}{3}),\text{ if }r<R_3\hfill \\ & D_1(r)\mathrm{vol}(Z(x^{},r,\frac{2R_2}{3}))D_2(r)x^{}M,\text{if }r<R_1.\hfill \end{array}$$
$`D_1(r)`$, $`D_2(r)`$ and $`r_0`$ can be chosen to depend only on the bounded geometry constants.
###### Proof.
Bounded geometry implies the existence of $`C_1,C_2>0`$ so that in normal coordinates
$$(g^{ij})_{i,j}<C_1,(g_{ij})_{i,j}<C_1,\text{and }C_2\sqrt{\left|det(g_{ij})\right|}C_1.$$
Observe that $`d(Z(x,\frac{R_1}{2},\frac{2R_2}{3}),MZ(x,\frac{3R_1}{4},\frac{9R_2}{10}))`$ is bounded independent of $`xM`$, using the bounds on the metric tensor. With all sets and distances in $`M`$, $`d(B(x,\frac{3R_1}{4}),MB(x,\frac{9R_1}{10}))R_1/10`$. Choose $`r_0`$ smaller than half the minimum of these two bounds. If $`r,s<r_0`$ and $`Z(x^{},s,\frac{2R_2}{3})U_r(Z(x,\frac{R_1}{2},\frac{R_2}{2}))\mathrm{}`$ for $`x,x^{}M`$ then $`Z(x^{},s,\frac{2R_2}{3})Z(x,\frac{3R_1}{4},\frac{9R_2}{10})\mathrm{}`$ which in turn implies $`Z(x^{},s,\frac{2R_2}{3})Z(x,\frac{9R_1}{10},\frac{2R_2}{3})`$ by the choice of $`r_0`$. This proves the first assertion.
The assertion about the volume bounds follows immediately from the upper and lower bounds of $`\sqrt{\left|det(g_{ij})\right|}`$.
We can choose all constants to depend only on the bounded geometry constants. ∎
The following is important to do analysis on manifolds of bounded geometry. The corresponding result for empty boundary is due to Shubin \[10, A1.2 and A1.3\].
###### 3.2 Proposition.
(Partition of unity)
Let $`M`$ be a manifold with boundary and of bounded geometry as in Definition 2.4. There are $`r_m>0`$ and, for $`0<r<r_m`$ constants $`C_K>0`$ ($`K`$), $`M_f`$, all depending only on the bounded geometry constants (and $`r`$) such that a covering of $`M`$ exists by sets $`\{U(x_i,r)\}_{iI}`$ which has the following properties:
1. $`x_iM`$ for $`i0`$ and $`U(x_i,r)=Z(x_i,r,\frac{2R_2}{3})`$;
$`x_iMN(\frac{R_2}{2})`$ for $`i<0`$ and $`U(x_i,r)=B(x_i,r)`$.
2. If $`s<r_m`$ and $`xM`$ then $`B(x,s)U(x_i,r)\mathrm{}`$ for at most $`M_f`$ of the $`x_i`$.
3. $`\{U(x_i,r/2)\}_{iI}`$ is a covering of $`M`$, too.
Denote with $`\kappa _i:B(0,r)U(x_i,r)`$ ($`i<0`$) and $`\kappa _i:B(0,r)\times [0,\frac{2R_2}{3})U(x_i,r)`$ for $`i0`$ the corresponding normal charts.
To this covering, a subordinate partition of unity $`\{\phi _i\}`$ exists such that
$$|D^\alpha \phi _i|C_Ki\left|\alpha \right|K\text{(in normal coordinates).}$$
###### Proof.
Set $`r_m:=\mathrm{min}\{R_1/2,R_2/12,R_3,r_0/2\}`$, where $`r_0`$ is given by Lemma 3.1. Let $`0<r<r_m`$. First choose a maximal set of points $`\{x_iM;i=0,1,2,\mathrm{}\}`$ such that all $`(B(x_i,r/4)M)`$ are disjoint. Next, choose a maximal set of points $`\{x_iMN(\frac{R_2}{2});i=1,2,\mathrm{}\}`$ such that all $`B(x_i,r/4)`$ ($`i<0`$) are disjoint. Note that the set $`I`$ of $`i`$ obtained this way may be a proper subset of $``$. For $`0<sr_0`$ set $`U(x_i,s):=B(x_i,s)`$ if $`i<0`$ and $`U(x_i,s):=Z(x_i,s,\frac{2R_2}{3})`$ if $`i0`$. Then
$$\underset{i<0}{}U(x_i,r/2)=\underset{i<0}{}B(x_i,r/2)\}\text{covers }MN(\frac{R_2}{2}).$$
This is true because else we find $`zMN(\frac{R_2}{2})`$ which has distance $`r/2`$ to all of the $`x_i`$. Then $`B(z,r/4)B(x_i,r/4)=\mathrm{}`$ $`i<0`$, violating the maximality of $`\{x_i\}_{i<0}`$. Similarly, $`\{B(x_i,r/2)M\}_{i0}`$ covers $`M`$ $``$ $`\{U(x_i,r/2)\}_{i0}`$ covers $`N(\frac{2R_2}{3})`$.
Now we have to show that the covering $`\{U(x_i,r)\}_{iI}`$ has Property 2. So fix $`0<s<r_m`$ and $`xM`$.
* If $`xN(\frac{R_2}{3})`$ and $`i<0`$ then $`B(x,s)U(x_i,r)=\mathrm{}`$ since $`d(N(\frac{R_2}{3}),MN(\frac{R_2}{2}))=\frac{R_2}{6}>r+s`$.
* If $`xMN(\frac{R_2}{3})`$ then the number $`N_1`$ of $`x_i`$ ($`i<0`$) with $`U(x_i,r)B(x,s)\mathrm{}`$ is by Lemma 3.1 bounded by
$$N_1\frac{\mathrm{vol}(B(x,s+r))}{inf_{x_iMN(\frac{R_2}{2})}\mathrm{vol}(B(x_i,r/4))}\frac{D_2(2r_m)}{D_1(r/4)}$$
since for such $`x_i`$ we have $`B(x_i,r/4)B(x,s+r)`$ and all of these are disjoint.
* If $`xMN(R_2)`$ then $`B(x,s)U(x_i,r)=\mathrm{}`$ for $`i0`$ since $`d(N(\frac{2R_2}{3}),MN(R_2))=\frac{R_2}{3}>s`$.
* If $`xN(R_2)`$ then the number $`N_2`$ of $`x_i`$ ($`i0`$) with $`B(x,s)U(x_i,r)\mathrm{}`$ is bounded by
$$N_2\frac{sup_{x_iM}\mathrm{vol}(Z(x_i,\frac{9R_1}{10},\frac{2R_2}{3}))}{inf_{x_iM}\mathrm{vol}(Z(x_i,\frac{r}{4},\frac{2R_2}{3}))}\frac{D_2(9R_1/10)}{D_1(r/4)}$$
since if there is one such $`i_0`$ then for all other such $`i`$ by Lemma 3.1 $`Z(x_i,\frac{r}{4},\frac{2R_2}{3})Z(x_{i_0},9R_1/10,\frac{2R_2}{3})`$, and all the $`Z(x_i,\frac{r}{4},\frac{2R_2}{3})`$ are disjoint.
It follows in all cases
$$M_f(r)\frac{D_2(9R_1/10)+D_2(2r_0)}{D_1(r/4)}<\mathrm{}.$$
It remains to construct the subordinate partition of unity. Choose a smooth cutoff function $`\phi :^m[0,1]`$ with $`\phi (x)=1`$ if $`\left|x\right|r/2`$ and $`\phi (x)=0`$ if $`\left|x\right|r`$. Denote the restriction to $`^{m1}`$ also with $`\phi `$. Choose smooth $`\psi :[0,1]`$ with $`\psi (x)=0`$ if $`x2R_2/3`$ and $`\psi (x)=1`$ if $`xR_2/2`$. Via the normal coordinates this yields cutoff functions $`f_i`$ on $`U(x_i,r)`$ with $`f_i\kappa _i(y^{},t)=\phi (y^{})\psi (t)`$ if $`i0`$ and $`f_i\kappa _i=\phi `$ if $`i<0`$. Therefore, if $`\kappa `$ is any normal chart, $`f_i\kappa =\phi (\kappa _i^1\kappa )`$ ($`i<0`$) and $`f_i\kappa =(\phi \psi )(\kappa _i^1\kappa )`$ ($`i0`$). The chain rule shows that the bounds on derivatives up to order $`K`$ of the coordinate changes (Proposition 3.3) yield bounds on the partial derivatives up to order $`K`$ of $`f_i`$ in normal coordinates. To construct the partition of unity, set $`F=_{iI}f_i`$ (at each point there are at most $`M_f`$ non-zero summands).
$$\text{Since}MN(\frac{R_2}{2})\underset{i<0}{}U(x_i,r/2)\text{and}N(\frac{R_2}{2})\underset{i0}{}U(x_i,r/2),$$
for each $`zM`$ at least one of $`f_i(z)=1`$ $``$ $`F1`$. Define
$$\phi _i:=f_i/F.$$
Obviously, $`\{\phi _i\}_{iI}`$ is a smooth partition of unity subordinate to our covering. Pick one $`\phi _i`$ and one normal chart $`\kappa `$. For partial derivatives up to order $`K`$ in normal coordinates observe
$$\begin{array}{cc}\hfill \left|D^\alpha (\phi _i\kappa )\right|& =\left|D^\alpha \frac{f_i\kappa }{F\kappa }\right|=\frac{\left|P_\alpha (D^\beta (f_i\kappa ),D^\gamma (F\kappa );\left|\beta \right|,\left|\gamma \right|\left|\alpha \right|)\right|}{\left|F\kappa \right|^{2^{\left|\alpha \right|}}}\hfill \\ & \stackrel{\left|F\right|1}{}\left|P_\alpha (D^\beta (f_i\kappa ),D^\gamma (F\kappa ))\right|.\hfill \end{array}$$
$`P_\alpha `$ is a polynomial entirely determined by $`\alpha `$. At every point $`xM`$, $`D^\gamma (F\kappa )|_x`$ is the sum of at most $`M_f`$ summands of the type $`D^\gamma (f_s\kappa )|_x`$. Therefore, we have bounds for all the entries of $`P_\alpha `$. This yields a bound $`C_K`$, depending only on the bounded geometry constants, for $`\left|D^\alpha (\phi _i\kappa )\right|`$ if $`\left|\alpha \right|K`$. ∎
Changes of normal coordinates
###### 3.3 Proposition.
Suppose $`M`$ is a Riemannian manifold with boundary and of bounded geometry. More precisely, suppose $`C>0`$ is a bound for partial derivatives up to order $`k+1`$ of $`g^{ij}`$ and $`g_{ij}`$ in normal coordinates. Then $`D>0`$ exists, depending only on $`C`$ so that, if $`\kappa _1:U_1^mM`$ and $`\kappa _2:U_2^mM`$ are normal charts as in 2.5, the following holds for $`f:=\kappa _1^1\kappa _2:U_0^m^m`$ ($`U_0`$ the domain of definition of the composition):
$$\left|D^\alpha f\right|D\left|\alpha \right|k.$$
Since the maps $`\kappa _i`$ are solutions of certain ordinary differential equation, namely the equation for geodesics, we first recall a result about differential equations.
###### 3.4 Lemma.
Let $`x^{}(t)=F(t,x(t))`$ be a system of ordinary differential equations ($`t`$, $`x(t)^n`$), $`FC^{\mathrm{}}(\times ^n,^n)`$. Suppose $`\phi (t,x)`$ is the flow of this equation. We find a universal expression $`Expr_\alpha `$, only depending on $`\alpha `$ such that for all $`t0`$ where $`\varphi (t,x_0)`$ is defined
$$\left|D_x^\alpha \varphi (t,x_0)\right|Expr_\alpha (\underset{0\tau t}{sup}\{|D_x^\beta F(\tau ,\varphi (\tau ,x_0)|\}|\beta \alpha ,t).$$
(3.5)
###### Proof.
The theory of ordinary differential equations \[7, V.3.1.\] tells us that we have the linear differential equation
$$\alpha ^{}(t)=\frac{F}{x}(t,\phi (t,x))\alpha (t);\alpha (0)=e_k=(0,\mathrm{},1,\mathrm{},0)$$
for $`_k\phi (t,x)`$. For linear differential equations \[7, IV.4.2\] gives inequalities which directly imply (3.5) if $`\left|\alpha \right|=1`$.
Inductively one shows that higher derivatives fulfill the linear differential equation
$$\begin{array}{c}(D_x^\alpha \phi )^{}(t,x)=(D_xF)(t,\phi (t,x))D_x^\alpha \phi (t,x)+\hfill \\ \hfill P_\alpha (D_x^\gamma \phi ,(D_x^\beta F)(t,\phi (t,x));\gamma <\alpha ,\beta \alpha )\end{array}$$
(3.6)
with $`D_x^\alpha \phi (0,x)=0`$ if $`\left|\alpha \right|>1`$. Here $`P_\alpha `$ is a polynomial matrix which depends only on $`\alpha `$. By induction and using \[7, IV.4.2\] again, the proposition follows. ∎
Reduction of order implies:
###### 3.7 Corollary.
A statement corresponding to Lemma 3.4 holds for ordinary differential equations of order $`k`$.
###### 3.8 Lemma.
Let $`pU^m`$, $`g_{ij}`$ a Riemannian metric on $`U`$ and $`\mathrm{exp}_p:B(r,0)U`$ the exponential map at $`p`$ (we identify $`T_pU`$ with $`^m`$ via an orthonormal frame). If $`r`$ is sufficiently small then $`\mathrm{exp}_p`$ is a diffeomorphism onto some open subset of $`U`$, and the derivatives up to order $`k`$ of $`\mathrm{exp}_p`$ and its inverse are bounded in terms of $`g_{ij}`$, $`g^{ij}`$, their derivatives up to order $`k+1`$, and $`r`$.
###### Proof.
We have $`\mathrm{exp}_p(x)=\phi (x,p,1)`$, where $`\phi `$ with $`\phi (x,q,0)=q`$, $`\phi ^{}(x,q,0)=x`$ is the flow of the differential equation for geodesics. Corollary 3.7 applies to this equation $`x^{\prime \prime }=F(x)`$, and $`F(x)=_{i,j}\mathrm{\Gamma }_{ij}(x)x_i^{}x_j^{}`$ is given by $`g_{ij}`$ and its first order derivatives.
For the inverse, by Lemma 2.17 it suffices to study its first order derivatives. Bounds on these follow from Proposition 2.6. ∎
###### 3.9 Lemma.
Suppose $`U^{},V^{m1}`$, $`\kappa :U^{}\times [0,r_C)V\times [0,r_C)`$ is a normal boundary chart centered at $`pV`$ on the Riemannian manifold $`V\times [0,r_C)`$ with metric $`g_{ij}`$ ($`g_{im}=\delta _{im}=g_{mi}`$). Then the derivatives of $`\kappa `$ and its inverse up to order $`k`$ are bounded in terms of $`g_{ij}`$, $`g^{ij}`$ and their derivatives up to order $`k+1`$.
###### Proof.
$`\kappa (q,s)=\phi _1(s_m,\phi _2(q,p,1),0,1)`$, where $`\phi _1`$ is the flow of the differential equation for the geodesics in $`V\times [0,r_C)`$ ($`\phi _1(v,p,\tau ,0)=(p,\tau )`$, $`\phi _1^{}(v,(p,\tau ),0)=v`$), and $`\phi _2`$ is the flow of the differential equation for geodesics on $`V`$. Hence $`\kappa `$ is the composition of two flows to which Corollary 3.7, and then Lemma 2.17 and Proposition 2.6 applies exactly as in the previous lemma. ∎
We prove Proposition 3.3 using these Lemmas as follows: By Theorem 2.5 we have bounds for $`g_{ij}`$ and their derivatives up to order $`k+1`$ in normal coordinates. Write
$$\kappa _1^1\kappa _2=(\kappa _1^1\kappa _0)(\kappa _0^1\kappa _2).$$
with $`\kappa _0`$ either being an exponential map or a normal boundary map with suitable range, respectively. If we use $`\kappa _1`$ or $`\kappa _2`$ to pull back the given Riemannian metric to the domain of the charts, $`\kappa _0^1\kappa _2`$ and $`\kappa _1^1\kappa _0`$ each fulfill exactly the assumptions of one of the two lemmas. The conclusion of these lemmas is then true for there composition, as well, and the Proposition follows.
###### 3.10 Corollary.
To check condition (B1) in Definition 2.4 it suffices to do this for an atlas of such charts. |
warning/0001/cond-mat0001171.html | ar5iv | text | # Infrared and Raman spectra of LiV2O5 single crystals
## I Introduction
During the past several years low dimensional quantum spin systems such as the spin-Peierls, spin-ladder and an antiferromagnetic Heisenberg linear chain systems have attracted much attention . The vanadate family of AV<sub>2</sub>O<sub>5</sub> oxides (A = Li, Na, Cs, Mg and Ca), which have common V<sup>4+</sup>O<sub>5</sub> square pyramids in the structure, have demonstrated a variety of the low dimensional quantum spin phenomena: charge-ordering transition in NaV<sub>2</sub>O<sub>5</sub> , spin-gap behavior in Ca(Mg)V<sub>2</sub>O<sub>5</sub> and CsV<sub>2</sub>O<sub>5</sub> and typical 1-D behavior without spin gap in LiV<sub>2</sub>O<sub>5</sub> . The magnetic susceptibility data, fitted in the framework of the Bonner-Fisher model with the exchange interaction J=308 K , suggest that the magnetic properties of this system can be described using homogenous Heisenberg antiferromagnetic linear chain model. The nuclear magnetic resonance (NMR) measurements showed the formation of the staggered spin configurations due to the existence of the finite-size effect.
However, in spite of good understanding of the magnetic properties, the study of the vibrational properties of LiV<sub>2</sub>O<sub>5</sub> is of a great importance because of the still puzzling interplay between the charge and the magnetic ordering in NaV<sub>2</sub>O<sub>5</sub> . Thus, in this work we present the polarized far-infrared (FIR) reflectivity as well as the Raman scattering spectra of LiV<sub>2</sub>O<sub>5</sub> single crystals. The 16$`A_g`$, 7$`B_{1g}`$, 6$`B_{3g}`$, 9$`B_{3u}`$, and 5$`B_{2u}`$ symmetry modes are experimentally observed. The assignment of the vibrational modes is done by comparing the phonons in LiV<sub>2</sub>O<sub>5</sub> and NaV<sub>2</sub>O<sub>5</sub>. Furthermore, our spectra demonstrate the dominance of the layer symmetry in LiV<sub>2</sub>O<sub>5</sub>.
## II Experiment
The present work was performed on single crystal plates with dimensions typically about $`1\times 4\times 0.5`$ mm<sup>3</sup> in the a, b, and c axes, respectively. The details of sample preparation were published elsewhere . The infrared measurements were carried out with a BOMEM DA-8 FIR spectrometer. A DTGS pyroelectric detector was used to cover the wave number region from 100 to 700 cm<sup>-1</sup>; a liquid nitrogen cooled HgCdTe detector was used from 500 to 1500 cm<sup>-1</sup>. Spectra were collected with 2 cm<sup>-1</sup> resolution, with 1000 interferometer scans added for each spectrum. The Raman spectra were measured in the backscattering configuration using micro-Raman system with DILOR triple monochromator including liquid nitrogen cooled CCD-detector. An Ar-ion laser was used as an excitation source.
## III Results and discussion
LiV<sub>2</sub>O<sub>5</sub> has an orthorhombic unit cell with parameters a=0.9702 nm, b=0.3607 nm, c= 1.0664 nm, Z=4 and space group Pnma (D$`{}_{}{}^{16}{}_{2h}{}^{}`$). Each vanadium atom is surrounded by five oxygen atoms, forming V$`O_5`$ pyramids. These pyramids are mutually connected via common edges to form zigzag chains along the b-direction. Such a crystalline structure is characterized by two kinds of vanadium double chains along the b-axis. One is magnetic V<sup>4+</sup> (S=1/2) and the other one is nonmagnetic V<sup>5+</sup> (S=0) chain. These chains are linked by corner sharing to form layers in (001) - plane, Fig. 1 (a). The Li atoms are situated between these layers, as shown in Fig. 1.1.(b). The LiV<sub>2</sub>O<sub>5</sub> unit cell consists of four formula units comprising 32 atoms in all (Fig. 1). The site symmetry of all atoms in Pnma space group is C<sub>s</sub>. The factor-group-analysis (FGA) yields :
(Li, V<sub>1</sub>, V<sub>2</sub> , O<sub>1</sub>, O<sub>2</sub>, O<sub>3</sub>, O<sub>4</sub>, O<sub>5</sub>) (C<sub>s</sub>):
$`\mathrm{\Gamma }`$ = 2$`A_g`$ \+ $`A_u`$ \+ $`B_{1g}`$ \+ 2$`B_{1u}`$ \+ 2$`B_{2g}`$ \+ $`B_{2u}`$ \+ $`B_{3g}`$ \+ 2$`B_{3u}`$,
Summarizing these representations and subtracting the acoustic ($`B_{1u}`$ \+ $`B_{2u}`$ \+ $`B_{3u}`$) and silent (8$`A_u`$) modes, we obtained the following irreducible representations of LiV<sub>2</sub>O<sub>5</sub> vibrational modes of Pnma space group:
$`\mathrm{\Gamma }`$ = 16$`A_g`$(aa,bb,cc) + 8$`B_{1g}`$(ab) + 16$`B_{2g}`$(ac) + 8$`B_{3g}`$(bc) + 15$`B_{1u}`$(E$`||`$c) + 7$`B_{2u}`$(E$`||`$b) + 15$`B_{3u}`$(E$`||`$a)
Thus, 48 Raman and 37 infrared active modes are expected to show up in the LiV<sub>2</sub>O<sub>5</sub> spectra. The room temperature polarized far-infrared reflectivity spectra of LiV<sub>2</sub>O<sub>5</sub> are given in Fig. 2. The open circles are the experimental data and the solid lines represent the spectra computed using a four-parameter model for the dielectric constant .
$$ϵ=ϵ_{\mathrm{}}\underset{j=1}{\overset{n}{}}\frac{\omega _{LO,j}^2\omega ^2+ı\gamma _{LO,j}\omega }{\omega _{TO,j}^2\omega ^2+ı\gamma _{TO,j}\omega }$$
(1)
where $`\omega _{LO,j}`$ and $`\omega _{TO,j}`$ are longitudinal and transverse frequencies of the j<sup>th</sup> oscillator, $`\gamma _{LO,j}`$ and $`\gamma _{TO,j}`$ are their corresponding dampings, and $`ϵ_{\mathrm{}}`$ is the high-frequency dielectric constant. The static dielectric constant, given in Table I, is obtained using the generalized Lyddane-Sachs-Teller relation $`ϵ_0=ϵ_{\mathrm{}}_{j=1}^n\omega _{LO,j}^2/\omega _{TO,j}^2`$.
The best-oscillator-fit parameters are listed in Table I. The agreement between observed and calculated reflectivity spectra is rather good. For the E$`||`$a polarization, nine oscillators with TO frequencies at about 202, 243, 255, 340, 397, 545, 724, 948, 1006 cm<sup>-1</sup> are clearly seen. In the E$`||`$b polarization (Fig. 2 (b)) five oscillators at 180, 255, 323, 557 and 595 cm<sup>-1</sup> are observed. We failed to obtain usefull signal for E$`||`$c spectra because of the very small thickness (c-axis) of the sample. The room and the low temperature Raman spectra of LiV<sub>2</sub>O<sub>5</sub>, for parallel and crossed polarizations, are given in Fig. 3. The spectra for parallel polarizations consist of $`A_g`$ symmetry modes. Thirteen modes at 100, 123, 172, 209, 328, 374, 398, 528, 550, 639, 725, 965 and 989 cm<sup>-1</sup> are clearly seen for the (aa) polarization, two additional modes at 197 and 456 cm<sup>-1</sup> for the (bb) polarization and one additional mode at 267 cm<sup>-1</sup> for the (cc) polarization. For the crossed (ab) polarization seven Raman active $`B_{1g}`$ symmetry modes at 168, 250, 270, 333, 546, 646 and 737 cm<sup>-1</sup> are found. In the case of the (bc) polarization, six $`B_{3g}`$ symmetry modes were observed with almost the same frequencies as $`B_{1g}`$ modes. The Raman spectra for (ac) polarization is given in Fig. 3 (f). For this polarization ($`B_{2g}`$ symmetry) we could not resolve any new mode. Namely, all modes observed for this polarization have been already seen in parallel or in other crossed polarizations, probably because of low-quality of the (010) surface. The frequencies of all observed Raman active modes are given in Table II.
We will first consider the Raman spectra shown in Fig. 3. From the 16 $`A_g`$ , 8$`B_{1g}`$, 8$`B_{3g}`$ and 16$`B_{2g}`$ modes predicted by FGA of LiV<sub>2</sub>O<sub>5</sub>, we clearly observe 16$`A_g`$, 7$`B_{1g}`$, and 6$`B_{3g}`$ modes. The missing modes seam to be of a very weak intensity. The (ac) polarized spectra (Fig. 3 (f)) consist mostly of $`A_g`$ modes (and some $`B_{1g}`$). The appearance of $`B_{1g}`$ and $`B_{3g}`$ symmetry modes at the same frequencies lead us to consider the crystal structure of this oxide once again. The LiV<sub>2</sub>O<sub>5</sub> crystal is a layer crystal (see Fig. 1 (b)). The vibrational properties of such crystals demonstrate the dominance of the layer symmetry . The unit cell of LiV<sub>2</sub>O<sub>5</sub> consists of two layers with 16 atoms in all. The full symbol of space group is P 2<sub>1</sub>/n 2<sub>1</sub>/m 2<sub>1</sub>/a. The first symbol represents $`2_1`$-screw axis along a-axis followed by diagonal glide plane (n) perpendicular to a\- axis with translation (b+c)/2; the second symbol is the $`2_1`$ -screw axis parallel to b-axis with mirror plane perpendicular to it and third symbol means $`2_1`$-screw axis parallel to c-axis with glide plane perpendicular to c-axis with translation of a/2 (see the lower part of Fig. 4). If we consider only one layer we break the periodicity in the direction perpendicular to the layer. As a consequence, there are no more symmetry operations along the a\- and c-axes. Now we should consider the layer symmetry in terms of diperiodic groups rather than triperiodic space groups. We found four operation of layer symmetry, Fig. 4:
1, identity
$`2_1`$, twofold screw axis parallel to the b-axis
$`\overline{1}`$, center of symmetry located between VO<sub>5</sub> pyramids
m, mirror plane perpendicular to the b-axis.
The DG15 diperiodic group is the only symmetry operation group of 80 that has these symmetry operations. The full symbol of this diperiodic group is P 1 $`2_1`$/m 1. This group is isomorphic with the C$`{}_{}{}^{2}{}_{2h}{}^{}`$ (P$`2_1`$/m) space group (second setting) . All atoms are in symmetry position (e) of this space group (with C<sub>s</sub> point symmetry). The normal mode distribution for the layer is
$`\mathrm{\Gamma }`$=16$`A_g`$ (xx,yy,zz, xz) +8$`B_g`$ (xy, yz) +8$`A_u`$ (E$`||`$y) +16$`B_u`$ (E$`||`$x, E$`||`$z)
The compatibility diagram, relating the layer and the crystal vibration of LiV<sub>2</sub>O<sub>5</sub>, together with the schematic representation of the layer and the crystal symmetry operations as well as the experimental conditions for the observation of the optical modes, are given in Fig. 4. According to Fig. 4 the $`A_g`$ modes for the layer symmetry appear for parallel and for (ac) crossed polarization and the $`B_{1g}`$ and $`B_{3g}`$ modes of crystal symmetry merge into $`B_g`$ modes of the layer symmetry. These facts are fully in agreement with our Raman spectra, Fig. 3. Thus, we can conclude that the vibrational properties of crystal are predominantly due to vibrational properties of the layer. As we already mentioned, the size of samples in (010) plane and worse quality of this surface do not allow us to check this conclusion also by infrared spectroscopy.
By comparing the LiV<sub>2</sub>O<sub>5</sub> spectra with the corresponding spectra and the lattice dynamics of NaV<sub>2</sub>O<sub>5</sub> we analyze the LiV<sub>2</sub>O<sub>5</sub> phonon properties. The basic building blocks of LiV<sub>2</sub>O<sub>5</sub> crystal structure are VO<sub>5</sub> pyramids which are mutually connected by edge to build the chains. Such chains are also present in NaV<sub>2</sub>O<sub>5</sub>. The difference between crystal structures of Na- and Li- vanadate is illustrated in Fig. 5. In NaV<sub>2</sub>O<sub>5</sub> there is only one vanadium site while in LiV<sub>2</sub>O<sub>5</sub> two nonequivalent vanadium atom positions are present. Besides that, the V<sub>1</sub>-O-V<sub>2</sub> angle in LiV<sub>2</sub>O<sub>5</sub> is smaller (120<sup>o</sup>) than in NaV<sub>2</sub>O<sub>5</sub> (140<sup>o</sup>). Distance between atoms in VO<sub>5</sub> pyramids does not differ significantly in Li- and Na- vanadates. This produces only a small frequency shift of the corresponding modes in these compounds. The existence of the two different VO<sub>5</sub> chains with nearly the same inter-atomic distance can produce the appearance of mode doublets. Raman spectra for (aa) and (bb) polarizations of both compounds are shown in Fig. 5. As it can be seen from Fig. 5, each mode of NaV<sub>2</sub>O<sub>5</sub> in the bond stretching region (above 450 cm<sup>-1</sup>) appears as a doublet in LiV<sub>2</sub>O<sub>5</sub>. The highest intensity mode in NaV<sub>2</sub>O<sub>5</sub> originates from V-O<sub>1</sub> bond stretching vibration. The distance between these atoms in Na-vanadate is 1.62 $`\stackrel{̊}{A}`$. In Li vanadate there are two such bonds, one longer bond 1.65 $`\stackrel{̊}{A}`$ (V<sub>1</sub>-O<sub>3</sub>) and one slightly shorter 1.61 $`\stackrel{̊}{A}`$ (V<sub>2</sub>-O<sub>2</sub>) than V-O<sub>1</sub> bond in Na-vanadate. Thus, we can expect the appearance of two modes with frequencies higher and lower than the Na V-O<sub>1</sub> bond stretching frequency. In fact, two modes are observed in Fig. 5 at lower and higher frequency than Na V-O<sub>1</sub> mode. Applying the same analysis to the another doublet in bond stretching energy region we conclude that 528/550 cm<sup>-1</sup> modes originate from V<sub>1</sub>-O<sub>4</sub> (V<sub>2</sub>-O<sub>5</sub>) bond stretching vibration. This doublet corresponds to the mode at 534 cm<sup>-1</sup> of NaV<sub>2</sub>O<sub>5</sub>.
Similar non - phononic broad structure, observed in Na - vanadate for (aa) polarization at about 640 cm<sup>-1</sup> (see Fig. 5), is found in LiV<sub>2</sub>O<sub>5</sub> as a doublet at about 639/725 cm<sup>-1</sup>. There are several scenarios to explain this structure . First, electric dipole transitions between split crystal field levels, second two-magnon scattering, and third possible electron-phonon coupled modes observed in antiferromagnets. In the case of NaV<sub>2</sub>O<sub>5</sub> the wide structure shifts to lower energy upon cooling . Fischer et al. concluded that such kind of behavior comes from electron-phonon coupling like in antiferromagnetic FeCl<sub>2</sub>. The opposite frequency shift vs. temperature of wide structure maximum in NaV<sub>2</sub>O<sub>5</sub> then in FeCl<sub>2</sub> is connected with the strong spin fluctuations in low dimensions .
Two-magnon excitations lead to broad band structures in Raman scattering spectra. The very recent neutron scattering measurements of LiV<sub>2</sub>O<sub>5</sub> show antiferromagnetic periodicity along the b-axis (chain direction). These results suggests that LiV<sub>2</sub>O<sub>5</sub> can be regarded as a collection of independent S=1/2 AF linear chains with zone boundary energy of $`42\pm 3`$ meV. The frequencies of observed modes (640/725 cm<sup>-1</sup>) are comparable with the values of exchange coupling energy J=308 K (3J=640 cm<sup>-1</sup>) and zone boundary energy range (2E<sub>ZB</sub>= 628-724 cm<sup>-1</sup>), respectively. However, the polarization selection rules allow the appearance of modes for polarization of incident and scattered light parallel to the dominant exchange path (b-axis direction). Because we observed broad modes only for (aa) and (cc) polarization and there is no frequency shift of these modes to higher energies by lowering of the temperature we concluded that two-magnon scenario is not realistic.
Finally, we consider the appearance of these modes as crystal field induced . These transitions between different crystal field levels should be discrete, and no significant frequency shift vs. temperature can be expected. This is in accordance with our spectra given in Fig.3. The appearance of two peaks can be understood due to existence of two structurally different strongly deformed VO<sub>5</sub> pyramids.
The mode at 788 cm<sup>-1</sup> is probably IR active LO mode (790 cm<sup>-1</sup>, see Table I) but its appearance is not understood at the moment.
The modes at frequencies below 450 cm<sup>-1</sup> originate from bond bending vibrations. In the Na-vanadate the V-O<sub>3</sub>-V bending mode appears at 448 cm<sup>-1</sup>. The corresponding mode in Li-vanadate could be the one at 374 cm<sup>-1</sup>. Note that, the frequency of this mode is about 16 % lower than in Na-vanadate, as a consequence of the change of the bond-bending force constant. We make an approximate estimation for the change of the bond-bending force constant by scaling the frequency as the square root of the force constant, and assuming a scaling of the force constants as R<sup>-6</sup> (R is the bond length) . Similarly, scaling of the phonon frequency for the bond stretching mode is R<sup>-3</sup>. In our case the R (O-V-O)<sub>Na</sub>/R(O-V-O)<sub>Li</sub> is 0.98. This parameter can produce the decrease of the phonon frequency of only 5 %. The remaining part of frequency decrease probably comes from modification of bond-bending force constant due to redistribution of the electron density . Namely, according to Refs. the electrons in NaV<sub>2</sub>O<sub>5</sub> are located in a V-O<sub>3</sub>-V molecular orbital. This is not the case in LiV<sub>2</sub>O<sub>5</sub> where the electrons are located in V<sup>4+</sup> chains.
The highest intensity mode for (bb) polarization in Li sample is the mode at 328 cm<sup>-1</sup>. This mode originates from O-V-O bending vibrations as the highest intensity (bb) polarized mode in Na-sample. The broad structure at about 398 cm<sup>-1</sup> seems to correspond to Li vibration. If we consider mass effect only, the replacement of heavier Na atoms with lighter Li produces the shift of corresponding mode towards higher frequency according to ($`\omega _{Na}`$=179 cm<sup>-1</sup>) x (m<sub>Na</sub>/m<sub>Li</sub>)<sup>1/2</sup> = 326 cm<sup>-1</sup>, which is not to far from the observed mode. Besides, the Li-O vibrations are expected at about 390 cm<sup>-1</sup> . The pair of modes at 172/209 cm<sup>-1</sup> originates from O-V-O bending vibration, and the pair of lowest frequency modes at 100/123 cm<sup>-1</sup> represents chain rotation modes.
The similar analysis could be conducted for B<sub>1(3)g</sub> modes. Namely, the modes at 646 and 737 cm<sup>-1</sup> are V<sub>1</sub>-O<sub>4</sub> and V<sub>2</sub>-O<sub>5</sub> bond stretching vibrations; the three modes at 250, 270 and 333 cm<sup>-1</sup> are bond bending vibrations, and the lowest frequency mode at 168 cm<sup>-1</sup> represents chain rotation mode. We should mention that similar mode exists in NaV<sub>2</sub>O<sub>5</sub> for crossed polarization too.
Identification of the infrared modes, obtained by comparison with IR spectra and lattice dynamics of NaV<sub>2</sub>O<sub>5</sub> is given in Table I and we do not repeat it here again.
In conclusion, we have measured the infrared and Raman spectra of LiV<sub>2</sub>O<sub>5</sub>. The assignment of the phonons is done by comparing the infrared and Raman spectra of LiV<sub>2</sub>O<sub>5</sub> and NaV<sub>2</sub>O<sub>5</sub>. The factor-group-analysis of the LiV<sub>2</sub>O<sub>5</sub> crystal symmetry and of its constituent layers is performed to explain the symmetry properties of the observed modes. Finally, our spectra demonstrate the dominance of the layer symmetry in LiV<sub>2</sub>O<sub>5</sub>.
Acknowledgments
Z.V.P. acknowledges support from the Research Council of the K.U. Leuven and DWTC. The work at the K.U. Leuven is supported by Belgian IUAP and Flemish FWO and GOA Programs. MJK thanks Roman Herzog - AvH for partial financial support. |
warning/0001/cond-mat0001410.html | ar5iv | text | # Three-Dimensional Elastic Compatibility: Twinning in Martensites
## Abstract
We show how the St.Venant compatibility relations for strain in three dimensions lead to twinning for the cubic to tetragonal transition in martensitic materials within a Ginzburg-Landau model in terms of the six components of the symmetric strain tensor. The compatibility constraints generate an anisotropic long-range interaction in the order parameter (deviatoric strain) components. In contrast to two dimensions, the free energy is characterized by a “landscape” of competing metastable states. We find a variety of textures, which result from the elastic frustration due to the effects of compatibility. Our results are also applicable to structural phase transitions in improper ferroelastics such as ferroelectrics and magnetoelastics, where strain acts as a secondary order parameter.
Strain plays a crucial role in many structural phase transitions, either as a primary order parameter (OP), e.g., in technologically important martensites and shape memory alloys , or as a secondary OP, e.g., in ferroelectric and magnetoelastic materials. Moreover, self-consistent coupling of strain to either charge (e.g., high-temperature superconductors ) or spin (e.g., colossal magnetoresistance manganites ) degrees of freedom determines key physical properties. However, the physical degrees of freedom in a continuum elastic medium (such as described by Ginzburg-Landau models) are contained in the displacement field, $`u(r)`$, even though it is the strain fields, $`ϵ_{ij}`$, which appear in the free energy. Instead of treating the strains as independent fields, one must assume that they correspond to a physical displacement field, i.e., they are derivatives of a single continuous function. This is achieved by requiring that they satisfy a set of nontrivial (Saint-Venant) compatibility relations concisely expressed by the equation $`\times (\times ϵ)=0`$ (in the absence of voids, dislocations, etc.). Indeed, ignoring the geometrical compatibility constraint and minimizing the free energy directly would lead to the erroneous result that non-OP strain components are identically zero, and that the OP responds trivially to perturbations, such as stress or local disorder, which is incorrect.
We show for the first time that a consistent Ginzburg-Landau (GL) understanding of martensitic textures in terms of the OP-strain alone, and involving long-range OP strain-strain forces, is possible in three dimensions (3D). A 3D free energy expressed entirely in the order parameter strain variable(s), rather than the displacement field, provides a unified understanding of martensitic textures. We use 3D compatibility equations, linking the strain tensor components in the bulk and at interfaces, that induce anisotropic order-parameter strain interactions (masked in a conventional displacement-field treatment). These two long-range bulk and interface potentials, together with local compositional fluctuations, drive the formation of global elastic textures. In contrast to 2D, the rich microstructure is the result of frustration and competing metastable states in a complicated energy landscape. Here we will focus only on twinning for illustrative purposes. Other complex textures follow from our formalism and will be reported elsewhere. Our work is an essential step towards understanding results of High Resolution Electron Microscopy (HREM) and neutron scattering studies in alloys such as FePd and NiAl. In $`Fe_xPd_{1x}`$, (011) twins preferentially grow at the expense of (101) twins as the temperature is gradually lowered to the martensitic transition temperature. In $`Ni_xAl_{1x}`$, martensite forms with a long period stacking structure involving (110) planes. Due to the directional aspects of the long-range interactions encoded by the underlying symmetry of the lattice, our analysis predicts such twinning planes and directions.
Model. As a specific example, we examine a cubic to tetragonal transition. In this case the symmetry-adapted strains $`e_i`$ can be defined in terms of the Lagrangian strain tensor components $`ϵ_{ij}`$ as the dilatation $`e_1=(1/\sqrt{3})(ϵ_{11}+ϵ_{22}+ϵ_{33})`$, the deviatoric (i.e., OP) strains $`e_2=(1/\sqrt{2})(ϵ_{11}ϵ_{22})`$, $`e_3=(1/\sqrt{6})(ϵ_{11}+ϵ_{22}2ϵ_{33})`$ and the three shear strains $`e_4=2ϵ_{23}`$, $`e_5=2ϵ_{13}`$, $`e_6=2ϵ_{12}`$. The elastic energy in 3D is given by
$$F=F_L(e_2,e_3)+F_{grad}(e_2,e_3)+F_{cs}(e_1,e_4,e_5,e_6),$$
where the Landau free energy (Fig. 1), which provides the driving force for the transition, is
$$F_L=A(e_2^2+e_3^2)+Be_3(e_3^23e_2^2)+C(e_2^2+e_3^2)^2,$$
and the Ginzburg (or gradient) contribution, which is responsible for the interfacial energies, to the free energy is
$`F_{grad}`$ $`=`$ $`g[(e_{2,x}^2+e_{2,y}^2)+{\displaystyle \frac{1}{3}}(e_{3,x}^2+e_{3,y}^2)`$ (1)
$`+`$ $`{\displaystyle \frac{2}{\sqrt{3}}}(e_{2,x}e_{3,x}e_{2,y}e_{3,y})+{\displaystyle \frac{4}{3}}e_{3,z}^2]`$ (2)
$`+`$ $`h[(e_{3,x}^2+e_{3,y}^2)+{\displaystyle \frac{1}{3}}(e_{2,x}^2+e_{2,y}^2)`$ (3)
$``$ $`{\displaystyle \frac{2}{\sqrt{3}}}(e_{2,x}e_{3,x}e_{2,y}e_{3,y})+{\displaystyle \frac{4}{3}}e_{2,z}^2]`$ (4)
The harmonic elastic energy contribution due to non-OP compression-shear (CS) strain components is
$$F_{cs}=A_ce_1^2+A_s(e_4^2+e_5^2+e_6^2).$$
It is also possible to add a coupling, $`F_{compos}(\eta ,\eta ,e_2,e_3)`$, of the OP strain to compositional fluctuations $`\eta `$, which are responsible for tweed, but we do not consider such effects in the present paper.
The coefficients $`A=A_o(TT_o)`$, $`B`$ and $`C`$ are related to the second, third and fourth order elastic constants, respectively, and can be obtained from experimental structural data. The martensitic transition temperature is denoted $`T_o`$ and the gradient coefficients $`g`$ and $`h`$ are determined from the phonon dispersion at the appropriate high-symmetry point in the Brillouin zone of the material. $`A_c`$ and $`A_s`$ denote the bulk compression and shear moduli, respectively. Here we do not consider the gradient energy contribution from the non-OP components $`e_1`$, $`e_4`$, $`e_5`$, and $`e_6`$, since their contribution is of secondary nature compared to the terms reported above.
3D compatibility and analysis. We express the non-OP strains in terms of the OP strains ($`e_2`$, $`e_3`$) using the three compatibility constraints $`ϵ_{12,12}=ϵ_{11,22}+ϵ_{22,11}`$, $`ϵ_{23,23}=ϵ_{22,33}+ϵ_{33,22}`$, $`ϵ_{31,31}=ϵ_{33,11}+ϵ_{11,33}`$. Analogous to the 2D case the constrained minimization of the compression-shear energy leads to an anisotropic long-range interaction between the OP strains in the bulk:
$$F_{cs}^{bulk}=d^3re_\alpha (\alpha )U_{\alpha \alpha ^{}}(|rr^{}|)e_\alpha ^{}(r^{})(\alpha ,\alpha ^{}=2,3),$$
where the kernels $`U_{22}`$, $`U_{23}`$, and $`U_{33}`$ can be expressed explicitly in Fourier space with each of the three kernels becomes a ratio of algebraic combinations of $`k_x`$, $`k_y`$ and $`k_z`$. All these combinations have the same order in the length, $`k`$, of the wavevector $`\stackrel{}{k}`$. Hence, the kernels do not have any scale dependence. That is, the kernels only depend on the angles $`\varphi `$ and $`\theta `$, defined in terms of spherical coordinates. The minima of each kernel occur at certain $`\theta `$ and $`\varphi `$ values. These angles vary from kernel to kernel and the system, in general, experiences frustration. Figure 2 shows contours of the kernel $`U_{22}`$ and $`U_{33}`$ as functions of the two angles $`\theta `$ and $`\varphi `$. The minima define a direction in 3D in which the compression-shear energy arising from the particular kernel is minimal.
In contrast to 2D, $`F_{cs}`$ can be viewed as a weighted sum of the three kernels, where the fields $`e_2`$ and $`e_3`$ act as the weights. There is thus an interplay between $`F_L`$ and the directional dependence of $`F_{cs}`$ to accommodate the frustration and choose minima, which in general do not have zero energy. Thus, there exist metastable minima corresponding to different microstructures. In 2D there is no such interplay, as the directional minimum is independent of the amplitude of the field: There is only one minimum corresponding to zero energy and hence there is no frustration. Moreover, these long-range kernels in 3D fulfill the full cubic symmetry properties associated with the transformation. In particular, $`F_{cs}`$ is invariant with respect to two-fold, three-fold, and four-fold rotation symmetries of the cube. We do not present here the mathematical structure underpinning this, but it is reflected in the frustration seen in 3D that is lacking in 2D.
The bulk contribution comes from a solid with periodic boundaries while the the effects of interfaces/surfaces such as habit planes (i.e., austenite-martensite or parent-product boundaries) may be included through a free energy contribution, $`F_{cs}^{surf}`$, where $`F_{cs}=F_{cs}^{bulk}+F_{cs}^{surf}`$ and
$$F_{cs}^{surf}=d^3k\mathrm{\Gamma }(k_x,k_y,k_z)[e_2(k)^2+e_3(k)^2].$$
In analogy with the 2D case the function $`\mathrm{\Gamma }(\stackrel{}{k})`$ has the form
$$\mathrm{\Gamma }(\stackrel{}{k})=\nu \left[\frac{1}{|k_x\pm k_y|}+\frac{1}{|k_y\pm k_z|}+\frac{1}{|k_z\pm k_x|}\right]$$
for the family of habit planes.
We remark that the physics of twinning is quite different in 2D and 3D. In 2D both the gradient and the bulk compatibility terms vary as $`1/r^2`$ and thus do not lead to a length scale. However, the surface compatibility term in 2D is essential for introducing a length scale. It varies as $`1/r`$ which then competes with the combined bulk compatibility and gradient terms to give rise to a twinning width. In 3D, the gradient terms behave as $`1/r^2`$ and $`F_{cs}^{bulk}1/r^3`$. Thus, there is already a length scale present and $`F_{cs}^{surf}`$ merely renormalizes the length scale arising from the bulk compatibility and gradient terms. We have numerically verified the above scaling relations. On the basics of the above arguments we expect the twin width to scale as the square-root of the length in 3D to hold on the basis of above arguments.
Texture simulations. The various martensitic textures that realize minima of the energy are found from random initial conditions and relaxational dynamics for the OP strains. That is, $`\dot{e}_\alpha (r)=\frac{\delta F(e_\alpha ,e_1(e_\alpha ),e_4(e_\alpha ),e_5(e_\alpha ),e_6(e_\alpha ))}{\delta e_\alpha (r)},`$ where the time $`t`$ is scaled with a characteristic relaxation rate and $`\alpha =2,3`$. Depending on the details of the initial configurations, various twinning textures reflecting the metastability emerge as $`\dot{e}_\alpha `$ vanishes. The simulations are performed for a regular cube, and periodic boundary conditions are applied in all six directions to obtain fully relaxed textures in $`e_2(r)`$ and $`e_3(r)`$.
We illustrate the simulation results obtained with representative (scaled) parameters. The red/blue/green color represent ($`e_2,e_3`$) positive/negative/zero OP strain values, respectively (the actual values are approximately determined by the minima of $`F_L`$ as illustrated in Fig.1).
Figure 3 shows the simplest twinning texture ((110) twins) that is uniform in the $`z`$ direction. The strain $`e_2`$ changes from negative (minimum I in Fig. 1) to positive (minimum II in Fig. 1) (i.e. blue to red). The strain $`e_3`$ is not shown, as it is essentially constant with a small modulation at the twin boundaries. This particular martensitic texture is the trivial extension of the 2D results since the OP fields are uniform in the $`z`$ directions. Notice also that the directional dependence is the same as observed in 2D studies .
A richer texture is shown in Fig. 4, where the strain $`e_3`$ changes from negative (minimum III in Fig. 1) to positive (minimum I or II in Fig. 1) (i.e. blue to red). The strain $`e_2`$ would either be zero (green) when $`e_3`$ is negative (minimum III) or positive/negative when $`e_3`$ is positive. The figure clearly shows that twins with strain $`e_2`$ can be oriented in two possible ways. This reflects the inversion symmetry of the kernels with respect to $`x`$, $`y`$. We note that a similar alternating twin microstructure has been implied by neutron and x-ray scattering studies in layered high-temperature superconductors .
To illustrate the of frustration that is inherent in the compatibility relations, we explain the orientation of the twins in Fig. 4. We construct an effective kernel for the red and blue twins, respectively, using the appropriate values of the strains $`e_2`$ and $`e_3`$ in the twins. Figure 5 depicts the minima of these effective kernels in the $`\varphi `$-$`\theta `$ plane. The dashed line shows the minimum corresponding to the blue twins and the solid line the minimum for the red twins. We note that the minima correspond to different $`\varphi `$ angles ($`45+\delta `$ and $`45\delta `$). The shift $`\delta `$ depends on the parameters of the Landau free energy. Ideally, the red and the blue twins would not run parallel; however, the system accommodates these two conflicting directions by choosing the average angle of 45 degrees. This aspect of competing metastable states is a novel feature associated with the cubic symmetry kernels that leads to a rich landscape of microstructures.
Conclusion. We have derived a complete symmetry based, fully 3D model that describes the cubic to tetragonal structural transitions observed in many martensitic and shape-memory alloys. We obtained analytically the compatibility-induced anisotropic long-range potential in the OP (deviatoric strain tensor components) in the bulk and at interfaces. Unlike the 2D case , there is a length scale in the system from bulk compatibility alone; the surface compatibility potential also introduces a length scale akin to 2D. Many twin orientations are possible in 3D as a result of elastic frustration and the “landscape” (probably “rugged”) of metastable energy states. Our model can be generalized straightforwardly to improper ferroelastics transitions by including symmetry allowed polarization (magnetization, etc.) nonlinear terms, and couplings to strain. Other symmetries can be handled within our approach. For example, we can study the cubic to trigonal (rhombohedral) transition in lead orthovanade, and NiTi- and AuCd-based shape-memory alloys using the three shear strains $`e_4`$, $`e_5`$, and $`e_6`$ as OP and with $`e_1`$, $`e_2`$, and $`e_3`$ expressed in terms of the shear strains via compatibility. Image reconstruction methods of systematic experimental data from HREM and neutron scattering will allow a more quantitative comparison of strain patterns with those obtained from our model.
We wish to thank G. R. Barsch, I. Mitkov, and S. R. Shenoy for insightful discussions. This work was supported by the U.S. Department of Energy (Contract number W-7405-Eng-36). This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098. |
warning/0001/astro-ph0001450.html | ar5iv | text | # Frequency resolved spectroscopy of Cyg X-1: fast variability of the reflected emission in the soft state.
## 1 Introduction
The importance of the reprocessed/reflected component in the X–ray spectra of accreting X–ray sources for exploring the geometry of the accretion flow is well known. Reflection of the primary Comptonized radiation from neutral or partially ionized matter located in the vicinity of the compact object – presumably the optically thick accretion disk, leads to appearance of characteristic features in the spectra of X–ray binaries. The main signatures of the emission reflected from cold neutral medium are well known – the fluorescent K<sub>α</sub> line of iron at 6.4 keV, iron K-edge at 7.1 keV and a broad hump at $`2030`$ keV (Basko, Sunyaev & Titarchuk 1974, George & Fabian 1991). The exact shape of these spectral features in the X-ray binaries depends on ionization state of the reflecting medium and might be modified by strong gravity effects and intrinsic motions in the reflector (e.g. Fabian et al., 1989). The amplitude of the reflection signatures depends primarily on the ionization state and the solid angle subtended by the reflector as seen from the source of the primary radiation.
Recently, Revnivtsev, Gilfanov & Churazov (1999, hereafter Paper I) proposed Fourier frequency resolved spectral analysis to study spectral variability in the X-ray binaries. Although interpretation of the Fourier frequency resolved spectra in general is not straightforward and requires a priori assumptions to be made, one of the areas where this method can be efficiently used is the fast variability of the reflected component. It has been found that in the low spectral state of Cyg X-1 and GX339-4 the energy spectra corresponding to the shorter time scales ($`0.11`$ sec) show less reflection than those of longer time scales (Revnivtsev, Gilfanov & Churazov 1999, 2000). The simplest, although not unique, interpretation of this result is smearing of the short term variations of the reflected emission due to finite light crossing time of the reflector. Based on the Fourier frequency dependence of the equivalent width of the Fe K<sub>α</sub> fluorescent line Revnivtsev et al. (1999) estimated the characteristic size of the reflector: $`80160R_g`$ for a $`2010M_{\mathrm{}}`$ black hole.
In this paper we investigate the fast variability of the reflected component in the soft spectral state of Cyg X-1 and compare it with that in the hard spectral state.
## 2 Observations and data analysis.
We used publicly available data of Cyg X–1 observations with the Proportional Counter Array aboard the Rossi X-ray Timing Explorer \[Brandt Rotschild & Swank 1996\] performed in June, 1996 during the soft spectral state of the source. The list of the observations is given in Table 1. The total live time was $`11.5`$ ksec. The “Generic Binned” mode data in configuration B\_4ms\_8A\_0\_35\_H, with time resolution of $`4`$ msec and covering 2.9-13.1 keV energy range was used for frequency resolved spectral analysis.
The data screening and selection was performed using FTOOLS 4.2 with standard screening criteria recommended by RXTE GOF. The frequency resolved spectra were calculated following the prescription detailed in Paper I. The dead time corrected value of the white noise level was determined from fitting of the power spectra in the 300-1000 Hz frequency range. The spectral analysis was performed in XSPEC v.10.0 \[Arnaud 1996\] with version 3.5 of the PCA response matrix. A uniform systematic error of 0.5% was added quadratically to the statistical error in each energy channel.
The observations of the source during the hard state discussed in the text were performed between March 26 and 31, 1996. The details of these observations are given in Paper I.
## 3 Results.
The broad band energy spectra of Cyg X-1 in the hard and soft spectral state are shown in Fig.1. The spectra were obtained using the data of overlapping ASCA and RXTE observations of the source on March 26, 1996 (hard state) and on May 30, 1996 (beginning of the soft state).
The power spectra of Cyg X-1 in the 3–13 keV energy band in the hard and soft spectral states obtained from the complete sets of the data used for the frequency resolved spectral analysis in this paper and in Paper I are shown in Fig.2. The power density is plotted in units of power$`\times `$frequency representing squared fractional rms at a given frequency per factor $`e`$ in frequency. This way of representing the power spectra most clearly characterizes relative contribution of variations at different frequencies to the total observed rms.
The Fourier frequency resolved spectra for the soft spectral state were obtained in 10 frequency bins of logarithmically equal width in the 0.016-128 Hz frequency range. In order to study the frequency dependence of the amplitude of the reflected component we fit the spectra in the 3–13 keV band with a simplified model consisting of an absorbed power law<sup>1</sup><sup>1</sup>1Contrary to the average spectra of Cyg X-1 in the soft state (Fig.1), the contribution of the soft component to the frequency resolved spectra is negligible (to be discussed in more detail in a separate paper). Therefore use of a power law to model continuum emission in the soft state is justified. with superimposed Gaussian line at 6.4 keV. The low energy absorption was fixed at $`N_H=610^{21}`$ cm<sup>-1</sup>. The line width (standard deviation for a Gaussian profile) was fixed at 1 keV which corresponds to the average value for the high state observations used for the analysis. The power law photon index and line flux were the only free parameters of the fit. To facilitate comparison with the hard state data we reanalyzed the data set used in Paper I in identical frequency bins covering 0.016-32 Hz frequency range (the time resolution of the hard state data was $`16`$ msec) and using the same spectral model. The width of the line for the hard state spectra was fixed at the corresponding average value of 0.8 keV \[Gilfanov, Revnivtsev & Churazov 1999\].
Such a spectral model is obviously oversimplified. Neither it is justified from the physical point of view. It is, however, suitable to quantify the amplitude of the characteristic “wiggle” usually seen in the spectra of the accreting X–ray sources in the $`515`$ keV energy range and commonly attributed to the effects of reflection. Use of more sophisticated models is restricted by insufficient statistics (especially in the soft spectral state) and the low number of energy channels below $`13`$ keV in the B\_4ms\_8A\_0\_35\_H configuration used in the most of the soft state observations.
The best fit values of the spectral parameters for both spectral states are given in Table 2. The dependence of the equivalent width on the Fourier frequency is shown in Fig.3. Variations of the parameters of the spectral model, in particular the change of the line centroid from 6.0 to 6.7 keV and reducing the intrinsic line width to zero, do not change the general trend. These variations, however, affect the particular values of the equivalent width and, to lesser extent, the shape of the curve in Fig. 3.
According to the $`\chi ^2`$–test two distributions, shown in Fig.3 differ at the confidence of level of $`98.7\%`$ ($`\chi ^2=19.4`$ for 8 d.o.f. in the 0.016–32 Hz frequency range). It should be noted, however, that the errorbars assigned to the Fourier frequency resolved spectra were propagated from the corresponding power density spectra and are likely to be somewhat overestimated, especially in the low frequency bins (this fact can be noticed in Fig.3). The confidence level, calculated using 0.45–32 Hz frequency range is $`99.7\%`$ ($`\chi ^2=18.1`$ for 5 d.o.f.).
## 4 Frequency dependence of the equivalent width and time response of the reflector.
The geometry, element abundances and ionization state of the reflector being fixed, the equivalent width of the Fe fluorescent line determined from a conventional energy spectrum is proportional to the relative amplitude of the reflected component and approximately measures the solid angle subtended by the reflector. The equivalent width of the fluorescent line determined from a Fourier frequency resolved spectrum measures the ratio of the rms amplitudes of variations of the reflected component and the primary emission in a given Fourier frequency range. The constancy of the equivalent width of the line at Fourier frequencies $`f30`$ Hz observed in the soft state (Fig.3) implies that the rms amplitude of variations of the reflected component has the same frequency dependence as that of the primary radiation. This would naturally appear if the reflected emission was reproducing variations of the primary radiation down to the time scales of $`3050`$ msec. Such behavior is in contrast to the hard spectral state, in which variations of the reflected flux are notably suppressed in comparison with the primary emission on the time scales shorter than $`500`$ msec.
The most straightforward explanation of this effect would be in terms of a finite light crossing time of the reflector $`\tau _{\mathrm{refl}}l_{\mathrm{refl}}/c`$ due to its finite spatial extend $`l_{\mathrm{refl}}`$. In this case the frequency dependence of the fluorescent line equivalent width, $`EW(f)`$, is determined by the geometry of the primary source and the reflector. The characteristic width of the $`EW(f)`$ is mainly defined by the spatial extent of the part of the accretion disk giving the main contribution to the reflected emission. Based on these arguments Revnivtsev, Gilfanov & Churazov (1999) estimated the characteristic size of the reflector in the low spectral state as $`l_{\mathrm{refl}}5\times 10^8`$ cm which would correspond to $`150R_\mathrm{g}`$ for a $`10M_{\mathrm{}}`$ black hole.
Below we consider this problem in a more quantitative way. The time dependence of the reflected emission is defined by the following relation:
$$F_{refl}(t)=\underset{0}{\overset{\mathrm{}}{}}F_0(t\tau )T(\tau )𝑑\tau $$
where $`F_0(t)`$ and $`F_{refl}(t)`$ are primary and reflected flux, $`T(\tau )`$ – the transfer function of the reflector, defined by the geometry. This function accounts for the propagation time of the photons from the primary source to different parts of the reflector and then to the observer. The Fourier transform of the reflected flux is:
$$\widehat{F}_{refl}(f)=\widehat{F}_0(f)\times \widehat{T}(f)$$
where $`\widehat{F}_0`$, $`\widehat{F}_{refl}`$ and $`\widehat{T}`$ are corresponding Fourier transforms and $`f`$ is Fourier frequency. The equivalent width of the fluorescent line determined from the Fourier frequency resolved spectra is
$$EW(f)\frac{|\widehat{F}_{refl}(f)|}{|\widehat{F}_0(f)|}=|\widehat{T}(f)|$$
i.e. is proportional to the Fourier transform of the transfer function of the reflector.
Fig.4 shows the transfer function and its Fourier transform for an isotropic point source located at the height $`h`$ above a flat disk with the inner radius $`R_{in}`$ and inclination angle $`i`$ for different values of $`R_{in}`$ and $`i`$. A Lambert law for the angular dependence of the reflected flux has been assumed. No general or special relativity effects have been taken into account. The characteristic width of the $`EW(f)`$ dependence is mainly defined by the distance $`d=\sqrt{h^2+R_{in}^2}`$ between the primary source and the inner edge of the disk and depends only weakly on the inclination angle $`i`$. A small offset of the primary source, $`\mathrm{\Delta }d`$ or non-zero opening angle of the disk do not significantly affect the characteristic width of Fourier transform of the response function (Fig.5). However, the transfer function itself and the high frequency part of its Fourier transform are sensitive to the details of the geometry (Fig. 4 and 5).
In Fig.6 we compare Fourier transform of the transfer function of a flat disk with inclination $`i=50\mathrm{°}`$, appropriate for Cyg X-1 \[Gies & Bolton 1986\], with the frequency dependence of the equivalent width observed in the soft and hard spectral states of Cyg X-1. As seen from Fig.6, suppression of the high frequency variations in the reflected emission observed in the hard state can be satisfactorily described by reflection from a disk with an inner radius of $`R_{in}100R_g`$ around a $`10M_{\mathrm{}}`$ black hole. Significantly larger, $`R_{in}1000R_g`$, or smaller values of the inner radius, $`R_{in}10R_g`$, are inconsistent with the data. The soft state data, on the other hand, requires much smaller values of the inner radius of the disk, $`R_{in}10R_g`$.
It should be noted that since we use the equivalent width of the Fe fluorescent line as a measure of the amplitude of the reflected emission, the results might be somewhat affected by the non-uniformity of the ionization state of the accretion disk with radius and geometrical effects (e.g. radial and azimuthal dependence of the reflection angle). Results of more detailed modeling of the disk transfer function and rigorous comparison with the data will be published elsewhere (a paper in preparation).
## 5 Discussion.
Study of fast variability of the reflected emission by means of Fourier frequency resolved spectroscopy offers a new independent method to probe the geometry of the accretion flow which complements conventional X-ray spectral analysis. The presence of a luminous soft component dominating the X-ray spectrum in the soft spectral state of black hole candidates (Fig.1) strongly favors small values of the inner radius of the disk (e.g. Gierlinski et al., 1999), in good agreement with the above result, $`R_{in}10R_g`$. In the hard spectral state conventional estimates of the inner radius of the accretion disk range from several tens $`R_g`$ in coronal model (e.g. Poutanen, Krolik & Ryde, 1997; Done & Zicky, 1999) to several hundred $`R_g`$ in ADAF model (e.g. Esin et al., 1998). Our result, $`R_{in}100R_g`$, falls in the middle of this range.
The inner radius of the accretion disk determined in the above analysis refers to the inner radius of the “reflective” part of the disk, where the ionization state is such that the disk is capable to produce a fluorescent iron line. Therefore substantial change of the ionization state of the surface layer of the inner disk (e.g. Young et al., 1999) may have similar effect on the frequency dependence of the equivalent width of the iron line as physical change of the inner disk radius.
In the simplified geometry of an isotropic point source above flat disk solid angle subtended by the reflector is $`\frac{\mathrm{\Omega }}{2\pi }=\frac{1}{\sqrt{1+R_{in}^2/h^2}}`$ (in the notation of the previous section). For the parameters from Fig.6 respective values of the solid angle are: $`\frac{\mathrm{\Omega }}{2\pi }0.1`$ for the hard state ($`h=10R_g`$, $`R_{in}=100R_g`$) and $`\frac{\mathrm{\Omega }}{2\pi }0.7`$ for the soft state ($`h=10R_g`$, $`R_{in}=10R_g`$). Equivalent width of the iron fluorescent line expected for these values of the solid angle and solar abundance of iron are several times smaller than those given in the Table 2, especially for the hard spectral state. However, as was noted above, the absolute values of the equivalent width quoted in Table 2 are subject to some uncertainty due to simplified model used for the spectral fits. Comparison of the model predictions with more accurately determined values of the equivalent width and amplitude of the reflected component can further constrain geometry of the accretion flow.
Finally, we should note that interpretation of the frequency dependence of the equivalent width of the fluorescent line in terms of the finite light crossing time of the reflector is not unique. An alternative explanation might be that the short time scale, $``$ 50–100 msec, variations appear in a geometrically different, likely inner, part of the accretion flow and give a rise to significantly weaker, if any, reflected emission than the longer time scale events presumably originating in the outer regions. This might be caused, for instance, by a smaller solid angle of the reflector as seen by the short time scale events and/or due to screening of the reflector from the short time scale events by the outer parts of the accretion flow. However, independent of the nature of the fall off of the equivalent width at high frequency, the flat response of the reflected emission to variations of the primary radiation observed at low frequencies puts an upper limit on the spatial extent of the reflector, i.e. on the inner radius of the accretion disk. In particular, large values of $`R_{in}`$, significantly exceeding $`100R_g`$ in the hard and $`10R_g`$ in the soft spectral state are excluded by our analysis.
## 6 Conclusions.
We have exploited Fourier frequency–resolved spectral analysis to study fast variability of the reflected emission on time scales of $`100`$ sec – 10 msec in the soft and hard spectral states of Cyg X-1. Our conclusions are:
1. In the soft spectral state variations of the reflected component have the same frequency dependence of the rms amplitude as the primary emission up to the frequencies $`30`$ Hz. This would be expected if, for instance, the reflected flux was reproducing, with flat response, variations of the primary radiation down to the time scales of $`3050`$ msec. The sensitivity of the present analysis is insufficient to study shorter time scales.
2. In the hard spectral state variability of the reflected flux is significantly suppressed in comparison with the direct emission on the time scales shorter than $`0.51`$ sec (see also Paper I).
3. Assuming that suppression of the short-term variability of the reflected emission is caused by the finite light-crossing time of the reflector, we estimated the inner radius of the accretion disk $`R_{in}100R_g`$ in the hard spectral state and $`R_{in}10R_g`$ in the soft spectral state.
## acknowledgments
This research has made use of data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center. The work was done in the context of the research network ”Accretion onto black holes, compact objects and protostars” (TMR Grant ERB-FMRX-CT98-0195 of the European Commission). M.Revnivtsev acknowledges partial support by RBRF grant 97-02-16264 and INTAS grant 93–3364–exit. |
warning/0001/astro-ph0001200.html | ar5iv | text | # APM 08279+5255: Keck Near- and Mid-IR High-Resolution Imaging Based on observations obtained at the W. M. Keck Observatory which is operated jointly by the California Institute of Technology, the University of California, and NASA.
## 1 INTRODUCTION
In the course of a survey for distant cool carbon stars in the Galactic halo, Irwin et al. (1998) serendipitously discovered an extremely luminous broad absorption line QSO at $`z=3.91`$, APM 08279+5255.<sup>1</sup><sup>1</sup>1The redshift obtained by Irwin et al. (1998) was 3.87, but the CO line observation by Downes et al. (1999) later showed that the redshift of this system is 3.9110, and that the absorption lines measured by Irwin et al. (1998) are probably blueshifted due to the gas outflow in this object. To avoid confusion, we adopt a redshift of 3.91 throughout this paper. Its phenomenally large apparent luminosity is immediately clear from its observed magnitude: the observed $`R`$ magnitude of 15.2 mag corresponds to an absolute magnitude of $`M_R=33.2`$ mag after a K correction. Furthermore, this object was also detected at 25, 60, and 100 $`\mu `$m in the IRAS Faint Source Catalog with flux densities of 0.23, 0.51, and 0.95 Jy. All together, the apparent bolometric luminosity of this object reaches an unprecedented level of $`5\times 10^{15}`$ L, which would make APM 08279+5255 the most luminous object known in the universe<sup>2</sup><sup>2</sup>2Throughout this paper, we adopt the values of H$`{}_{0}{}^{}=50`$ km s<sup>-1</sup> Mpc<sup>-1</sup> and q$`{}_{0}{}^{}=0.5`$..
The observations to date indicate that APM 08279+5255 is likely to be gravitationally lensed. Irwin et al. (1998) showed that the $`R`$-band image of this object, taken under a 0$`\stackrel{}{\mathrm{.}}`$9 seeing, is slightly elongated, and that the image likely consists of two point sources separated by 0$`\stackrel{}{\mathrm{.}}`$3–0$`\stackrel{}{\mathrm{.}}`$45 with a flux ratio of 1.05–1.15, probably being two gravitationally lensed images of the same QSO. Subsequently, Ledoux et al. (1998) performed adaptive-optics observations in the $`H`$ band, which achieved a spatial resolution of 0$`\stackrel{}{\mathrm{.}}`$3. Their image shows two point sources separated by 0$`\stackrel{}{\mathrm{.}}`$35$`\pm `$0$`\stackrel{}{\mathrm{.}}`$02 with a flux ratio of 1.21$`\pm `$0.25, in good agreement with the values derived by Irwin et al. (1998). In addition, by using an optical integral-field spectrograph, Ledoux et al. (1998) showed that the optical spectra of these two sources are quite similar, strengthening the gravitational lensing hypothesis for this object.
At the same time, the far-IR and submillimeter observations have shown that APM 08279+5255 contains a large amount of gas and dust, making this object an extreme example of a hyper/ultra-luminous IR galaxy/QSO. In addition to the IRAS far-IR detections, its continuum was also detected in the submillimeter by Lewis et al (1998). The black-body dust temperature was estimated from the submillimeter SED to be 220 K, which in turn results in a dust mass with no magnification correction of $`3.7\times 10^9`$ M (Lewis et al. (1998)). Later, Downes et al. (1999) detected two CO lines (4–3 and 9–8) by millimeter interferometry, and derived the molecular gas temperature of $`200`$ K from the line ratio. The mass of molecular gas was calculated to be $`16\times 10^9`$ M with a magnification factor of 7–20. From a simultaneous millimeter continuum observation, they also derived the dust mass of $`17\times 10^7`$ M with a magnification factor of 7–30. The detection of the CO (9–8) line is especially important because this is direct evidence that this object contains hot and dense molecular gas.
APM 08279+5255 was also found to have a significant visual linear polarization ($`p>1`$ %, Hines et al. 1999). Hines et al. also suggest that the broad absorption trough formerly identified as a $`z=3.07`$ damped Ly-$`\alpha `$ absorption system by Irwin et al. (1998) may be a O iv/Ly$`\beta `$ broad absorption line intrinsic to the QSO because there is an increase of polarization in the trough.
From the currently available data, APM 08279+5255 looks very much like the hyperluminous IR QSO H1413+117 (Cloverleaf QSO) and the warm ultraluminous IR galaxy Mrk 231, in the sense that it contains a clearly visible QSO nucleus surrounded by a large amount of gas and dust. The major difference between these two objects and APM 08279+5255 seems to be the exceptionally large luminosity of the latter, which is likely the effects of gravitational lensing. In these systems, the existence of a powerful QSO nucleus seems to indicate that the QSO is the dominant luminosity source, generating the large IR luminosity by heating dust.
Here, we present high-resolution images of APM 08279+5255 in the near- and mid-IR taken with the Keck telescope. Our high-resolution (FWHM $``$ 0$`\stackrel{}{\mathrm{.}}`$15) $`K`$-band image shows a third component between the two components previously detected. We argue that it is another lensed image of the same background QSO, and examine the intrinsic properties of the lensed QSO based on the lens model. We especially try to constrain the effects of differential magnification since they could introduce a significant distortion in the observed spectral energy distribution (SED) as shown by Blain (1999).
After the initial submission of this paper, we were informed by Drs. Ibata and Lewis that a similar work was going to be submitted based on the analysis of HST/NICMOS images of APM 08279+5255 (Ibata 1999). The two sets of observational data are in excellent agreement. Both studies (1) detect a third image, (2) derive similar positions and relative brightnesses for all three images, and (3) conclude that the third image is likely to be another lensed image of the same background QSO and construct a similar lens model based on this assumption. Although our work was carried out independently of their work, we have opted to incorporate their results whenever they provide critical pieces of information to understand this object.
## 2 OBSERVATIONS AND DATA REDUCTION
### 2.1 Near-IR Imaging
The $`K`$-band images of APM 08279+5255 were taken on the night of UT 1998 October 3 with the Near Infra-Red Camera (NIRC; Matthews & Soifer (1994)) on the Keck I telescope on Mauna Kea in Hawaii. NIRC uses a Hughes-SBRC 256 $`\times `$ 256 InSb array as the detector, and is attached to the f/25 forward Cassegrain focus of the telescope, producing a pixel scale of 0$`\stackrel{}{\mathrm{.}}`$15/pixel with a field of view of 38″ on a side. For this observation, the NIRC image converter (Matthews et al. (1996)) was used with NIRC to achieve a high spatial resolution. The image converter changes the beam from f/25 to f/180, producing a pixel scale of 0$`\stackrel{}{\mathrm{.}}`$0206/pixel with a field of view of 5$`\stackrel{}{\mathrm{.}}`$3 on a side. For $`K`$-band imaging, the integration time was 10 seconds per image. The highest resolution image was produced by combining the four images (40 seconds total) with the minimum image size (i.e., selective shift & add). The extremely good seeing that night resulted in a FWHM of 0$`\stackrel{}{\mathrm{.}}`$15 in this $`K`$-band image.
### 2.2 Mid-IR Imaging
The 12.5 $`\mu `$m and 17.9 $`\mu `$m images were initially taken on the night of UT 1998 October 3 using the MIRLIN mid-IR camera (Ressler et al. (1994)) on the Keck II telescope. MIRLIN uses a Boeing 128 $`\times `$ 128 Si:As array, and is attached to the f/40 bent Cassegrain visitor port of the telescope, producing a pixel scale of 0$`\stackrel{}{\mathrm{.}}`$138/pixel with a field of view of 17″on a side. The total integration time was 15 minutes at each wavelength. These images were used to measure the mid-IR flux density of APM 08279+5255. Calibration was done by observing bright stars tied to the IRAS calibration at 12 and 25 $`\mu `$m (IRAS Explanatory Supplement 1988).
The higher spatial-resolution 12.5 $`\mu `$m image of APM 08279+5255 was taken on the night of UT 1999 November 25 using the Long Wavelength Spectrometer (LWS; Jones et al. (1993)) on the Keck I telescope in the imaging mode. LWS uses a Boeing 128 $`\times `$ 128 Si:As array, and is attached to the f/25 forward Cassegrain focus of the telescope, producing a pixel scale of 0$`\stackrel{}{\mathrm{.}}`$08/pixel with a field of view of 10$`\stackrel{}{\mathrm{.}}`$2 on a side. The seeing was 0$`\stackrel{}{\mathrm{.}}`$4 FWHM, and the total integration time was 27 minutes.
### 2.3 Near-IR Photometry
Near-IR photometry was obtained on UT 1998 November 3 with the InSb camera on the 200-inch Telescope at Palomar Observatory. Calibration was done by measuring stars listed by Elias et al. (1982).
The magnitudes and flux densities reported here are listed in Table 1. All the photometry was done with a 4″-diameter beam.
## 3 RESULTS
### 3.1 The $`𝑲`$-band Image
The $`K`$-band grey-scale image with scale 0$`\stackrel{}{\mathrm{.}}`$0206/pixel is shown in Figure 1a with the contour map of the same data in Figure 1b. It is immediately clear from this image that the $`K`$-band image consists of two bright components, which is consistent with the previous observations by Irwin et al. (1998) and Ledoux et al. (1998). The northern brighter source is slightly extended to the south-west. Since the southern source is quite circularly symmetric, this suggests that the extension in the northern source is a third fainter component.
We performed a point-spread-function (PSF) fitting using DAOPHOT (Stetson (1987)) on the $`K`$-band image. First, it was assumed that the two main bright components are basically point sources, and that we could construct the PSF by using the upper half of the upper component and the lower half of the lower component. Two axisymmetric PSFs were first produced from two half PSFs by self-reflection, and their average was taken to produce the final PSF. In other words, we assumed that the PSF is symmetric with respect to the horizontal axis in Figure 1.
To determine the position of the third component accurately enough as the starting point for the PSF fitting, we first performed the two-PSF subtraction. The position of the third component was determined in the residual image, and with this added information, we then performed the three-PSF subtraction.
The three components we have finally found are shown in Figure 2a and 2b together with their contour maps Figure 2c and 2d. We refer to these components as A, B, and C in descending order of brightness. The positions and relative brightnesses determined from the PSF fitting for each component are listed in Table 2. Using the flux ratios, the $`K`$ magnitudes of individual components were derived from magnitude of $`K=12.08`$ mag for the total system. The derived FWHMs indicate that the seeing was 0$`\stackrel{}{\mathrm{.}}`$15 when the image was taken. The FWHM of the component C indicates that this component is also point-like. For our further discussions, we assume that the $`K`$-band image of APM 08279+5255 consists of three point sources.
Overall our measured positions and relative brightnesses of the three components are in good agreement with the values derived by Ibata et al. (1999).
### 3.2 The 12.5 $`𝝁`$m Image
The LWS 12.5 $`\mu `$m image of APM 08279+5255 is shown in Figure 3a. Figure 3b is an artificial image showing how the three components detected in the $`K`$ band would look like if observed with the LWS pixel scale (0$`\stackrel{}{\mathrm{.}}`$08/pixel) under the same seeing condition (FWHM $`=`$ 0$`\stackrel{}{\mathrm{.}}`$4). This image was simulated using the parameters listed in Table 2. Contour maps of the images are also shown in Figure 3c and 3d, respectively.
The morphological resemblance is clear. Although there might be a small difference in morphology between the two images, we cannot say with confidence that it is real. The 12.5 $`\mu `$m image shape changes from image to image considerably because of the lower signal-to-noise ratio, and the image shape might have been smeared when a large number of images were combined. Therefore, based on the overall ellipticity of the image contours, we conclude that the image configuration at 12.5 $`\mu `$m is similar to that in the $`K`$ band.
## 4 DISCUSSION
### 4.1 The Lens Model
To construct a gravitational lens model, we use the elliptical effective lensing potential $`\psi `$ in the following form (Blandford & Kochanek 1987; Narayan & Bartelmann 1999):
$$\psi (\theta _1,\theta _2)=\theta _E[\theta _c^2+(1\epsilon )\theta _1^2+(1+\epsilon )\theta _2^2]^{1/2},$$
(1)
where $`\theta _E`$ is the Einstein radius, $`\theta _c`$ and $`\epsilon `$ are the core radius and ellipticity of the lensing potential, and $`\theta _1`$ and $`\theta _2`$ are rectangular coordinates in radians with respect to an arbitrarily defined optic axis. Although we treat the Einstein radius itself as a free parameter, it can also be expressed as,
$$\theta _E=\frac{D_{ls}}{D_s}4\pi \frac{\sigma _v^2}{c^2},$$
(2)
where $`D_{ls}`$ is the standard angular distance between the lens and the source (cf., Fukugita et al. (1992)), $`D_s`$ is the angular distance from the observer to the background source, $`\sigma _v`$ is the internal velocity dispersion of the lensing galaxy (i.e., the mass), and $`c`$ is the speed of light. For a given Einstein radius, this equation sets the constraints on the mass and redshift of the lens. In addition to these parameters specifying the shape of the potential, the center position ($`x_l`$,$`y_l`$) and the position angle ($`\gamma `$) of the lensing potential must be specified. All together, the model contains six unknown parameters ($`\theta _E`$, $`\theta _c`$, $`\epsilon `$, $`\gamma `$, $`x_l`$ and $`y_l`$).
The major uncertainty with APM 08279+5255 is the nature of component C. There are two possibilities: it is either a third lensed image of the same background source or the lensing galaxy. We will construct models based on both these assumptions, and evaluate their validity based on the available observational data.
#### 4.1.1 Three-Image Model
Three lensed images provide six constraints: two relative brightnesses and four coordinates giving two relative image positions with respect to the other. Since the number of model parameters is also six, we can determine the values of the parameters, but cannot assess the goodness of the fit. The best-fit model was searched by taking component B as the reference and varying the parameters such that the source positions of components A and C coincide with that of B while the derived relative magnifications approach the observed flux ratios. Our best-fit model parameters are shown in Table 3. Figure 4 shows the profile of the effective lensing potential, the time-delay surface, and the expected image positions. Basically, this model reproduces the positions and relative brightnesses of three images well within the observational uncertainties. The lens is almost round ($`\epsilon 0.01`$) and has a large core radius ($`\theta _c0\stackrel{}{\mathrm{.}}2`$). This is because the three images are almost in a straight line (i.e., small $`\epsilon `$) and the third image is very bright (i.e., large $`\theta _c`$). The total magnification factor for a point source<sup>3</sup><sup>3</sup>3We originally derived a value of 71, which was quoted by Ibata et al. (1999), but our later calculation increased the value to 86, which agrees with Ibata et al.’s value of 90. was calculated to be 86.
#### 4.1.2 Two-Image Model
Ibata et al. (1999) presented the possibility that component C might be the lensing galaxy rather than a third lensed image. In this case, the number of the model parameters reduces to three ($`\theta _e`$, $`\epsilon `$, and $`\gamma `$). The core radius $`\theta _c`$ can be set to 0 because the lack of a third image implies a singular potential core while the position of component C directly determines the position of the lensing galaxy ($`x_l`$, $`y_l`$)<sup>4</sup><sup>4</sup>4Our model is slightly different from that of Ibata et al. (1999) in that we fix the lens position at the position of component C. Their model treats the lens position as a free parameter.. On the other hand, two lensed images provide three constraints: one relative brightness and two coordinates giving one relative image position with respect to the other. Again, the number of the model parameters and that of the constraints are the same. The derived parameters are shown in Table 3. The major difference from the three-image model is the much lower value of total magnification (7). Also, the ellipticity of the potential is becoming significant (0.08). The overall structure of the time-delay surface is very similar to that of the three-image model.
### 4.2 Magnification of Extended Sources
The background source is likely to be spatially extended at longer wavelengths. Therefore, it is necessary to understand how the lens distorts and magnifies an extended source as the outer edge of the source approaches/crosses the caustics. For a background source that is a uniform circular disk, the lensed image can be constructed with the methods of Schramm & Kayser (1987). Since the surface brightness is preserved in gravitational lensing, the areal ratio between the source and the images gives the magnification factor.
For the two models, we show how the source crosses the caustics (Figures 5 and 6) and how the shapes of the lensed images change (Figure 7 and 8). The behavior is similar in the two models: as the source size increases, it first forms an arc (b) and later turns into a ring (c). In the case of the three-image model, the ring quickly becomes a filled disk due to the third image while the ring is not completely filled in the two-image model even with a source radius of 650 pc (d). Figure 9 shows how the total magnification factor changes as the source radius increases in the two models.
There are three major differences between the two models in terms of their response to extended sources. First, a much larger source size is required in the two-image model for each transition of the lensed image shape as shown in Figure 8. Second, the magnification is much greater in the three-image model. Figure 9 shows that the magnification factor of the three-image model could be as large as $`120`$ when the source radius is 50–90 pc while that of the two-image model is more than an order of magnitude less ($`7`$) for the same radius range. Third, the magnification factor of the two-image model hardly changes with an increasing source size while the magnification factor of the three-images model is very sensitive to such a change (Figure 9). In other words, the effects of differential magnification could be significant with the three-image model.
### 4.3 Differential Reddening
One uncertainty underlying the discussion so far is differential reddening. For example, if the three-image model is correct, the line of sight to component C intersects the lensing galaxy at a point only $``$ 0$`\stackrel{}{\mathrm{.}}`$03 from the galaxy’s center (Figure 5b). Such an angular separation corresponds to 200–250 pc at a lensing galaxy’s redshift of $`0.5<z<4`$. If this is the case, the brightness of component C might be more heavily affected by the reddening in the lensing galaxy than those of the other two components. This would produce observed flux ratios considerably different from the intrinsic ones.
However, the currently available data seem to indicate that differential reddening is not significant in this system. Ibata et al. (1999) noted that the three components have almost identical colors from 1.10 $`\mu `$m to 2.05 $`\mu `$m. Our $`K`$-band image also do not show any sign of change in the flux ratio or image morphology, which would have resulted if component C brightens significantly at longer wavelengths. The 12.5 $`\mu `$m image also seems to be consistent with the $`K`$-band image. Therefore, it seems unlikely that the derived models are in serious error due to differential reddening.
### 4.4 The Nature of the Third Image
As seen in Figure 9, depending on whether component C is a third lensed image or a lensing galaxy, the magnification factor could be drastically different. Therefore, the nature of component C is a decisive factor when determining the intrinsic properties of APM 08279+5255. We list three arguments favoring the idea that component C is a third lensed image rather than the lensing galaxy.
#### 4.4.1 The large apparent brightness of component C
If component C is the lensing galaxy, then it cannot be an ordinary field galaxy. The small separation between the lensed images (0$`\stackrel{}{\mathrm{.}}`$4) requires the lensing galaxy to be either a low-mass low-redshift galaxy or a massive high-redshift galaxy in order not to split the lensed images too far apart. However, neither type of galaxy with a normal mass-to-light ratio could produce such a bright apparent magnitude as $`K=14.5`$ mag. More quantitatively, inserting $`\theta _E=0`$$`\stackrel{}{\mathrm{.}}`$2 in equation (2), we obtain possible combinations of the lensing galaxy redshift and velocity dispersion $`(z_l,\sigma _v(\mathrm{km}/\mathrm{s}))`$ as (0.5, 100), (1, 120), (2, 170), and (3, 280). To illustrate how much the lensing galaxy must be overluminous, if we take an L galaxy (i.e., $`\sigma 200`$ km/s) as the lens, it implies that $`z_l2`$, but the expected observed $`K`$ magnitude of such a galaxy would be $`K20`$ mag. Therefore, it requires such a lensing galaxy to be overluminous by more than five magnitudes. Together with the evidence that component C is point-like, this would require that component C is a luminous AGN/QSO nucleus of the lensing galaxy. Although such a QSO–AGN/QSO gravitational lensing may happen (Gott & Gunn 1974), the probability for such an event is in general very small.
#### 4.4.2 Dark Ly$`\alpha `$-forest line cores
The high-resolution Keck HIRES spectrum of APM 08279+5255 (Ellison et al. 1999a ; Ellison et al. 1999b ) shows no residual flux in the core of strong Ly$`\alpha `$ forest lines up to an observed wavelength of 5715 Å. If component C is a lower-redshift AGN/QSO, its continuum flux should easily be detectable in the core of these strong Ly$`\alpha `$ lines, given the high signal-to-noise ratio of this HIRES spectrum. This means that if component C is an AGN/QSO, its redshift must be larger than 3.7, and it seems rather contrived if a $`z=3.9`$ QSO is gravitationally lensed by another QSO system at $`z>3.7`$.
#### 4.4.3 Color
As already mentioned in the discussion of differential reddening, components A, B, and C have almost identical colors from 1.1 $`\mu `$m to 12.5 $`\mu `$m. This is easy to understand if component C is also a lensed image of the same background source.
Based on these arguments, we conclude that component C is a third lensed image of the QSO. Therefore, we investigate the intrinsic properties of APM 08279+5255 based on the three-image model, although in the end this question can be settled with spectroscopy of component C. If, however, the two-image model is correct, its consequences are easy to derive because this model is not sensitive to differential magnification: the magnification factor is $``$ 7-10 over the relevant spectral range, and therefore APM 08279+5255 must be an extremely luminous object with the intrinsic bolometric luminosity $`5\times 10^{14}`$ L. Also, the shape of the intrinsic SED must be close to what is observed.
### 4.5 Differential Magnification
In this section, we discuss the consequences for differential magnification of our preferred three-image model. Unlike the two-image model, the magnification factor of the three-image model is sensitive to the source size (Figure 9). Therefore, the effects of differential magnification must be evaluated before the intrinsic properties of the lensed QSO can be discussed. For the discussions below, references are made with respect to the restframe wavelength of APM 08279+5255.
#### 4.5.1 Magnification in the Restframe UV/Optical
The fact that three lensed images are completely point-like in the restframe $`B`$ band (observer’s $`K`$ band) sets the upper limit of $``$ 20 pc on the source radius. If the source radius were larger than 20 pc, we should see the effects of image elongation seen in Figure 7a. In the restframe UV, the source size is expected to be either comparable to or smaller than that in the restframe optical. From Figure 9, it can be seen that the magnification factor corresponding to this range of source size is $``$ 90, which is same as the value derived by Ibata et al. (1999).
#### 4.5.2 Magnification in the Restframe Near-IR
Since the spatial resolution of the restframe near-IR (observed 12.5 $`\mu `$m) image is not high enough to detect the morphology of each component, it is not possible to set as stringent a limit on the source radius as in the restframe optical. As seen in Figure 7, once the source radius reaches 50 pc, components A and B will connect with each other and form an arc, and this would produce a lower ellipticity in the resultant image (Figure 10). However, this has not been seen in Figure 3a and c. Based on the overall ellipticity of the restframe near-IR image, we conclude that the source radius in the restframe near-IR (observer’s 12.5 $`\mu `$m) must be less than 50 pc, which corresponds to a magnification factor of 90–120 (Figure 9).
#### 4.5.3 Magnification in the Restframe Mid-IR
There exists no high-resolution spatial information in the restframe mid-IR (observer’s far-IR) which can be used to determine the magnification factor directly. However, it is possible to put constraints on the magnification factor based on the restframe mid-IR luminosity.
With the assumption that the mid-IR emitting region is intrinsically a circular disk on the sky with a constant specific intensity, the mid-IR SED can be modeled with the following expression:
$$f_{\nu _{obs}}=mQ_{em}(\nu _e)B_{\nu _e}(T_d)(1+z_s)^3\pi \left(\frac{r}{D_s}\right)^2.$$
(3)
Here, $`m`$ is the magnification factor, $`Q_{em}`$ is the dust emissivity, $`B_{\nu _e}`$ is the Planck function evaluated at the emitted frequency, $`T_d`$ is the dust temperature, $`z_s`$ is the redshift of the background source, and $`r`$ is the intrinsic radius of the emitting region. The Planck function is evaluated at the emitted frequency ($`B_{\nu _e}`$) while the observed flux density is expressed at the observed frequency ($`f_{\nu _{obs}}`$; $`\nu _e=\nu _{obs}(1+z_s)`$).
As the dust temperature, we adopt the CO gas temperature of 200 K derived by Downes et al. (1999) from the ratio of the CO (4–3) and (9–8) lines. If this 200 K region is strongly magnified, the dust global temperature could be much lower than 200 K, but it does not affect our discussion here because dust at a temperature significantly below 200 K would not provide a significant flux in the restframe mid-IR.
Only specific combinations of $`m`$ and $`r`$ are allowed by the lens model (Figure 9). For each allowed combination, the corresponding value of $`Q_{em}`$ is determined such that equation (3) reproduces the observed IRAS 100 $`\mu `$m flux. Since 100 $`\mu `$m is close to the peak of a 200 K blackbody at a redshift of 3.91, the value of $`Q_{em}`$ essentially scales the total energy output of such a blackbody. Therefore, its value can be regarded as equivalent to that of the Planck-averaged dust emissivity (Draine & Lee 1984).
As long as the dust emissivity $`Q_{em}`$ is larger than $``$ 0.03, it is always possible to find a set of a source radius and a magnification factor which reproduces the observed 100 $`\mu `$m flux. However, the emissivity cannot be smaller than this because the intrinsic source size would become larger than 0$`\stackrel{}{\mathrm{.}}`$3, which would produce a gravitationally magnified image with a diameter $``$ 1″, roughly the upper limit on the millimeter source size set by Downes et al. (1999). A Planck-averaged emissivity of 0.03 is similar to that of astronomical silicate grains with a radius 0.01-1 $`\mu `$m at 200 K as found by Draine and Lee (1984).
Depending on the dust emissivity, the magnification factor in the mid-IR could be anywhere between 4 ($`Q_{em}=0.03`$) and 120 ($`Q_{em}=1`$), which corresponds to the source radius of 1.8 kpc and 60 pc, respectively. The case of $`Q_{em}=1`$ corresponds to an optically thick source of temperature 200 K. The source radius of the 200 K component becomes larger with a smaller dust emissivity because dust grains can be heated to 200 K at a farther distance from the central source.
#### 4.5.4 Magnification in the Restframe far-IR/submm
The range of possible magnification factors in the restframe far-IR/submm is the same as that in the mid-IR (i.e., 4–120). However, it cannot exceed the magnification factor in the mid-IR since the far-IR/submm emitting region is expected to be larger than the mid-IR emitting region. The magnification factor could be exactly the same in the restframe mid-IR and far-IR/submm if the emission from the 200 K component also dominates in the latter wavelength regime.
Table 4 summarizes the derived magnification factors at different wavebands and the corresponding source sizes. It can be seen that the effects of differential magnification are negligible through the restframe UV, optical, and near-IR while it could potentially be significant in the restframe mid-IR/far-IR/submm.
### 4.6 Intrinsic Properties
#### 4.6.1 Spectral Energy Distribution
From the restframe visible to near-IR, the magnification factors are estimated to be $``$ 100, and the intrinsic SED, basically unaffected by differential magnification, is extremely flat ($`f_\nu \nu ^1`$, Figure 11). On the other hand, in the restframe mid-IR/far-IR/submm, the exact magnification factors are not known. Therefore, we derive instead the possible range of the intrinsic SED allowed by the model.
The allowed range can be calculated based on the fact that the lens model requires the magnification factor to be 4–120 in the restframe mid-IR/far-IR/submm. In Figure 11, we empirically fit the observed SED with the following expression,
$$f_{\nu _{obs}}B_{\nu _e}(T_d=200\mathrm{K})(1e^{\tau _{\nu _e}}),$$
(4)
where $`\tau _{\nu _e}`$ is the optical depth parametrized as,
$$\tau _{\nu _e}=\left(\frac{\nu _e}{\nu _0}\right)^2.$$
(5)
Here, $`\nu _0`$ is set to be 1.5 THz (i.e., 200 $`\mu `$m). The possible range of the intrinsic SED (the shaded area in Figure 11) is determined by demagnifying the observed SED fit by factors between 4 and 120. The upper limit corresponds to the case in which the dust emissivity is low (0.03) and therefore the 200 K dust region is large, resulting in a small magnification factor while the lower limit corresponds to the case in which the dust emissivity is unity (i.e., blackbody) and therefore the 200 K region is small, resulting in a large magnification factor.
As seen in Figure 11, in principle the intrinsic SED of APM 08279+5255 could strongly peak at $`\lambda _{rest}20\mu `$m. However, such an SED is unlikely. If the SED strongly peaks in the restframe mid-IR, the dust covering factor around the central energy source must be large. However, as can be seen in Figure 11, the intrinsic SED is extremely flat at $`\lambda _{rest}<2\mu `$m, which means that our line of sight to the central QSO is relatively dust free. Therefore, if the SED strongly peaks in the mid-IR, it must mean that we are looking into a heavily dust-enshrouded QSO through a relatively transparent hole, which seems unlikely. We believe it more likely that the intrinsic SED is relatively flat up to $`\lambda _{rest}=20\mu `$m and drops at longer wavelengths (dark-shaded region in Figure 11).
#### 4.6.2 Bolometric Luminosity
If we assume that the intrinsic SED is flat up to the restframe mid-IR, the characteristic magnification factor of APM 08279+5255 can be taken as $`100`$. Therefore, the intrinsic bolometric luminosity is estimated to be $`5\times 10^{13}\mathrm{L}_{}`$. The IR luminosity at $`\lambda >10\mu `$m is $`1\times 10^{13}\mathrm{L}_{}`$. Because most of the luminosity is coming out at shorter wavelengths, we can determine the luminosities without knowing the precise shape of the restframe far-IR/submm SED.
#### 4.6.3 Dust Properties
Figure 11 shows that the mid-IR/far-IR/submm SED of APM 08279+5255 is roughly consistent with those of lower-redshift IR-luminous galaxies/QSOs. However, since we cannot constrain the intrinsic SED at $`\lambda _{rest}>100\mu `$m to better than a factor of 10, it is not possible to set meaningful limits on intrinsic properties of this object such as the mass and temperature of dust.
The main sources of uncertainty are the dust emissivity and the effects of differential magnification. As seen in equation (3), the magnification factor is inversely proportional to the dust emissivity. Since the dust emissivity is not well known even in the local universe, this introduces a large uncertainty in the magnification factor, and therefore in the intrinsic luminosity of the dust emission. In addition, there is a possibility that the magnification factor may systematically decrease at longer wavelengths because cooler regions have larger spatial extent (Eisenhardt 1996; Blain 1999). If this is the case, there may exist a cold (e.g., 50 K) massive dust component which is not seen in the observed spectrum due to its small magnification factor. Because of these uncertainties, a self-consistent model can be constructed with a broad range of dust masses ($`10^610^9\mathrm{M}_{}`$) and temperatures (50–200 K).
Figure 11 also illustrates the difficulty of determining dust properties without knowing the exact shape of the mid-IR/far-IR/submm SED. For example, one noticeable feature in the figure is that most of the objects have comparable submillimeter luminosities in spite of the large spread in the bolometric luminosity. This strongly suggests that the dust distribution in these objects are “matter-bounded” (cf. Barvainis et al. 1995): in other words, these objects are likely to have comparable dust masses, and the difference in the IR ($`>10\mu `$m) luminosities is caused by the difference in dust temperature (from 50 K (ULIRGs) to 200 K (PG 1206+459) for the dust component dominating the restframe submm emission). On the other hand, the large submillimeter luminosity of BRI1202-0725 indicates that even if the dust temperature is at 200 K, its dust mass is still as large as those of local ULIRGs, and could be larger by an order of magnitude if the dust temperature is low (e.g., 50 K). The fact that the allowed IR/submm SED range for APM 08279+5255 is broadly consistent with the SEDs of such a variety of objects indicates that with the currently available data, little can be said about the dust mass and temperature of APM 08279+5255.
### 4.7 The Lensing Galaxy
If the three-image model is correct, there is no observational image that can be associated with the lensing galaxy. The redshift of the lensing galaxy is not determined by the model because the redshift of the lens enters only through the Einstein radius, and this can be produced by a distant massive galaxy or a nearby low-mass galaxy.
The derived core radius of 0$`\stackrel{}{\mathrm{.}}`$2 corresponds to a physical length of 1.3–1.6 kpc at a redshift of $`0.5<z<4`$. A galaxy with a core radius of this size must be a large elliptical-type galaxy with a velocity dispersion $`\sigma _v`$ of $``$ 300 km/s. To be compatible with the derived small Einstein radius, the lensing galaxy must be at a high redshift, probably $`z3`$, although at the moment there is no observational evidence to support this hypothesis. The mass of such a lensing galaxy is calculated to be $`2\times 10^{11}`$ L within the Einstein radius of 0$`\stackrel{}{\mathrm{.}}`$29. Such a galaxy would not have been detected by any observations to date of this system.
## 5 SUMMARY AND CONCLUSIONS
We have obtained high-resolution images of APM 08279+5255 in the $`K`$-band (FWHM $``$ 0$`\stackrel{}{\mathrm{.}}`$15) and at 12.5 $`\mu `$m (FWHM $``$ 0$`\stackrel{}{\mathrm{.}}`$4). We have constructed a gravitational lens model using the $`K`$-band image, and determined the magnification factors at longer wavelengths based on the 12.5 $`\mu `$m image and other existing data. The basic conclusions are as follows:
1. APM 08279+5255 consists of three components, which are distributed over $``$ 0$`\stackrel{}{\mathrm{.}}`$4. The third image (component C) is bright ($`K=14.5`$ mag) and compact (FWHM $`0\stackrel{}{\mathrm{.}}15`$).
2. Component C could be either a third lensed image (three-image model) or the luminous QSO/AGN nucleus of the lensing galaxy (two-image model). If the three-image model is correct, the magnification factor is high ($`100`$) and the effects of differential magnification could be significant. If the two-image model is correct, the magnification factor is low ($`10`$), and the effects of differential magnification are negligible. The three-image model is preferred by a number of observational arguments, and therefore we adopt this model to deduce the intrinsic properties.
3. The derived magnification factors for APM 08279+5255 are 90, 90–120, and 4-120 in the restframe UV/optical, near-IR, and mid-IR/far-IR/submillimeter, respectively. The corresponding intrinsic source radii are $`<20`$ pc, $`<`$ 50 pc, and 60–1800 pc.
4. By assuming that the intrinsic SED is flat up to $`\lambda _{rest}<20\mu `$m, we estimate an overall magnification factor of $`100`$. From this, the intrinsic bolometric luminosity of APM 08279+5255 is derived to be $`5\times 10^{13}\mathrm{L}_{}`$. Its intrinsic IR luminosity is $`1\times 10^{13}\mathrm{L}_{}`$. Therefore, APM 08279+5255 is intrinsically luminous, but it is not the most luminous object known.
5. With the currently available data, neither the dust mass nor the dust temperature can be determined due to the uncertainties associated with the dust emissivity and the possible effects of differential magnification.
###### Acknowledgements.
We thank Drs. Ibata and Lewis for communicating us their results prior to the submission of their paper. E.E. thanks Drs. Chris Fassnacht and Andrew Blain for helpful discussions. The W. M. Keck Observatory is operated as a scientific partnership between the California Institute of Technology, the University of California, and the National Aeronautics and Space Administration. It was made possible by the generous financial support of the W. M. Keck Foundation. Infrared astronomy at Caltech is supported by grants from the NSF and NASA. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. |
warning/0001/physics0001015.html | ar5iv | text | # Using atomic interference to probe atom-surface interaction
## I Introduction
The interaction between a ground–state atom and a dielectric or conducting wall has been investigated theoretically (, ,,, , , , , , ) and experimentally (, , , ). Theoretical studies have been performed on different levels, from a simple model of a dipole–dipole interaction of the atom and its mirror image, to the full QED relativistic quantum treatment. Interesting in particular are the long–range Casimir interactions that were recently observed in cavity QED experiments , . When the atom–wall distance $`z`$ is not small compared to the wavelength of the dominant atomic transitions, the $`z^3`$ law associated with the instantaneous electrostatic interaction is no longer valid. The full QED treatment leads to the famous long distance $`z^4`$ law.
Recent experimental developments enable precise manipulation of cold atoms by lasers, see e.g. Ref. . Small and well defined velocities of the atoms can be achieved using advanced cooling and launching techniques and a detuned laser field can be used to create controlled and adjustable potentials for the atoms. Under these conditions, atoms can be used to explore the atom–surface potential, for example using evanescent–wave mirrors. Classical reflection from such an atomic mirror was used to measure the van der Waals force between a dielectric surface and an atom in its ground state . This experiment though, could not fully discriminate between the electrostatic and the QED expressions. Segev et al. considered a similar experiment in the quantum regime (atoms undergoing above–barrier, classically forbidden reflection). Unlike classical reflection, which can only be used to identify thresholds and to measure the height of the potential barrier, quantum probabilities are determined by the complete potential curve, and are sensitive to the short- and long–range behavior of the potential. It was found that, for velocities of the order of the recoil speed, the quantum reflection probabilities are indeed sensitive to the long–range (Casimir) interaction.
In this work we study how the form of the atom–surface interaction can be observed using atomic interference of the type reported in . We consider atoms with multiple ground state sub-levels, which feel different potentials in the evanescent radiation field. These potentials can be probed by using stimulated Raman transitions within the evanescent wave. These transitions exhibit interference effects whose phases depend on the atomic trajectories and on the entire potential, as in quantum reflection. An important aspect of the effects we discuss here is that they occur for higher incident velocities than those considered in and may therefore be experimentally easily accessible. Furthermore, even for small velocities (near the recoil velocity) the effects are more dramatic than those shown by quantum reflection.
As discussed in Refs. and , the experiment is analogous to an atom interferometer whose size is of the order of a fraction of the evanescent wave decay length, $`\kappa ^1`$. We will focus on three experimental cases: in the first, the atomic de Broglie wavelength $`\lambda _{\mathrm{dB}}\kappa ^1`$. This leads to interference fringes analogous to those in any other interferometer. It corresponds to the experiment described in Ref. in which rubidium (Rb) atoms are dropped from a height of about 2 cm. Second, we will repeat the situation of the first case, but with Rb replaced by meta-stable helium (He\*). In this case we have $`\lambda _{\mathrm{dB}}/2\pi `$ a few $`\kappa ^1`$, which allows for only a few interference fringes. This allows us to illustrate how the signal changes when only the atomic species is changed. Finally, we will examine the case where $`\lambda _{\mathrm{dB}}/2\pi \kappa ^1`$, taking He\* atoms at the recoil velocity as our example. No fringes are present, but a strong dependence on the nature of the atom-wall potential is demonstrated. To simplify the discussion, we work with a two-level model, between the initial atomic state and the adjacent one in the ground state manifold, thus neglecting the populations of the other ground state sub-levels as well as the excited levels. For the atom-wall interaction we will use the published values for each atom.
In the short $`\lambda _{\mathrm{dB}}`$ limit, the motion of the atom can be treated semi-classically. We thus calculate the transition probability between the two atomic levels according to the Landau-Zener model for adiabatic transitions. In the other cases, the study of the atomic motion requires a fully quantum mechanical treatment. We study these cases by solving the associated coupled Schrödinger equations numerically.
In Sec.II we describe the physical system and the theory for atomic interference. In Sec.III the results are presented. Discussion and conclusions are given in Sec.IV.
## II Model for interference of multiple ground state sub-levels
We will focus on an experimental setup along the lines of Fig. 1. Here, a strong laser beam with frequency $`\omega `$ and transverse magnetic (TM) polarization (i.e. magnetic field perpendicular to the plane of incidence) creates an evanescent wave, which is nearly $`\sigma ^{}`$ circularly polarized with respect to the y-axis. A second weak laser beam with frequency $`\omega \mathrm{\Delta }\omega `$ and transverse electric (TE) polarization (i.e. electric field perpendicular to the plane of incidence) creates another evanescent wave with $`\pi `$ polarization along the y-axis.
Atoms normally incident on the evanescent wave move in an effective optical potential $`\widehat{V}_{\mathrm{light}}`$ (which in general depends on the internal state of the atom) and an attractive atom–wall interaction $`\widehat{V}_{\mathrm{wall}}`$. The total potential is given by
$$\widehat{U}(z)=\widehat{V}_{\mathrm{light}}(z)+\widehat{V}_{\mathrm{wall}}(z).$$
(1)
### A Optical potential
Let us consider an atom in the evanescent field and assume for simplicity that the $`\sigma ^{}`$ polarized wave is strong compared to the $`\pi `$ wave. The $`\sigma ^{}`$ component lifts the magnetic (Zeeman) degeneracy of the atomic ground state (the quantization axis is in the $`\widehat{y}`$ direction), so that each magnetic sub-level feels a different optical potential. To first order, the $`\pi `$ polarized wave produces a coupling between the atomic sub-levels via a Raman transition: for example, starting from the sub-level $`m_i`$, the atom absorbs a $`\sigma ^{}`$ polarized photon and emits a stimulated photon with $`\pi `$ polarization. The atom thus ends up in the $`m_i1`$ sub-state with its total energy increased by $`\mathrm{}\mathrm{\Delta }\omega `$. This transition is resonant when $`\mathrm{}\mathrm{\Delta }\omega `$ is equal to the energy difference between the magnetic sub-levels. If the two evanescent waves are counter-propagating, in the reference frame moving with the optical grating, the situation corresponds to grazing incidence diffraction . For a review of the theoretical understanding of atomic diffraction and interference from evanescent waves, see .
In the limit of low saturation and a detuning $`\delta `$ large compared to both the frequency difference $`\mathrm{\Delta }\omega `$ and the natural linewidth, the excited state manifold may be eliminated adiabatically, and, for a ground state of total angular momentum (including nuclear spin) $`J_g`$, the atomic wave function is described by the $`2J_g+1`$ Zeeman components $`|m>,m=J_g,\mathrm{},+J_g`$. An atom at a distance $`z`$ from the surface of the mirror is subject to an optical potential $`\widehat{V}_{\mathrm{light}}(z)`$ whose matrix elements are of the form \[we suppose that the frequency difference $`\mathrm{\Delta }\omega `$ and Zeeman shift are negligible compared to the detuning $`\delta `$ and the hyperfine structure of the excited level\]
$$<m|\widehat{V}_{\mathrm{light}}(z)|m^{}>=\frac{d^2}{\mathrm{}\delta }\underset{q,q^{},m_e,J_e}{}E_q^{}(z)E_q^{}(z)(J_gm;1q|J_em_e)(J_em_e|J_gm^{};1q^{}),$$
(2)
where a product of Clebsch–Gordon–coefficients appears on the rhs, the electric field polarization is expanded in the usual spherical basis with coefficients $`E_q`$, ($`q=1,0,+1`$) and $`d`$ is the reduced dipole moment. The optical potential couples different Zeeman sub-levels if the field is not in a pure polarization state with respect to this basis, as in the setup considered in this work. The optical potential due to the two evanescent waves can then be written as
$$<m|\widehat{V}_{\mathrm{light}}(z)|m^{}>=\frac{d^2}{\mathrm{}\delta }C_{mm^{}}\mathrm{exp}(2\kappa z),$$
(3)
where the $`C_{mm^{}}`$ coefficients are given by the Clebsch-Gordon coefficients times the field amplitudes at the surface $`z=0`$ (see Eq. 2). The inverse decay length $`\kappa `$ is $`\kappa =k[(n\mathrm{sin}\theta )^21)]^{1/2}`$, where $`k=\omega /c`$ is the free field wave vector, assumed to be the same for each laser, $`n`$ is the refraction index of the prism and $`\theta `$ is the angle of incidence of the lasers with the surface. In our calculations we will always use $`n=1.87`$ and $`\theta =53^{}`$. The wave-vector $`k`$ is different for each atom.
### B Atom–wall potential
The simplest model for the interaction of a ground state atom and a wall of dielectric constant $`ϵ`$ considers the interaction between a dipole d and its mirror image and yields the Lennard–Jones potential,
$$V_{\mathrm{wall}}^{\mathrm{LJ}}(z)=\frac{ϵ1}{ϵ+1}\left(\frac{<d_{}^2>+2<d_{}^2>}{64\pi ϵ_0}\right)\frac{1}{z^3}\frac{ϵ1}{ϵ+1}\frac{C^{(3)}}{z^3},$$
(4)
where $`<d_{}^2>`$ and $`<d_{}^2>`$ are the expectation values of the squared dipole parallel and perpendicular to the surface ,. This expression for the potential is approximately valid for $`ϵ`$ independent of frequency and $`kz`$ much smaller than unity.
If we take into account retardation effects, the Casimir-Polder potential is obtained, where the finite propagation time between the dipole and its image results in a different power-law behavior for large $`z`$,
$$\underset{z\mathrm{}}{lim}V_{\mathrm{wall}}^{CP}(z)z^4.$$
(5)
In the complete QED theory the interaction potential between an atom of polarizability $`\alpha (\omega )`$ at a distance $`z`$ from a dielectric wall can be written as :
$$V_{\mathrm{wall}}^{\mathrm{QED}}(z)=\frac{\mathrm{}}{8\pi ^2c^3}_0^{\mathrm{}}𝑑\omega \omega ^3\alpha (\omega )\times \left(_0^1𝑑p+_0^i\mathrm{}𝑑p\right)H(p,ϵ)\mathrm{exp}(2ip\omega z/c)$$
(6)
with
$$H(p)=\frac{\sqrt{ϵ1+p^2}p}{\sqrt{ϵ1+p^2}+p}+(12p^2)\frac{\sqrt{ϵ1+p^2}ϵp}{\sqrt{ϵ1+p^2}+ϵp}.$$
(7)
The numerical values of the constant $`C^{(3)}`$ for Rb and He\* atoms used in this paper are given in Table I, as well as the numerical values of some other important parameters. Since the atom is in a ground state with $`l=0`$, the value of $`C^{(3)}`$ is the same for any magnetic or hyperfine sub-level. The detailed interaction between a ground state Rb or meta-stable He atom and a dielectric surface were recently calculated. For the van der Waals potential we have used data from for Rb and from for He\*. In the former work an interpolation formula for the van der Waals potential is given as $`V_{\mathrm{wall}}^{QED}(z)=V_{\mathrm{wall}}^{LJ}(z)0.987[(1+1.098z)^10.00493z(1+0.00987z^30.00064z^4)^1]`$ where $`z`$ is expressed in units of the laser wavelength $`\lambda /(2\pi )`$. This formula approximates the numerical calculation with a 0.6% accuracy between 0 and $`10\lambda /(2\pi )`$.
### C Population transfer
In this work we will consider only population transfer from the incident sub-level $`i`$ to the final one $`f`$ ($`m_f=m_i1`$), that is only two levels. This is a good approximation if the coupling is weak enough for the population of the other levels to be negligible. An example of the total interaction potential from Eq. 1 is shown in Fig.2. The $`U_{ff}`$ potential curve has been shifted vertically by $`\mathrm{}\mathrm{\Delta }\omega `$, corresponding to the kinetic energy change. Then the two (adiabatic) potential curves cross at the point of resonance. The coupling turns the exact crossing into an avoided crossing. An incoming wave function in the $`m_i`$ channel is split in two parts that are subsequently reflected from their respective repulsive potentials and recombined after the second passage at the crossing. Thus, the evanescent wave realizes a “Michelson interferometer” with a single beam splitter and two mirrors.
When the atomic $`\lambda _{\mathrm{dB}}`$ is short, the avoided crossing can be treated by means of the semi-classical Landau–Zener model for non–adiabatic transitions . Assuming that the atom moves through the crossing with a constant velocity $`v_c`$ (fixed by energy conservation), the Landau–Zener formula allows one to compute the probability amplitude for the two atomic levels after the crossing. The transition probability $`w_{if}`$ from the initial sub-level $`i`$ to the final sub-level $`f`$ is given by
$$w_{i,f}=4T_{LZ}R_{LZ}\mathrm{cos}^2(\delta \varphi ),$$
(8)
where $`R_{LZ}=1T_{LZ}`$ and the transmission coefficient is $`T_{LZ}=\mathrm{exp}(\pi \mathrm{\Lambda })`$, with $`\mathrm{\Lambda }=2|<i|\widehat{U}|f>|^2/(\mathrm{}^2\kappa \mathrm{\Delta }\omega v_c).`$
The phase difference $`\delta \varphi `$ is given by the difference in the phase shifts between the crossing and the turning points in the semi-classical approximation, plus a correction term :
$$\delta \varphi =\frac{1}{\mathrm{}}\left[_{z_{f,r}}^{z_c}𝑑zp_f(z)_{z_{i,r}}^{z_c}𝑑zp_i(z)\right]+\frac{\pi }{4}+\frac{\mathrm{\Lambda }}{2}\mathrm{log}\left(\frac{\mathrm{\Lambda }}{2e}\right)+\mathrm{arg}\left(\mathrm{\Gamma }\left(1i\frac{\mathrm{\Lambda }}{2}\right)\right),$$
(9)
where $`z_c`$ is the position of the crossing point, $`z_{n,r}`$ the classical return point for an atom in the n-th level, $`p_n(z)`$ its momentum and $`\mathrm{\Gamma }`$ the Gamma function.
Changing the frequency difference $`\mathrm{\Delta }\omega /(2\pi )`$ causes a change in the length of one of the interferometer arms, thus a change in the phase difference $`\delta \varphi `$ between the two paths. As a consequence we expect the transition probability to show oscillations in $`\mathrm{\Delta }\omega `$ (Stückelberg oscillations). We will see in the next section that $`\delta \varphi `$ is very sensitive to the exact shape of the potential. The amplitude of the oscillations also depends on $`\mathrm{\Delta }\omega `$ both explicitly and implicitly through $`v_c`$ (the crossing point moves with changing $`\mathrm{\Delta }\omega `$).
The Landau–Zener model is a good approximation only when the atom speed is approximately constant during the crossing. In particular it is not valid when the classical return point and the crossing are close to each other or when the de Broglie wavelength of the atom is of the order of the width of the interaction region. In order to explore this long-wavelength regime, we have to forgo the semi-classical Landau-Zener model and solve numerically the coupled Schrödinger equations for the system. Since atoms that cross the potential barrier stick to the dielectric surface, the appropriate boundary conditions at the surface are those for a running wave propagating downward ($`z\mathrm{}`$), while for $`z\mathrm{}`$ the solution is a superposition of downward (incident) and upward (reflected) waves. We have integrated the system of Schrödinger equations using the renormalized Numerov method . To avoid the singularity at the surface we have modified the potential to be a large negative constant near the interface and verified that the transition probability does not change by varying the value of the constant.
## III Results
We will present our calculations in two parts. First we discuss the short wavelength regime, and point out various experimental strategies to observe retardation effects. Then we discuss two cases in which the de Broglie wavelength is not small compared to the evanescent wave decay length (the ”long wavelength regime”), in which a numerical integration of the Schrödinger equation is necessary.
### A Short wavelength regime
Fig. 3 shows a calculation of the population transfer $`w_{if}`$ as a function of the frequency difference $`\mathrm{\Delta }\omega `$ for the Lennard–Jones (LJ) and QED model of the van der Waals interaction. The value of the light potential is the same for the two curves (i.e. the laser intensity is the same), and the incident momentum is $`115\mathrm{}k`$, which corresponds to Rb atoms dropped from a height of 2.3 cm. We use the Landau-Zener approach to calculate the transfer probability (see Eq. 8). As in and , we observe Stückelberg oscillations in the transfer probability. These oscillations can be understood as the variations in the accumulated phase difference between the two different paths taken by the atoms after the level crossing shown in Fig.2. The de Broglie wavelength $`\lambda _{\mathrm{dB}}`$ is such that several fringes appear as the position of the level crossing is moved through its possible range. The last oscillation at the higher frequencies is where the crossing point and the classical return point are very close to each other and the Landau-Zener model breaks down. We have set the probability to zero beyond this limit. In reality, the transition probability falls roughly exponentially with frequency, as one finds solving the Schrödinger equations numerically (see next subsection).
In general, we find that the dephasing between the two curves (with and without retardation effects) is greatest when the atoms are incident at an energy close to the top of the potential barrier. Note that the height of the barrier is greater when retardation effects are included, since these reduce the strength of the atom-surface interaction. The effect of retardation is roughly to shift these fringes by half a fringe to the left. The major cause of this shift is the increase in the height of the total potential which is greater for the $`i`$ level than the $`f`$ level. A 10% increase in the value of the light-shift potential would exhibit nearly the same shift. Therefore, since it is not possible to turn retardation on and off, it would be necessary to measure absolutely the light-shift potential to better than 10% in order to distinguish a retardation effect. Experimentally this is rather difficult. Instead of attempting to measure the absolute light shift, one could rather measure the absolute height of the potential by observing the threshold of reflection as in , and using the known kinetic energy of the atoms to get an absolute calibration of the height. Let us assume then that the barrier height, instead of the light intensity, is known. In Fig.4a we show the result with the same parameters as in Fig.3 except for the light intensity in the LJ model, which has been changed so as to have the same barrier height as the QED model. In this case the shift is much smaller, about 1/5 of a fringe. We have verified that even taking into account an experimental uncertainty of a few % in the height of the corresponding potentials, the two models are still clearly resolved.
This approach seems feasible, but a third method of observing the effects of retardation is possible if one uses more of the information available in the oscillation curve. Fig. 4b shows the same curves as Fig. 4a but with the QED curve numerically shifted so as to coincide with the LJ curve at its maximum. One sees that the period of these oscillations is not the same. It decreases with decreasing detuning, faster for the full QED potential, so that there is a difference in the spacing of the minima in the population transfer. Thus with fringes with sufficiently high signal to noise, one can distinguish retardation while leaving the absolute barrier height as a free parameter in a fit to the data. It seems to us that a viable experimental method is to use a combination of the second two approaches. Careful measurements of the barrier height can be used to cross check a fit to the Stückelberg oscillations with the barrier height as a free parameter.
The incident energy, momentum and barrier height used in Figs. 3 and 4a were arbitrarily chosen to correspond to the experiment in , but it would be interesting to know how Figs. 3 and 4a would change, especially for different incident momenta (de Broglie wavelengths). We studied this question by repeating the calculation of Fig. 4a, for different incident momenta, while always keeping the barrier height 10% above the incident kinetic energy. We find that, roughly speaking, the number of oscillations increases as the incident momentum. This is because the number of fringes in the interferogram increases with decreasing wavelength. The accumulated phase difference between the oscillations in the LJ and QED models, over the corresponding frequency range, also increases approximately linearly with incident momentum. Thus if we consider the fringe shift divided by the fringe period as a figure of merit, the sensitivity of the experiment to retardation effects increases with increasing incident momentum.
### B Long wavelength regime
We now examine the large de Broglie wavelength limit, where the semi-classical model breaks down since the atomic wavelength is not small compared to the interferometer size. We first consider the case of He\* atoms dropped from a height of a few cm as for Rb. We show the results in Fig. 5. For an initial distance of 2.3 cm from the mirror, the incident momentum of He\* is $`7.4\mathrm{}k`$, much lower than for Rb. Again we will chose the intensities of the strong $`\sigma ^{}`$ wave so as to have the same barrier height for the two potentials, about 10% above the incident kinetic energy. In this regime, to observe interference, the detuning between the evanescent waves must be of only a few MHz, in fact, beyond about 10MHz the crossing point is closer to the surface than the return point. One only sees one or two Stückelberg oscillations since the momentum involved is small and the atoms do not accumulate enough phase difference to show more oscillations. The two potentials give similar results (Fig.5), the main difference being in the shape and height of the big peak.
We have also looked at even lower incident momenta. Near the recoil limit, one can expect interference only for detunings of less than one MHz and the accumulated phase difference is too small in this range to show any Stückelberg oscillations. Both for He\* and Rb one obtains a big peak in the transition probability and no Stückelberg oscillations, as expected. The difference between the van der Waals potential and the full QED potential still shows up in the different shape and height of the peaks (Fig.6). Here the width of the curve is delimited by the frequency for which the classical return point and the crossing point coincide.
Finally we note that the qualitative features of the short and long $`\lambda _{\mathrm{dB}}`$ limit are the same for Rb and He\*, e.g. He\* with an incident momentum of about $`100\mathrm{}k`$ gives the same type of oscillation pattern as Rb.
## IV Conclusions
In summary, we have proposed an experiment to probe van der Waals like surface interactions by exploiting interference mechanisms for well defined Zeeman sub-levels of atoms moving in two evanescent waves. Retardation can be resolved using atoms incident at speeds which are easily obtained in free fall over a few centimeters. The controlling parameter is here the detuning between the two evanescent waves. One then measures the fraction of the atoms which have undergone a change of magnetic sub-level as a function of this detuning.
For the situation in which the atomic de Broglie wavelength is sufficiently small the experiment resembles typical interferometry experiments. The theoretical description is semi-classical, employing well defined atomic trajectories, while experimentally, one seeks a particular (non sinusoidal) fringe pattern as a signature of retardation. This should be possible with an improved version of the experiment of Ref. .
Another approach is to investigate the interaction of atoms whose de Broglie wavelength is not small compared to the length scale of the interferometer. In this regime most of the information is to be found in the shape of the population transfer curve, since there are very few or zero interference fringes. Note though, that we have assumed throughout that the incident atoms are mono-energetic. This means that the velocity spread of the incident atoms must be small compared to the atomic recoil. Nevertheless this may be worth the effort, because the predicted effect of retardation is quite dramatic.
For a quantitative comparison however, the presence of all the sub-levels (which give rise to the multiple crossing of the dressed potentials at the same distance $`z_c`$ from the surface) and, possibly, losses from spontaneous emission have to be taken into account, but the results are not qualitatively different.
## V Acknowledgments
R. M. acknowledges support from the Training and Mobility of Researchers (European Union, Marie–Curie Fellowship contract n. ERBFMBIC983271) and would like to thank Paul Julienne and Olivier Dulieu for their help with the numerical code, and J. Babb for providing the data for the atom-wall interaction for meta-stable helium. This work was also supported by the Région Ile de France. |
warning/0001/astro-ph0001242.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The study of the physics of accretion powered sources, whether on galactic (X-ray binaries) or extragalactic systems (AGN), involves length scales much too small to be resolved by current technology. As such, this study is conducted mainly through the theoretical interpretation of their spectral and temporal properties. Until recently studies of this class of objects focused, for technical mainly reasons, on their X-ray spectra. These are generally fit very well with those of Comptonization of soft photons by hot $`(T_e\stackrel{<}{}10^9`$ K) electrons, a process explored in great depth over the past twenty or so years (; R. Sunyaev this volume). Because the electron temperatures of matter accreting onto a black hole are expected to be similar to those necessary to produce the observed spectra, it has been considered that detailed spectral fits of these sources would lead to insights on the dynamics of accretion onto the compact object.
However, the determination of accretion dynamics requires the knowledge of the density and size of the emitting region, neither of which is provided by radiative transfer and spectral fitting considerations (the equations of radiative transfer involve the optical depth as the independent variable). Indeed, as shown explicitly in and , plasmas of very different radial extent and density profiles can yield identical Comptonization spectra. The degeneracy of this situation can lifted with the additional information provided by timing observations.
The timing properties of BHC, however, suggest length scales inconsistent with the prevailing notion that the observed X-rays are emitted from a region of size a few Schwarzschild radii, $`R_S`$: The power spectra (PSD) of BHC exhibit most of their power at scales $`1`$ s, far removed from the characteristic time scales associated with the dynamics in the vicinity of the black hole horizon, $`R_S/c10^3`$ s. Until recently rather little attention has been paid to this time scale discrepancy, generally attributed to the (unknown) mechanism “fueling” the black hole, presumably operating at much larger radii; rather, more attention was paid to the flicker noise–type ($`f^1`$) PSDs of this class of sources. Novel insight into the variability properties were introduced by , who showed that both the magnitude and the Fourier period dependence of the lags in the X-ray light curves at two different energies was inconsistent with Comptonization by a plasma of size $`\mathrm{a}\mathrm{few}R_S`$; the magnitude of time lags ($`0.1`$ s) suggested an emission region of much larger size, precluding thus an explanation of the observed PSDs due to a modulation of the accretion rate onto the black hole.
These discrepancies between the expected and the observed variability of GBHC led , , and to propose that, contrary to the prevailing notions, the size of the scattering region responsible for the X-ray emission is not $`R310R_S`$ but rather $`R\stackrel{>}{}10^3R_S`$, as implied by the PSDs and lag observations. Furthermore, the scattering medium (corona) is inhomogeneous, with the electron density following the law,
$$n(r)=\{\begin{array}{cc}n_1\hfill & \text{for }rr_1\hfill \\ n_1(r_1/r)^p\hfill & \text{for }r_2>r>r_1\hfill \end{array}$$
(1)
where $`r`$ is the radial distance from the center of the corona (assumed to be spherical) and $`r_1`$, $`r_2`$ are its inner and outer radii respectively. The index $`p>0`$ is a free parameter whose value depends on the specific dynamical model that determines the electron density. For example, the ADAF of suggest $`p=3/2`$ and $`T_e\stackrel{<}{}10^9`$ at radii as large as $`r(m_p/m_e)R_S`$, while models combining inflow and outflow , (ADIOS) allow in addition values $`p1`$. As pointed out in , most of the data analyzed to date suggest $`p=1`$ with $`p=3/2`$ also acceptable in certain cases.
It is intuitively obvious that scattering in the extended configuration of Eq. (1) produces time lags over a range of Fourier periods similar to the range of radii span by the hot corona: Scattering at a given radius $`R`$ increases the X-ray energy and introduces a lag $`\mathrm{\Delta }t\stackrel{<}{}R/c`$ between the scattered and unscattered photons, the lag appearing at a Fourier period $`P=R/c`$ (; N. Kylafis these proceedings). To compare this to the lags given in (which are the average over all such photon pairs scattered in a given decade in $`R`$, as a function of $`R`$) one has to multiply $`\mathrm{\Delta }t`$ by the probability of scattering at a given radius, $`𝒫(R)\tau (R)`$, with $`\tau (R)`$ the scattering depth over the radius $`R`$. For the density profile of Eq. (1), $`\tau (R)R^{p+1}`$ and since the Fourier period $`PR`$, $`\mathrm{\Delta }tR^{p+2}P^{p+2}`$. Monte Carlo simulations and analytical considerations have shown that these arguments are essentially correct . In addition, these simulations showed that the configuration of Eq. (1) produces light curves of high coherence over the range $`[r_2/c,r_1/c]`$ of the Fourier period, provided that the soft photons are injected near its center.
## 2 The Model Light Curves
Based on these considerations one can easily produce model light curves, whose properties in the time or the Fourier domain can then be compared to observations for consistency. The high coherence of the observations suggest injection of the soft photons at $`r<<r_2`$. The response function, $`g(t)`$, of a corona with density profile given by Eq. (1) to an instantaneous release of soft photons at $`rr_1`$ has been computed by a Monte Carlo simulation , . The result of a particular case is shown in Figure 1 along with an analytic fit. As pointed out in these functions can be approximated well by a Gamma function distribution i.e. $`g(t)t^{\alpha 1}e^{t/\beta }(0<\alpha <1,\beta >0)`$.
Assuming linearity, i.e. that $`T_e`$ is not affected by the photon flux, model light curves can be produced by an incoherent, random (Poisson) injection of shots of the form $`g(t)`$. The prescription for such a light curve is
$$F(t)=\underset{i=1}{\overset{N}{}}Q_i\theta \left(t\underset{i=1}{\overset{N}{}}t_i\right)g(tt_i).$$
(2)
$`Q_i`$ are the shot amplitudes, assumed constand, $`\theta (tt_1)`$ is the Heaviside function and $`t_i`$ is a collection of Poisson distributed time intervals obtained from the expression $`t_i=ft_0logR_i`$; $`R_i`$ is a random number uniformly distributed between 0 and 1, and $`f`$ is a real number, indicating the mean time between shots in terms of their rise time $`t_0`$. Figure 2 shows the light curve obtained using the above prescription with $`g(t)`$ as given in Figure 1 and $`f=3`$, while Figure 3 the light curve from an identical sequence of $`R_i`$’s and $`f=10`$.
These light curves look similar to those of BHC sources in their low, hard state . Due to the Poisson nature of $`t_i`$’s and the identical shape of the shots that make up these light curves, their PSDs are those of an individual shot, also shown in Figure 1 (dotted line). The form of the PSD is also quite similar to those of BHC (see ), with the high and low frequency breaks associated respectively with the inner $`r_1`$ and outer $`r_2`$ edges of the scattering corona.
One can easily produce, in addition, light curves for different photon energies, considering that for photons of a higher energy, $`g(t)`$ is approximately of the same form with only a small increase in the parameters $`\alpha `$ and $`\beta `$ . While the resulting light curves are visually identical to those of lower energies, their difference is nonetheless easily seen in the phases of the corresponding FFT. We have generated this way model light curves appropriate to three different energies and then computed the corresponding pairs of phase lags (a)-(b), (a)-(c) as a function of the Fourier frequency measured in Hz; the results are shown in Figure 4. The phase lags as presented in this figure are very similar to those of the sources analyzed in and in particular the X-ray transient GRO J0422+32 .
A correlation between the RMS variability and the value of $`f`$ is the most prominent feature of Figures 2 and 3. Because, in the linear regime, smaller $`f`$ is equivalent to higher luminosity (more photons per unit time), this correlation provides a natural account of the observed anticorrelation between the BHCs luminosity (in their hard spectral states) and their RMS variability. Furthermore, consideration of this model within its natural dynamic framework, namely that of ADAF, provides a direct correspondence between the dynamical and the timing properties of these systems: In ADAF all scales (and also $`r_1`$ and $`r_2`$) scale proportionally to $`v_{ff}\tau _{cool}`$ with $`v_{ff}`$ the free-fall velocity and $`\tau _{cool}`$ the local cooling time. For $`\tau _{cool}`$ inversely proportional to the local denisty (as is the case with ADAF), all length scales (and therefore frequencies) associated with the corona should depend on $`\dot{m}/\dot{m}_{Edd}`$, decreasing with increasing value of $`\dot{m}/\dot{m}_{Edd}`$; hence the corresponding frequencies should increase with increasing luminosity, a behavior which has apparently been observed in most accreting sources (whether neutron stars or black holes).
## 3 Conclusions
(a) The model of the extended inhomogeneous corona provides in a simple, well understood fashion model light curves for BHC which have morphology, PSDs and phase (or time) lags very similar to those observed. (b) The dependence of the phase lags on Fourier period allows the determination of the density profile of the corona; the data are consistent both with $`p=3/2`$ (ADAF) and $`p=1`$ (ADIOS) flows. (c) The dependence of the RMS variability and the PSD break frequencies of BHC on their luminosity are also in general agreement with these dynamical considerations and provide a well defined framework within which these ideas could be tested through more detailed comparison with observations. |
warning/0001/astro-ph0001224.html | ar5iv | text | # Error analysis for stellar population synthesis as an inverse problem.
## 1 Introduction
Important astrophysical issues could be solved if we were able to deduce the stellar population from the integrated light received from a far-away galaxy (or a region of a galaxy). Two methods try to solve this problem : the stellar data base synthesis and the evolutionary population synthesis. The first method is a more empirical approach, it relies strictly on the spectral data base of stars or stellar aggregates, while the second one is a more theoretical approach; it relies on our best knowledge on the formation and evolution of stars and stellar aggregates. Stellar data base synthesis is used in both inverse and forward modeling, while evolutionary population synthesis is mostly applied as a forward model. In our opinion the two methods are complementary and should be used jointly in an effort to solve this difficult problem. Both methods are subject to errors and ideally their results should always be presented with a set of acceptable solutions rather than just a solution. The purpose of this paper is to establish this set of acceptable solutions for the stellar data base synthesis.
Pelat (1997 and 1998; hereafter paper I and paper II) re-investigated the stellar data base synthesis method and proposed new and fast algorithms to search for physical solutions. The sensitivity of this solution to the observational errors was discussed in general terms: Monte-Carlo simulations were performed but no real error analysis had been done. This paper provides the user with practical formulæ which give an estimate of the domain of acceptable solutions around the solution.
In the following § 2, we recall briefly the solution we have proposed and then we focus in great detail on the question of error analysis.
## 2 Population synthesis as an inverse problem
First we specify what is meant by observables and parameters in the description of the physical model.
Observables. It is usually agreed to adopt equivalent widths as spectral observables, in order to avoid data reduction errors and to minimize extinction problems. These observables are noted here as the vector $`𝑾_{\text{obs}}`$. The aim of the stellar population synthesis is to find the combinations of stars which best reproduce the equivalent widths $`𝑾_{\text{obs}}`$ of the observed object (say, a galaxy).
Model parameters. The galaxy is described by the unknowns $`k_i`$. They give the proportional contribution to the luminosity due to stars of type $`i`$ at a reference wavelength $`\lambda _0`$. The assembly of $`k_i`$s is described by the vector $`𝒌`$.
The equation governing the stellar population synthesis problem is a non linear one, where a set of synthetic equivalent widths is a function of a set of stellar luminosity contributions $`k_i`$. We have:
$$W_{\text{syn}j}=\frac{_{i=1}^n_{}W_{ji}I_{ji}k_i}{_{i=1}^n_{}I_{ji}k_i},j=1,\mathrm{},n_\lambda .$$
(1)
Where, $`n_\lambda `$ is the number of observed equivalent widths; $`n_{}`$ is the number of stars used for the synthesis; $`W_{ji}`$ and $`I_{ji}`$ are respectively the equivalent width and continuum flux at wavelength $`\lambda _j`$ of stars of kind $`i`$. The continuum fluxes are normalized to one at the reference wavelength $`\lambda _0`$. For short we write Eq. (1): $`𝑾_{\text{syn}}=\phi (𝒌)`$.
In addition we have the condition that the contribution of a star to the luminosity cannot be negative and that the normalized proportions $`k_i`$ add up to one. A physical solution of Eq.(1) therefore lies in the vector set:
$$𝒮=\{(k_1,\mathrm{},k_n_{})|k_i0,_{i=1}^n_{}k_i=1\}.$$
(2)
This set $`𝒮`$ is a simplex (e.g. an equilateral triangle for a data base of three stars, a tetrahedron for a data base of four and so on). The set of all equivalent widths which are able to be exactly synthesized by the data base is the image of $`𝒮`$ by $`\phi `$. This is the synthetic domain $`𝒲`$ introduced in paper I:
$$𝒲=\{𝑾|𝑾=\phi (𝒌),𝒌𝒮\}.$$
(3)
If $`𝑾_{\text{obs}}𝒲`$ there is at least one exact solution $`𝒌`$ such that $`\phi (𝒌)=𝑾_{\text{obs}}`$.
The set $`𝒦`$ of all exact solutions is given formally by: $`𝒦=\phi ^1(𝑾_{\text{obs}})`$. This set may be empty. The solution set $`𝒦`$ is non-empty if there exists at least one vector $`𝒌`$ solution of the following system:
$$\begin{array}{c}\hfill 𝒌0,\\ \hfill \text{A}𝒌=\mathrm{𝟎},\\ \hfill 𝒃^T𝒌=1.\end{array}\}$$
(4)
The matrix elements of A are : $`[\text{A}]_{ji}=(W_{\text{obs}j}W_{ji})I_{ji}`$ and $`𝒃^T`$ is a line of ‘ones’. This system may or may not possess a solution.
### 2.1 Test for the existence of an exact solution
It is very easy to check if the above system (4) possesses at least one solution. If we add to it a linear equation to be maximized (e.g. $`𝒄^T𝒌=\mathrm{max}!`$), this system turns into a linear program. In fact, it is not necessary to explicit that supplementary linear equation, we only have to solve the linear program for a feasible solution (see for example Press et al. 1992 section 10.8). If no feasible solution exists this means that the galaxy cannot be exactly synthesized by the data base. This test is extremely rapid.
### 2.2 Under-determined case: $`n_\lambda n_{}1`$
If the galaxy can be exactly synthesized ($`𝒦`$ is non-empty), it has been demonstrated in paper II that $`𝒦`$ is a polytope, i.e. it is the convex hull of some finite extreme solutions. A solution $`𝒌`$ is extreme if it possesses at least $`n_01`$ components (stellar contributions) equal zero. The key parameter $`n_0`$ is the dimension of the null space associated with matrix A. The data base is non-degenerate for the observation if we have $`n_0=n_{}n_\lambda `$.
If the galaxy cannot be synthesized, an approximate solution is found with a data base reduced to at most $`n_\lambda `$ stars. The procedure is described in the next section.
### 2.3 Overdetermined case: $`n_\lambda n_{}`$
In the overdetermined case, there is most probably no exact solution of (4), the observation is not on the synthetic surface and one must content oneself with an approximate solution. This approximate solution $`𝒌`$ is usually accepted in the least square meaning, that is $`𝒌`$ must minimize a quadratic form :
$$D^2=(𝑾_{\text{obs}}\phi (𝒌))^T𝚺^1(𝑾_{\text{obs}}\phi (𝒌)),$$
(5)
where $`𝚺^\mathbf{}\mathrm{𝟏}`$ is a positive definite ‘weight’ matrix. It has been shown in paper I that a very successful estimate of $`𝒌`$ is found near the ‘first-guess’ $`𝒌_0`$:
$$𝒌_0=(\text{B}^T\text{B})^1𝒃[𝒃^T(\text{B}^T\text{B})^1𝒃]^1,$$
where B is equal to the matrix A augmented with the line $`𝒃^T`$. It is shown in appendix A how one can get rid of the positivity constraint $`𝒌\mathrm{𝟎}`$ when searching for a minimum in the neighbourhood of $`𝒌_0`$. The estimate $`𝒌_0`$ is the unique solution of the problem if the matrix $`𝚺^1`$ is diagonal with diagonal elements equal to $`I_{\text{syn}j}^2`$ ($`I_{\text{syn}j}=_{i=1}^n_{}I_{ji}k_i`$). It was argued in paper I that a reasonable weight matrix should not be very different from this particular diagonal matrix.
## 3 Purposes of the error analysis
An observation of a galaxy is not just the point $`𝑾_{\text{obs}}`$ in $`W`$-space, it includes all points interior to an error zone around the observation. If we assume the observational errors to be Gaussian, this error zone is an hyper-ellipsoid around $`𝑾_{\text{obs}}`$. We have observed $`𝑾_{\text{obs}}`$ but we consider that we could have observed, with probability $`\gamma `$, any other point at the interior of this hyper-ellipsoid. This set of ‘probable’ points: $`_\gamma `$, is formally defined as
$$_\gamma =\{𝑾|(𝑾𝑾_{\text{obs}})^T\text{V}^1(𝑾𝑾_{\text{obs}})F_{\chi ^2}^1(\gamma )\},$$
(6)
where V is the variance-covariance matrix of the observation and $`F_{\chi ^2}`$ is the distribution function of a $`\chi ^2`$ variate with $`n_\lambda `$ degrees of freedom. For short, we call $`_\gamma `$ the ‘error ellipsoid’. We recall that V is equal to $`\mathrm{\Delta }𝑾\mathrm{\Delta }𝑾^T`$, where $``$ stands for the expectation and $`\mathrm{\Delta }𝑾`$ for a variation of $`𝑾`$ around its mean.
The purpose of the error analysis is to determine how the observational errors are transformed by the inversion process. Usually one tries to solve two problems as follows: i) deduce $`\text{V}_k`$ the variance-covariance matrix of the solution knowing the matrix V for the observational errors; ii) construct $`𝒦_\gamma `$, the set of all solutions that fall within the ‘error ellipsoid’, i.e. $`𝒦_\gamma =\phi ^1(_\gamma )`$.
These problems are simplified if the error ellipsoid corresponds to small deviation $`d𝑾`$ around the observation. In that approximation one can linearize $`\phi ^1`$, which implies that $`𝒦_\gamma `$ is also an hyper-ellipsoid. We have:
$$𝒦_\gamma \{𝒌^{}|(𝒌^{}𝒌)^T\text{P}(𝒌^{}𝒌)F_{\chi ^2}^1(\gamma )\}.$$
(7)
The matrix P would be equal to $`\text{V}_k^1`$ if $`\text{V}_k`$ were invertible but this is not the case here: the matrix $`\text{V}_k`$ is singular because of the normalization constraint on $`𝒌`$. Using the definition of $`𝒦_\gamma `$ (7), one is able to test if an alternative solution $`𝒌^{}`$ is acceptable at the confidence level $`\gamma `$.
The results of the error analysis are given in § 5. Before that, we give in § 3.1 an overview of the main processes at the origin of the observational errors and in § 4 we introduce information content held by an observation with regard to the synthesis problem we want to solve.
### 3.1 Origin of the errors
Errors on equivalent widths from spectra of galaxies have several origins. The most important one is induced by the empirical process of continuum intensity plotting. This leads to a wide uncertainty on equivalent widths estimates, caused by the fact that the continuum intensity lies on the denominator in the equivalent width’s expression. Contributing to the uncertainties in this empirical approach are the blending of the absorption lines plus the possible presence of emission lines as in active galactic nuclei spectra. The limited wavelength range of the study plays also a crucial role in this domain.
Another phenomenon in active galactic nuclei spectra is the presence of an additional non stellar continuum which leads to the dilution of lines, i.e. the reduction of equivalent widths in comparison with those of normal galaxies. The effect of this non stellar continuum can be estimated and equivalent widths can be corrected; but obviously, this estimation is also subject to errors. Finally, velocity dispersion in the galaxies broadens lines and is also a source of errors. Considering these facts, typical errors of $`10\%`$ are present in the equivalent widths of strong lines.
We did not take into account the errors in the equivalent width of the lines in the data base itself because they are supposed to be negligible compared with errors in the equivalent width of the lines from the galaxy. However we provide the user in § B.1.3 with a test able to validate this hypothesis. If these errors were not negligible, one should not perform the synthesis.
## 4 Information content of an observation.
It is clear that if the synthetic domain $`𝒲`$ is entirely contained within the observational error $`_\gamma `$, the observation cannot discriminate, at the level $`\gamma `$, between the different stellar populations. All possible models which fall within $`_\gamma `$ must be considered as indistinguishable. In other words, the observation bears not enough information to differentiate between various possible stellar populations of the galaxy. The size of the error ellipsoid $`_\gamma `$ is in some way the resolution at which we try to sort between the different models.
On the contrary, if the observation is so good that one can consider the ‘error-bars’ to be null, the observation indeed convey a maximum information: $`𝑾_{\text{obs}}`$ must be synthesized by the data base. Under this extreme hypothesis, $`_\gamma `$ is reduced to the single point $`\{𝑾_{\text{obs}}\}`$ which has a zero measure in $`𝒲`$.
Following a suggestion of R. Barrett we define the information $`I_\gamma `$ brought by the observation as being a function of the measure of $`_\gamma `$ within $`𝒲`$. More precisely $`I_\gamma `$ will be a function of the probability $`p_\gamma `$ that the image by $`\phi `$ of a point drawn at random from the a priori distribution of the stellar populations falls within the error ellipsoid. That is:
$$p_\gamma =\underset{_\gamma 𝒲}{}\pi (𝒘)𝑑𝒘,𝒘=\phi (𝒌),$$
(8)
where $`\pi `$ is the a priori probability density function of the equivalent widths. We shall evaluate this $`n_\lambda `$-dimensional integral by a Monte-Carlo method (see appendix A).
The probability $`p_\gamma `$ may be zero under two circumstances: (i) $`𝑾_{\text{obs}}`$ is within the synthetic domain and $`_\gamma `$ is reduced to $`\{𝑾_{\text{obs}}\}`$ only; (ii) the observation and its associated error domain are outside the synthetic domain: $`_\gamma 𝒲=\mathrm{}`$. In each case we have maximum information of different nature conveyed by the observation: (i) we know that the true stellar population lies in the solution set (provided that all stars present in the galaxy are also present in the data base); (ii) we know that the data base is incomplete, more stars must be added.
In a first approach, it seems reasonable to define the information brought by an observation by: $`I_\gamma =1p_\gamma `$, i.e. it is the probability that a stellar population drawn at random from the simplex $`𝒮`$ induces equivalent widths that do not fall within the error domain. By ‘at random’ we mean according to $`\pi `$, the a priori distribution of the equivalent widths. In practice we use the transform by $`\phi `$ of an a priori distribution of the stellar populations which is uniform over $`𝒮`$. Defined like this, $`I_\gamma `$ is zero if all models are validated by the observation (no information) and one if the solution set has a zero probability to be reached by a random model (maximum information, one cannot get there at random).
An illustration of the information content within an observation is given in Fig 1, see also Fig. 4 of paper II. Here, at the 1-$`\sigma `$ confidence level ($`\gamma =0.683`$) the information content is $`I_\gamma =0.92`$. This observation has therefore a high information content. It is located on a part of the synthetic domain, which is seldom reached by a random synthesis following an a priori uniform distribution in $`𝒌`$.
## 5 Error analysis in practice.
In this section, we limit ourselves to the practical results of the error analysis. We refer to appendix B for all the computations needed to derive the various matrices which appear below.
The key parameter here is the number $`n_\text{s}`$ of stars necessary to define the solution (or the extreme solutions). In the overdetermined case $`n_\text{s}=n_{}`$, while in the under-determined case $`n_\text{s}=n_\lambda +1`$ if the galaxy can be synthesized and $`n_\text{s}=n_\lambda `$ if it is not. It is, in both under- and overdetermined cases, less or equal to $`n_\lambda +1`$. We have a regular case if $`n_\text{s}=n_\lambda +1`$ and a singular one if $`n_\text{s}<n_\lambda +1`$. We consider separately these two cases. In both cases, V stands for the variance-covariance matrix of the observations and $`𝒔`$ for the list of the indices of the stars retained in the solution.
### 5.1 Regular case $`n_\text{s}=n_\lambda +1`$.
Note that in this case the galaxy belongs to the synthetic domain. A solution $`𝒌`$ is any barycentric combination of $`n_K`$ extreme solutions denoted by $`𝒌_s`$ ($`n_K`$ may be equal to one). At least $`n_{}(n_\lambda +1)`$ components of $`𝒌_s`$ are equal to zero. The components of $`𝒌_s`$ are in addition subject to errors, due to the presence of errors in the data. In the tangent approximation (small errors), the variance-covariance matrix of the extreme solutions is a singular matrix given by:
$$\text{V}_{k_s}=\text{K}(\begin{array}{cc}\text{V}\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{K}^T,\text{K}=\text{W}^1.$$
(9)
The elements of the matrix W are:
$$\begin{array}{ccc}[\text{W}]_{ji}=\hfill & (W_{\text{obs}j}W_{js(i)})\frac{I_{js(i)}}{I_{\text{syn}j}},\hfill & j=1,\mathrm{},n_s1;\hfill \\ \hfill =& 1,\hfill & j=n_s,\hfill \end{array}$$
(10)
where $`𝒔(i),i=1,\mathrm{},n_\text{s}`$ run over the stars retained in the solution. The set of acceptable extreme solutions is an ellipsoid centered around the extreme solution $`𝒌_s`$. Its characteristic matrix $`\text{P}_s`$ is given by:
$$\text{P}_s=\text{W}^T(\begin{array}{cc}\text{V}^1\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{W},\text{W}=\text{J}^1\text{B},$$
(11)
where J is a $`n_\lambda +1`$ by $`n_\lambda +1`$ diagonal matrix:
$$\begin{array}{ccc}\hfill [\text{J}]_{jj}=& I_{\text{syn}j},\hfill & j=1,\mathrm{},n_\lambda ;\hfill \\ \hfill [\text{J}]_{jj}=& 1,\hfill & j=n_\lambda +1.\hfill \end{array}$$
The set of acceptable solutions may be considered as the convex hull of all these acceptable extreme solutions.
We have illustrated in Fig. 1 the error analysis performed on an observation subject to errors of $`10\%`$. The population synthesis was done with the same data base used in paper II. The results are given in Table 1. The $`𝒌_s,s=1,2,3`$ are the three extreme solutions; $`\sigma 𝒌_s`$ is the standard deviation vector associated with the solution (it is the square root of the diagonal of $`\text{V}_{k_s}`$); $`\mathrm{\Delta }`$ is the extreme range attained by the $`k_i`$s by a point, which is constrained to move on the error ellipsoid. The analysis was done at the confidence level of 1-$`\sigma `$, that is: $`\gamma 0.683`$.
#### 5.1.1 Merit order among extreme solutions
The product of the non-zero eigenvalues of $`\text{V}_k`$ is proportional to the surface of the ellipsoid $`𝒦_\gamma `$. This quantity allows to sort the extreme solutions in order of merit, from the smallest surface (highest merit) to the greatest surface (smallest merit). In a synthesis with many extreme solutions it is advisable to retain only the solutions which possess the highest merit.
### 5.2 Singular case $`n_\text{s}<n_\lambda +1`$.
In the singular case, an approximate solution is searched by minimizing a distance from the observation to the synthetic domain. This distance is usually the elliptical distance $`D`$, defined by Eq. (5), and it requires a positive definite matrix $`𝚺`$. However, we suppose below that $`D`$ is the Euclidean distance through the variable transformation described in appendix B.2.5. The ellipse has been transformed into the unit circle and $`𝚺=\text{I}`$.
The error analysis is complicated by the fact that one must perform a projection on a plane which is tangent to the synthetic domain around the approximate solution. We describe step by step the operations which lead to the result.
First, construct the $`n_\lambda `$ by $`n_\text{s}`$ matrix $`\dot{\text{A}}`$ of elements: $`[\dot{\text{A}}]_{js(i)}=(W_{\text{syn}j}W_{js(i)})I_{js(i)}`$. The index $`s(i)`$ runs over the $`n_\text{s}`$ stars defining the solution. We need also the diagonal matrix $`\text{J}_A=\text{diag}(I_{\text{syn}1},\mathrm{},I_{\text{syn}n_\lambda })`$.
The matrix $`\text{G}=\text{J}_A^1\dot{\text{A}}`$ is constructed. A rank deficiency QR algorithm is performed on G (e.g. using dgeqpf from lapack). Following this operation one gets: $`\text{G}𝚷=\text{Q}\text{R}`$. The matrix Q is a $`n_\lambda `$ by $`n_\lambda `$ square matrix partitioned into two blocks: $`\text{Q}=(\text{p}|\text{q})`$. The first block is formed by the $`n_\text{s}1`$ first columns of Q. An orthogonal projector $`\text{H}=\text{pp}^T`$ is constructed.
We define the matrix $`\dot{\text{B}}`$ equal to $`\dot{\text{A}}`$ with a supplementary line of ‘ones’ and the matrix $`\text{K}=(\dot{\text{B}}^T\dot{\text{B}})^1\dot{\text{B}}^T\text{J}`$ where J has already been defined ($`\text{J}=\text{diag}(I_{\text{syn}1},\mathrm{},I_{\text{syn}n_\lambda },1)`$ ).
The variance-covariance matrix of the solution is:
$$\text{V}_k=\text{K}(\begin{array}{cc}\text{H}\text{V}\text{H}^T& \mathrm{𝟎}\\ \mathrm{𝟎}^T& 0\end{array})\text{K}^T.$$
We define the oblique projector M:
$$\text{M}=\text{p}\text{p}^T+\text{q}\text{q}^T\text{V}\text{p}(\text{p}^T\text{V}\text{p})^1\text{p}^T,$$
and compute: $`\text{W}=\text{J}^1\dot{\text{B}}`$. The matrix P entering the definition of the acceptance zone in Eq. (7) is given by:
$$\text{P}=\text{W}^T(\begin{array}{cc}\text{M}^T\text{V}^1\text{M}& \mathrm{𝟎}\\ \mathrm{𝟎}^T& 0\end{array})\text{W},\text{W}=\text{J}^1\dot{\text{B}}.$$
Note that in the regular case we have: $`\text{K}=\text{B}^1\text{J}`$ and in the singular case we have: $`\text{K}=(\text{B}^T\text{B})^1\text{B}^T\text{J}`$. Therefore K is always the least square solution of the linear system $`\text{B}\text{K}=\text{J}`$. (In both cases B is constructed with $`W_{\text{syn}}`$ and the stars contributing to the synthesis.)
## 6 The example of NGC 3521.
To illustrate our study we take, as an example, the clusters of Bica’s data base (see Bica 1988) and its galaxy S3 (NGC 3521). As our aim is to visualize the error zone, we shall use only two lines (i.e. $`n_\lambda =2`$) this means two degrees of freedom for the error analysis. (Indeed our analysis is valid for any number of lines.) From the 35 clusters constituting the data base, we select according to Bica’s results the eight clusters (i.e. $`n_{}=8`$) contributing to the synthesis. We have chosen the two largest equivalent widths of the galaxy which correspond to the Ca II K line ($`\lambda _K=3933`$Å) and the CH G band ($`\lambda _G=4301`$Å); the reason of this choice is that the relative errors on the strongest lines are generally the smallest. Typical errors on such equivalent widths are in the order of $`10\%`$, but for the sake of argument, we take an error of only $`3\%`$. We also assume, that there is no correlation between equivalent widths errors ($`\rho =0`$) even if this is probably not the case in reality.
Fig. 2 shows that the galaxy can be synthesized exactly. Our method gives 16 different extreme solutions. Table 2 shows these several extreme solutions $`𝒌_s`$ where $`s=1,\mathrm{},16`$; also shown are the standard deviations vectors $`\sigma 𝒌_s`$ corresponding to 1-$`\sigma `$ confidence level; and the standard deviation $`\sigma ^{}𝒌_s`$ taking into account a $`1\%`$ error in the data base. The solutions are sorted by increasing order of merit (i.e. surface<sup>-1</sup>).
Figure 2 demonstrates clearly the impact of the synthetic domain structure $`D`$ on the solution errors. Indeed the clusters constituting the data base are nearly aligned in $`W`$-space and their continuum intensities are such that the synthetic domain is compressed (i.e the lines joining the clusters are very close and nearly parallel to each others). As a consequence large errors are present in the extreme solutions (see also figure 3). It is important to bear this situation in mind when a data base of clusters (or stars) is chosen. Ideally the resolution at which the equivalent widths space is sampled should be adapted to the quality of the observations. Better observations allow a finer coverage of the $`W`$-space i.e. more clusters (or stars) may be added. We see in table 2 that the contribution of data base errors to the solution errors is negligible. Finally, the best solution (as defined in §5.1.1) is solution Nr. 1; the solutions 1, 4 and 11 are the three solutions among the sixteen extreme solutions that agree well with Bica’s solution (see $`𝒌_{\text{Bica}}`$ in Tab. 2). This result is very satisfying, because we used minimal information by considering only two equivalent widths.
## 7 Conclusion.
We provide explicit formulæ to perform a complete error analysis on stellar population synthesis by means of a data base. The hypothesis underlying the results are that errors on the data base itself are negligible compared with errors on the observed galaxy. The method also supposes that the errors are reasonably Gaussian and further demands the knowledge of the error matrix (the variance-covariance matrix) of the observed equivalent widths.
The results provided are (i) the variance-covariance matrix of the extreme solutions (regular case) or of the approximate solution (singular case); (ii) the matrix defining a zone in the solution space where the stellar populations are indistinguishable at a confidence level $`\gamma `$.
We also introduced information contained in an observation with respect to the problem to be solved. This information is evaluated by a Monte-Carlo method using as input the a priori distribution of the stellar population. We further designed a method to generate a uniform a priori stellar population distribution. The same method allows us to transform the complex hyper-tetrahedron boundary constraints into simple bounds constraints. This transformation simplifies considerably the search, when needed, for a minimum in the stellar population space.
## Acknowledgments
We wish to thank Yvon Biraud, Catherine Boisson, Silvano Bonazzola, Monique Joly, Jean-Alain Marck and Dr. Yuen Keong Ng whose comments about this paper have been so helpful. Special thanks go to Dr. Richard Barrett who suggested the concept of information developed here.
## Appendix A Uniform mapping of a simplex
In this appendix we show how to transform an hyper-cube into an hyper-tetrahedron. This operation is continuous and transforms a uniform density on the hyper-cube into a density equally uniform on the hyper-tetrahedron.
The transformation $`\psi `$ allows a simplification of the constraints $`𝒌\mathrm{𝟎}`$ subject to $`_ik_i=1`$ ($`𝒌𝒮`$) through the mapping $`𝒌=\psi (𝒖)`$ of the variables $`u_i`$ subject to simple bounds constraints of the type: $`0u_i1`$. At the same time the uniform coverage of the hyper-tetrahedron $`𝒮`$ permits to evaluate the information integral (8).
We also want the change of variables $`\psi `$ to be continuous and twice derivable, so that $`\psi `$ can be used in an optimization program where the variables are subject to the constraints $`𝒌𝒮`$ as needed in § 2.3.
One can think of several ways to map $`𝒌=\psi (𝒖)`$, but if the transformation is not chosen carefully there is a risk that $`\psi `$ selects preferentially certain zones of the $`𝒌`$ parameters space. In order to avoid this possible bias, we would like to find a change of variable ensuring that a uniform coverage of the $`𝒖`$ space (the cube) transforms to a uniform coverage of the $`𝒌`$ space (the tetrahedron). We found two methods satisfying these requirements.
### A.1 Method 1.
Let us consider the $`n_{}`$ independent random variables $`K_i`$, following all the same exponential distribution. This distribution is, for $`k0`$ a probability density function: $`f(k)=\mathrm{exp}(k)`$. The joint distribution of the multiplet $`(K_1,\mathrm{},K_n_{})`$ has the density $`f_n_{}(k_1,\mathrm{},k_n_{})=_{i=1}^n_{}f(k_i)=\mathrm{exp}(_{i=1}^n_{}k_i)`$ and is constant on $`_{i=1}^n_{}k_i=a`$. The conditional density on $`_{i=1}^n_{}k_i=a`$ is therefore uniform for any $`a>0`$, in particular for $`a=1`$.
Now the random variable $`U_i=_0^{K_i}f(k)𝑑k=1\mathrm{exp}(K_i)`$ is uniformly distributed on $`0u_i<1`$ so, $`K_i=\mathrm{ln}(1U_i)`$ or equivalently $`K_i=\mathrm{ln}(U_i)`$ is exponentially distributed. In conclusion, the mapping $`\psi `$ defined by:
$$k_i=\mathrm{ln}(u_i),i=1,\mathrm{},n_{},0<u_i1,$$
(12)
ensures: (i) $`k_i0`$; (ii) a uniform coverage of the $`k_i`$ domain; and (iii) that $`u_i`$ is only subject to simple bounds constraints $`0u_i<1,i=1,\mathrm{},n_{}`$.
The only problem with this method is that we use $`n_{}`$ parameters while only $`n_{}1`$ are necessary. The search algorithm may loose time exploring those lines in $`𝒖`$ space where $`W_{\text{syn}j}`$ does not change because the $`k_i`$ differ only by a scaling factor. We thus designed ‘method 2’ in order to solve this problem.
### A.2 Method 2.
Here we want to map the $`k_i`$ using only $`n_{}1`$ parameters of type $`u_i`$. Again the $`K_i`$ are exponentially distributed as in ‘method 1’; this ensures, as shown above, that the density is uniform on $`_{i=1}^n_{}k_i=1`$.
We now introduce the new variables $`Q_i`$. For the sake of argument we limit ourself to a $`n_{}=4`$ example, but it should be clear that the method works for any $`n_{}>0`$. We define:
$$\begin{array}{ccc}\hfill Q_1=& K_4+K_3+K_2+K_1,\hfill & 0Q_1<\mathrm{};\hfill \\ \hfill Q_2Q_1=& K_4+K_3+K_2,\hfill & 0Q_21;\hfill \\ \hfill Q_3Q_2Q_1=& K_4+K_3,\hfill & 0Q_31;\hfill \\ \hfill Q_4Q_3Q_2Q_1=& K_4,\hfill & 0Q_41.\hfill \end{array}$$
The Jacobian of this change of variables, $`J_4={\displaystyle \frac{(k_1,\mathrm{},k_4)}{(q_1,\mathrm{},q_4)}}`$, is given by: $`J_4=q_1^3q_2^2q_3`$. The four values $`(Q_1,Q_2,Q_3,Q_4)`$ have the density:
$$f(q_1,q_2,q_3,q_4)=e^{q_1}q_1^3q_2^2q_3.$$
(13)
This demonstrates that the $`Q_i`$ are independent and have densities:
$$\begin{array}{ccc}f_1(q_1)=\hfill & \frac{1}{3!}q_1^3e^{q_1},\hfill & 0q_1<\mathrm{};\hfill \\ f_2(q_2)=\hfill & 3q_2^2,\hfill & 0q_2<1;\hfill \\ f_3(q_3)=\hfill & 2q_3,\hfill & 0q_3<1;\hfill \\ f_4(q_4)=\hfill & 1,\hfill & 0q_4<1.\hfill \end{array}\}$$
(14)
The variable $`Q_1`$ follows a gamma distribution, (as required for a sum of independent exponential random variables) and the other variables follow a power distribution of decreasing index, the last one being uniform.
Now if $`Q_1`$ is given, say $`Q_1_{i=1}^n_{}K_i=1`$, the remaining $`Q_2`$, $`Q_3`$ and $`Q_4`$ cover uniformly (because the $`K_i`$ are exponential) the hyper-tetrahedron defined by (2). We have:
$$\begin{array}{cc}K_4=\hfill & Q_4Q_3Q_2,\hfill \\ K_3=\hfill & (1Q_4)Q_3Q_2,\hfill \\ K_2=\hfill & (1Q_3)Q_2,\hfill \\ K_1=\hfill & 1Q_2.\hfill \end{array}\}$$
(15)
Finally if we define:
$$U_n=\underset{0}{\overset{Q_{n_{}n+1}}{}}nq^{n1}𝑑q=Q_{n_{}n+1}^n,n=1,\mathrm{},n_{}1,$$
(16)
the variables $`U_n`$ are uniform and independent. We then have a power distribution of index $`n`$: $`Q_{n_{}n+1}=U_n^{\frac{1}{n}}`$ (for $`n=1,\mathrm{},n_{}1`$) when $`U_n`$ is uniformly distributed in $`[0,1]`$. Therefore the change of variables $`\psi `$:
$$\begin{array}{cc}k_4=\hfill & u_1u_2^{\frac{1}{2}}u_3^{\frac{1}{3}},\hfill \\ k_3=\hfill & (1u_1)u_2^{\frac{1}{2}}u_3^{\frac{1}{3}},\hfill \\ k_2=\hfill & (1u_2^{\frac{1}{2}})u_3^{\frac{1}{3}},\hfill \\ k_1=\hfill & 1u_3^{\frac{1}{3}},\hfill \end{array}\}$$
(17)
maps the $`𝒌`$ space uniformly if $`u_1,u_2`$ and $`u_3`$ are drawn from three independent random variables $`U_1,U_2,U_3`$ uniformly distributed in $`[0,1]`$. Now for any $`n_{}>1`$, $`\psi `$ is defined by:
$$k_1=1u_{n_{}1}^{1/(n_{}1)},k_n_{}=\underset{i=1}{\overset{n_{}1}{}}u_i^{\frac{1}{i}},$$
(18)
and for $`n=2,\mathrm{},n_{}1`$:
$$k_n=[1u_{n_{}n}^{1/(n_{}n)}]\underset{i=n_{}n+1}{\overset{n_{}1}{}}u_i^{\frac{1}{i}},$$
(19)
where all $`u_i`$ are subject to the constraint: $`0u_i<1`$. We present in Fig. 4 an example of such a mapping for $`n_{}=3`$.
## Appendix B Computations needed for the error analysis.
We would like to establish a relationship between a small variation $`d𝑾`$ around $`𝑾_{\text{obs}}`$ and the resulting variation $`d𝒌`$ around the solution $`𝒌`$. The variation $`d𝒌`$ is constrained to $`_{i=1}^n_{}dk_i=0`$ by $`_{i=1}^n_{}k_i=1`$.
### B.1 The galaxy can be synthesized.
$`𝑾_{\text{obs}}`$ belongs to the synthetic domain ($`𝑾_{\text{obs}}𝒲`$) if the galaxy can be synthesized and the many solutions are contained within the convex hull of a set of extreme solutions (see paper II). An extreme solution has by definition at least $`n_{}(n_\lambda +1)`$ components $`k_i`$ equal to zero, what remains is the evaluation of $`n_\lambda +1`$ components of $`𝒌`$.
#### B.1.1 Error analysis of the extreme solutions.
We discard the stars with zero contribution and set $`n_{}=n_\text{s}`$ (recall that $`n_\text{s}=n_\lambda +1`$). This allows us to keep the same notation. Note that the new matrix A is formed of the $`n_\lambda +1`$ columns of the original matrix in system (4) corresponding to the stars which have not been set to zero. In the same way $`\phi `$ stands here for the original $`\phi `$ restricted to these stars.
By adding the line $`𝒃^T`$ to A we form a $`n_\lambda +1`$ by $`n_\lambda +1`$ matrix B and $`\text{B}^1`$ exists, because the problem has an (extreme) solution. We have:
$$𝒌=\text{B}^1(\begin{array}{c}\mathrm{𝟎}\\ 1\end{array}),\text{B}=(\begin{array}{c}\text{A}\hfill \\ 𝒃^T\hfill \end{array}).$$
(20)
From here we have $`d\text{B}𝒌+\text{B}d𝒌=\mathrm{𝟎}`$, the matrix $`d\text{B}`$ is $`d\text{A}`$ supplemented with a line of zeros. We have $`dA_{ji}=dW_jI_{ji}`$ and $`[d\text{A}𝒌]_j=dW_j_{i=1}^n_{}I_{ji}k_i=dW_jI_{\text{syn}j}`$. Let us introduce the diagonal matrix J, where $`J_{jj}=I_{\text{syn}j}`$ for $`j=1,\mathrm{},n_\lambda `$ and the last diagonal element is $`J_{n_\lambda +1,n_\lambda +1}=1`$. We then have:
$$\text{B}d𝒌=d\text{B}𝒌=\text{J}(\begin{array}{c}d𝑾\\ 0\end{array}).$$
(21)
As none of the continua $`I_{ji}`$ are null, $`\text{J}^1`$ exists and we define $`\text{W}=\text{J}^1\text{B}`$. The matrix W has a very simple form. For example: if we have $`n_\lambda =2`$, the number of stars of an extreme ray is $`n_{}=3`$ and W equals:
$$(\begin{array}{ccc}\left(W_{\text{obs}1}W_{11}\right){\scriptscriptstyle \frac{I_{11}}{I_{\text{syn}1}}}& \left(W_{\text{obs}1}W_{12}\right){\scriptscriptstyle \frac{I_{12}}{I_{\text{syn}1}}}& \left(W_{\text{obs}1}W_{13}\right){\scriptscriptstyle \frac{I_{13}}{I_{\text{syn}1}}}\\ \left(W_{\text{obs}2}W_{21}\right){\scriptscriptstyle \frac{I_{21}}{I_{\text{syn}2}}}& \left(W_{\text{obs}2}W_{22}\right){\scriptscriptstyle \frac{I_{22}}{I_{\text{syn}2}}}& \left(W_{\text{obs}2}W_{23}\right){\scriptscriptstyle \frac{I_{23}}{I_{\text{syn}2}}}\\ 1& 1& 1\end{array}).$$
If we define $`\text{K}=\text{W}^1=\text{B}^1\text{J}`$, equation (21) can be written as
$$\text{W}d𝒌=(\begin{array}{c}d𝑾\\ 0\end{array})\text{or}d𝒌=\text{K}(\begin{array}{c}d𝑾\\ 0\end{array}).$$
(22)
We can now compute the variance-covariance matrix of the $`𝒌`$’s. Let $`\text{V}_k`$ be that matrix. In the tangent approximation we have $`\text{V}_k=d𝒌d𝒌^T`$:
$$\begin{array}{cc}\text{V}_k=\hfill & \text{K}(\begin{array}{c}d𝑾\\ 0\end{array})(d𝑾^T\mathrm{\hspace{0.17em}0})\text{K}^T,\hfill \\ \hfill =& \text{K}(\begin{array}{cc}d𝑾d𝑾^T& \mathrm{𝟎}\\ \mathrm{𝟎}^T& 0\end{array})\text{K}^T.\hfill \end{array}$$
If V designates the variance-covariance matrix of the observation we finally have:
$$\text{V}_k=\text{K}(\begin{array}{cc}\text{V}\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{K}^T.$$
(23)
As a bonus, we note that it is possible to express the solution $`𝒌`$ in terms of K. We have $`\text{B}=\text{J}\text{W}`$, $`\text{B}^1=\text{K}\text{J}^1`$ and the solution (20) can be expressed by:
$$𝒌=\text{K}\text{J}^1(\begin{array}{c}\mathrm{𝟎}\\ 1\end{array})=\text{K}(\begin{array}{c}\mathrm{𝟎}\\ 1\end{array}).$$
#### B.1.2 Acceptance region of the extreme solutions
At the level $`\gamma `$, all points around the solution $`𝒌`$ are indistinguishable and give an image close to the observation $`𝑾_{\text{obs}}`$. The acceptable solutions $`𝒌^{}`$ are in a set $`𝒦_\gamma `$ called the ‘acceptance region’. We define: $`𝒦_\gamma =\phi ^1(_\gamma )`$.
In the tangent approximation around $`𝒌`$, we have:
$$𝒦_\gamma =\{𝒌^{}|(𝒌^{}𝒌)^T\text{P}(𝒌^{}𝒌)F_{\chi ^2}^1(\gamma )\},$$
where, as in Eq. (6), $`F_{\chi ^2}`$ is the repartition function of a $`\chi ^2`$ variate possessing $`n_\lambda `$ degrees of freedom. That is also the equation of an ellipsoid around $`𝒌`$. The ‘weight’ matrix P is usually equal to the inverse of the variance-covariance matrix $`\text{V}_k`$. The problem is that $`\text{V}_k`$, as it is clear from (23), does not possess an inverse. Back to the definition of $`\phi ^1(_\gamma )`$ we have:
$$\begin{array}{cc}𝒦_\gamma =\hfill & \{𝒌^{}|[\phi (𝒌^{})\phi (𝒌)]^T\text{V}^1[\phi (𝒌^{})\phi (𝒌)]F_{\chi ^2}^1(\gamma )\},\hfill \\ \hfill =& \{𝒌^{}|d𝑾^T\text{V}^1d𝑾F_{\chi ^2}^1(\gamma )\},\hfill \end{array}$$
where we have set $`d𝑾=\phi (𝒌^{})\phi (𝒌)`$. The above expression is equivalent with:
$$\begin{array}{cc}𝒦_\gamma =\hfill & \{𝒌^{}|(d𝑾^T0)(\begin{array}{cc}\text{V}^1\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})(\begin{array}{c}d𝑾\\ 0\end{array})F_{\chi ^2}^1(\gamma )\}.\hfill \end{array}$$
The choice of the extra elements in the central matrix is arbitrary. We chose zero for simplicity. We can now use Eq. (22) and write:
$$𝒦_\gamma =\{𝒌^{}|(𝒌^{}𝒌)^T\text{W}^T(\begin{array}{cc}\text{V}^1\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{W}(𝒌^{}𝒌)F_{\chi ^2}^1(\gamma )\}.$$
Therefore:
$$\text{P}=\text{W}^T(\begin{array}{cc}\text{V}^1\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{W}.$$
#### B.1.3 Contributions to errors coming from stellar library
Let’s call $`\text{J}_{Wi}`$ and $`\text{J}_{Ii}`$ the diagonal matrices such that for $`j=1,\mathrm{},n_\lambda `$ we have $`[\text{J}_{Wi}]_{jj}=I_{ji}k_i`$ and $`[\text{J}_{Ii}]_{jj}=(W_{\text{syn}j}W_{ji})k_i`$. In addition we have for $`j=n_\lambda +1`$ that $`[\text{J}_{Wi}]_{jj}=[\text{J}_{Ii}]_{jj}=1`$. Define further $`\text{K}_{Wi}=\text{B}^1\text{J}_{Wi}`$ and $`\text{K}_{Ii}=\text{B}^1\text{J}_{Ii}`$ Following the same reasoning as in section B.1.1 we find that the contribution of the stellar library to the total variance-covariance matrix of a solution is:
$$\text{V}_{k()}=\underset{i=1}{\overset{n_\lambda +1}{}}\{\text{K}_{Wi}(\begin{array}{cc}\text{V}_{Wi}\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{K}_{Wi}^T+\text{K}_{Ii}(\begin{array}{cc}\text{V}_{Ii}\hfill & \mathrm{𝟎}\\ \mathrm{𝟎}^T\hfill & 0\end{array})\text{K}_{Ii}^T\}.$$
Where $`\text{V}_{Wi}`$ and $`\text{V}_{Ii}`$ are the variance-covariance matrices of the $`W_{ji}`$ and $`I_{ji}`$ of star $`i`$. A comparison of $`\text{V}_{k()}`$ and $`\text{V}_k`$ is able to tell if the errors introduced by the stellar library are negligible relative to the errors on the observed galaxy.
### B.2 The galaxy cannot be synthesized
An approximate solution $`𝑾_{\text{syn}}`$ is found if the galaxy cannot be synthesized. A manifold of solutions is defined by a number of stars $`n_\text{s}`$ less or equal to the number of equivalent widths. Here we distinguish $`n_\text{s}`$ from the $`n_{}`$ number of stars in the data base. This manifold is called the synthetic ‘surface’ in paper I.
#### B.2.1 Projector on the tangent plane.
If the observational errors $`d𝑾_{\text{obs}}`$ are small around $`𝑾_{\text{obs}}`$, we can expect small deviations $`d𝑾_{\text{syn}}`$ around $`𝑾_{\text{syn}}`$. In this context, we can approximate the synthetic surface by a plane tangent to this surface at $`𝑾_{\text{syn}}`$. The results obtained in B.1 can be applied to $`𝑾_{\text{syn}}`$ since, by definition, $`𝑾_{\text{syn}}`$ belongs to the synthetic domain. In this context let’s $`\dot{\text{A}}`$ and $`\dot{\text{B}}`$ stand for the matrices A and B, where $`𝑾_{\text{obs}}`$ have been replaced by $`𝑾_{\text{syn}}`$. By (21) $`𝑾_{\text{syn}}+d𝑾_{\text{syn}}`$ belongs to the tangent plane if the $`d𝒌`$ are such that:
$$(\begin{array}{c}d𝑾_{\text{syn}}\\ 0\end{array})=\text{J}^1\dot{\text{B}}d𝒌\text{or}d𝑾_{\text{syn}}=\text{J}_A^1\dot{\text{A}}d𝒌,$$
(24)
with $`\text{J}_A=\text{diag}(I_{\text{syn}1},\mathrm{},I_{\text{syn}n_\lambda })`$. This means that $`d𝑾_{\text{syn}}`$ belongs to the linear hull of the columns of $`\text{J}_A^1\text{A}`$ provided that $`_idk_i=0`$. We maintain that $`d𝑾_{\text{syn}}`$ belongs to this linear hull even if $`_idk_i0`$.
In fact $`_idk_i=0`$ is the equation of an hyperplane $`\mathrm{\Sigma }_0`$ orthogonal to the subspace of dimension one generated by e.g. the vector $`(1,\mathrm{},1)`$. The hyperplane $`\mathrm{\Sigma }_0`$ is itself a subspace since it includes the origin, therefore any vector $`d𝒌`$ is the sum of a vector $`d𝒌_0\mathrm{\Sigma }_0`$ plus a vector $`𝒗\mathrm{\Sigma }_0`$. By construction the solution $`𝒌\mathrm{\Sigma }_0`$ since $`_ik_i=1`$, then $`d𝒌`$ can be written $`d𝒌=d𝒌_0+d\alpha 𝒌`$, where $`d\alpha `$ is a scalar. Now we have:
$$\begin{array}{cc}\hfill \text{J}^1\dot{\text{B}}d𝒌=& \text{J}^1\dot{\text{B}}(d𝒌_0+d\alpha 𝒌),\hfill \\ \hfill =& \text{J}^1\dot{\text{B}}d𝒌_0+d\alpha \text{J}^1\dot{\text{B}}𝒌,\hfill \\ \hfill =& (\begin{array}{c}d𝑾_{\text{syn}}\\ 0\end{array})+(\begin{array}{c}0\\ d\alpha \end{array}).\hfill \end{array}$$
Therefore $`d𝑾_{\text{syn}}`$ belongs to the linear hull of the columns of $`\text{G}=\text{J}_A^1\dot{\text{A}}`$ i.e. it belongs to the range of G.
The range of G is obtained by the rank deficiency QR algorithm (see Golub and Van Loan 1996 Chapter 5 §5.4.1). According to this algorithm it is possible to write, after a permutation $`𝚷`$, $`\text{G}𝚷=\text{Q}\text{R}`$. The array Q is a $`n_\lambda `$ by $`n_\lambda `$ matrix. The $`n_\text{s}1`$ first columns of Q span the range of G (i.e. they are an orthonormal basis of the tangent plane) and the remaining columns span the subspace orthonormal to it. If, following this partition, we write $`\text{Q}=(\text{p}|\text{q})`$, one can construct the orthogonal projector H on the tangent plane, we have:
$$\text{H}=\text{p}\text{p}^T.$$
The projector H is a $`n_\lambda `$ by $`n_\lambda `$ matrix of rank $`n_\text{s}1`$.
#### B.2.2 Least square solution of the linearized problem
One obtains $`𝑾_{\text{syn}}^{}=𝑾_{\text{syn}}+d𝑾_{\text{syn}}`$ from $`𝑾_{\text{obs}}^{}=𝑾_{\text{obs}}+d𝑾_{\text{obs}}`$ by minimizing a ‘distance’ between these $`𝑾_{\text{syn}}^{}`$ and $`𝑾_{\text{obs}}^{}`$. This distance is usually defined as a positive definite quadratic form, therefore the correction $`d𝑾_{\text{syn}}`$ is implicitly defined by:
$`(𝑾_{\text{obs}}^{}𝑾_{\text{syn}}^{})^T𝚺^1(𝑾_{\text{obs}}^{}𝑾_{\text{syn}}^{})=`$ (25)
$`\underset{𝑾^{}}{\mathrm{min}}(𝑾_{\text{obs}}^{}𝑾^{})^T𝚺^1(𝑾_{\text{obs}}^{}𝑾^{}).`$
Via a change of variables, one can describe the ‘weight’ matrix $`𝚺^1`$ by the identity matrix (see Sect. B.2.5). We hereafter consider that $`𝚺^1=\text{I}`$, then the synthesis is obtained by the orthogonal projection of $`𝑾_{\text{obs}}`$ on the synthetic surface i.e. using H. Consequently one obtains $`𝑾_{\text{syn}}^{}`$ by adding $`d𝑾_{\text{syn}}=\text{H}(𝑾_{\text{obs}}^{}𝑾_{\text{syn}})`$ to $`𝑾_{\text{syn}}`$. Taking into account that $`\text{H}(𝑾_{\text{obs}}𝑾_{\text{syn}})=0`$ we obtain:
$$d𝑾_{\text{syn}}=\text{H}d𝑾_{\text{obs}}.$$
(26)
#### B.2.3 Variance-covariance matrix of $`𝒌`$
The second step is to find a relation between $`d𝑾_{\text{syn}}`$ and $`d𝒌`$ where $`d𝒌`$ is restricted to the $`n_\text{s}`$ stars which contribute to the synthesis. Since, by definition $`𝑾_{\text{syn}}𝒲`$, the overdetermined system $`\dot{\text{B}}𝒌=(\mathrm{𝟎}\mathrm{\hspace{0.17em}1})^T`$ possesses an exact solution in the least-square sense. Therefore $`𝒌`$ is also the solution of the invertible square system $`\dot{\text{B}}^T\dot{\text{B}}𝒌=\dot{\text{B}}^T(\mathrm{𝟎}\mathrm{\hspace{0.17em}1})^T`$. In this linear analysis, the perturbed $`𝑾_{\text{syn}}^{}`$ is close to the synthetic surface and the normalization constraint $`𝒃^T𝒌=1`$ is approximately satisfied. It follows that $`\dot{\text{B}}^Td\dot{\text{B}}𝒌+\dot{\text{B}}^T\dot{\text{B}}d𝒌=0`$. Again $`d\dot{\text{B}}𝒌`$ is given by Eq. (21), where J is the diagonal matrix of the synthetic continua $`\text{J}=\text{diag}(I_{\text{syn}1},\mathrm{},I_{\text{syn}n_\lambda },1)`$. This leads to:
$$d𝒌=(\dot{\text{B}}^T\dot{\text{B}})^1\dot{\text{B}}^T\text{J}(\begin{array}{c}d𝑾_{\text{syn}}\\ 0\end{array}).$$
(27)
Note that $`\dot{\text{B}}`$ being of full column rank, the expression $`(\dot{\text{B}}^T\dot{\text{B}})^1\dot{\text{B}}^T`$ is equal to $`\dot{\text{B}}^{(1)}`$ the Moore-Penrose pseudo-inverse of $`\dot{\text{B}}`$ (for a definition see e.g. Harville 1997 Chap. 20.1). Let’s define here $`\text{K}=(\dot{\text{B}}^T\dot{\text{B}})^1\dot{\text{B}}^T\text{J}`$ and recalling that $`d𝑾_{\text{syn}}=\text{H}d𝑾_{\text{obs}}`$ we obtain the variance-covariance matrix of the synthesis:
$$\text{V}_k=\text{K}(\begin{array}{cc}\text{H}\text{V}\text{H}^T& \mathrm{𝟎}\\ \mathrm{𝟎}^T& 0\end{array})\text{K}^T,$$
(28)
where $`\text{V}=d𝑾_{\text{obs}}d𝑾_{\text{obs}}^T`$ is the variance-covariance matrix of the observations.
#### B.2.4 Acceptance region of the least-square solutions
An alternative solution $`𝒌^{}`$ is considered acceptable, at the level $`\gamma `$, if there exists a $`𝑾_{\text{obs}}^{}`$ within the ‘error’ region $`_\gamma `$ around $`𝑾_{\text{obs}}`$ such that $`\text{H}𝑾_{\text{obs}}^{}=𝑾_{\text{syn}}^{}=\phi (𝒌^{})`$. According to this definition, $`𝒌^{}`$ is acceptable if it belongs to the following set:
$$\begin{array}{c}𝒦_\gamma =\{𝒌^{}|d𝑾_{\text{obs}}^T\text{V}^1d𝑾_{\text{obs}}F_{\chi ^2}^1(\gamma )\},\hfill \\ \text{H}d𝑾_{\text{obs}}=d𝑾_{\text{syn}}=\phi (𝒌^{})\phi (𝒌),\hfill \end{array}\}$$
(29)
where V is the variance-covariance matrix of the observational errors.
Among all the solutions of $`\text{H}d𝑾_{\text{obs}}=d𝑾_{\text{syn}}`$ we choose the one which is ‘closest’ to $`𝑾_{\text{obs}}`$ in the $`\text{V}^1`$-norm. This ensures that the solution is within $`_\gamma `$. The minimum of $`d𝑾_{\text{obs}}^T\text{V}^1d𝑾_{\text{obs}}`$, which satisfies the constraint $`\text{H}d𝑾_{\text{obs}}=d𝑾_{\text{syn}}`$ is given by:
$$\begin{array}{cc}\hfill d𝑾_{\text{obs}}=& \text{M}d𝑾_{\text{syn}},\hfill \\ \hfill \text{M}=& \text{pp}^T+\text{qq}^T\text{V}\text{p}(\text{p}^T\text{V}\text{p})^1\text{p}^T.\hfill \end{array}$$
The matrix M is an oblique projector on the regression plane defined by the constraint. We indeed have:
$$(\begin{array}{c}d𝑾_{\text{obs}}\\ 0\end{array})=(\begin{array}{c}\text{M}d𝑾_{\text{syn}}\\ 0\end{array}).$$
(30)
One cannot invert Eq. (27) in order to introduce $`d𝒌`$, but we can use Eq. (24) since $`𝒦_\gamma `$ is in the set where $`_idk_i=0`$:
$$(\begin{array}{c}d𝑾_{\text{syn}}\\ 0\end{array})=\text{J}^1\dot{\text{B}}d𝒌.$$
(31)
Combining Eqs. (31) and (30) and proceeding like in Sect. § (B.1.2), we find:
$$𝒦_\gamma =\{𝒌^{}|(𝒌^{}𝒌)^T\text{P}(𝒌^{}𝒌)F_{\chi ^2}^1(\gamma )\},$$
(32)
$$\text{P}=\text{W}^T(\begin{array}{cc}\text{M}^T\text{V}^1\text{M}& \mathrm{𝟎}\\ \mathrm{𝟎}^T& 0\end{array})\text{W},\text{W}=\text{J}^1\dot{\text{B}}.$$
(33)
#### B.2.5 Weighted least square
We consider now the case where the distance between two points of the $`W`$-space is not the Euclidean distance but an elliptical distance. This elliptical distance is subordinated to a symmetric definite positive matrix $`𝚺`$. We have $`d_\mathrm{\Sigma }(𝑾_1,𝑾_2)=(𝑾_1𝑾_2)^T𝚺^1(𝑾_1𝑾_2)`$. Usually one chooses $`𝚺`$ equal to V the variance-covariance matrix of the data, but it may not be always the case.
A substitution of variables transforms the elliptical distance into a spherical (i.e. Euclidean) one. We want in other words that $`d(𝑾_1,𝑾_2)=d_\mathrm{\Sigma }(𝑾_{10},𝑾_{20}),`$where $`d`$ is the Euclidean distance. If $`𝑾_0`$ stands for the original variables and $`𝑾`$ for the new variables, we define the matrix N by: $`𝑾_0=\text{N}𝑾`$. It now follows that
$$d(𝑾_1,𝑾_2)=(𝑾_1𝑾_2)^T\text{N}^T𝚺^1\text{N}(𝑾_1𝑾_2).$$
This imposes the condition $`\text{N}^T𝚺^1\text{N}=\text{I}`$, which is satisfied if we choose N as the matrix appearing in the Choleski factorization of $`𝚺`$ (see Golub & Van Loan 1996, section 4.2.1). We have $`𝚺=\text{N}\text{N}^T`$, where N is lower triangular. It is straightforward to verify that N satisfies the condition.
Finally we need to compute the variance-covariance matrix V appearing in Eqs. (28) and (33) from the variance-covariance matrix $`\text{V}_0`$ of the original variables. We have $`\text{V}=d𝑾d𝑾^T=\text{N}^1d𝑾_0𝑾_0^T\text{N}^{1T}`$, therefore:
$$\text{V}=\text{N}^1\text{V}_0\text{N}^{1T}.$$
(34) |
warning/0001/hep-ph0001077.html | ar5iv | text | # Neutrino scattering in strong magnetic fieldsPlenary talk given by Palash B. Pal at the COSMO-99 conference held at the Abdus Salam ICTP, Trieste, Italy, during 29/9/1999 to 2/10/1999
## 1 Motivation
Neutrino interactions in strong magnetic fields have gained a lot of attention because of the problem of neutrino emission from a proto-neutron star. It has been argued that a small asymmetry in the neutrino emission can explain the high peculiar velocities of pulsars. Based on earlier work on neutrino dispersion relations in an external magnetic field, it has been shown that neutrino oscillations can probably account for the pulsar kicks. On the other hand, there is a parallel stream of argument with the idea that neutrino opacities may change appreciably in magnetic fields. Asymmetric opacities can presumably account for the kicks. The opacity has recently been calculated by Roulet, assuming the matrix element for the interactions remain unchanged in a magnetic field. He found that the opacities do change appreciably compared to the no-field case. However, he did not find any asymmetry vis-a-vis the direction of the magnetic field. The aim of the present work is to redo these calculations from first principles and show that the relevant cross-section is indeed asymmetric.
## 2 Solutions of the Dirac equation in a uniform magnetic field
We take the magnetic field $`B`$ in the $`z`$-direction, and choose
$`A_0=A_y=A_z=0,A_x=By.`$ (1)
The Dirac equation in this field can be exactly solved. The energy eigenvalues are given by
$`E_n^2=m^2+p_z^2+2neB,`$ (2)
where $`n`$ is a non-negative integer signifying the Landau level. All $`n>0`$ levels are doubly degenerate, whereas the $`n=0`$ level is non-degenerate. The eigenstates are of the form
$`e^{ipX_{\backslash y}}U_s(y,n,𝒑_{\backslash y}),`$ (3)
where the notation $`pX_{\backslash y}`$ stands for the dot product, setting the $`y`$-components equal to zero. $`U_s`$ is the spinor, given by
$`U_+=\left(\begin{array}{c}I_n(\xi )\\ 0\\ \frac{p_z}{E_n+m}I_n(\xi )\\ \frac{M_n}{E_n+m}I_{n1}(\xi )\end{array}\right),U_{}=\left(\begin{array}{c}0\\ I_{n1}(\xi )\\ \frac{M_n}{E_n+m}I_n(\xi )\\ \frac{p_z}{E_n+m}I_{n1}(\xi )\end{array}\right).`$ (12)
We have used the shorthand $`M_n=\sqrt{2neB}`$, and a dimensionless variable
$`\xi =\sqrt{eB}\left(y+{\displaystyle \frac{p_x}{eB}}\right).`$ (13)
The function $`I_n(\xi )`$ appearing in Eq. (12) is given by
$`I_n(\xi )=N_ne^{\xi ^2/2}H_n(\xi ),`$ (14)
where $`H_n`$ are Hermite polynomials, and $`N_n`$ is a normalization which can be chosen arbitrarily. We choose
$`N_n=\left({\displaystyle \frac{\sqrt{eB}}{n!\mathrm{\hspace{0.17em}2}^n\sqrt{\pi }}}\right)^{1/2}.`$ (15)
For the sake of consistency, we will define $`I_1=0`$ so that the solution $`U_{}`$ vanishes for the zeroth Landau level. From the spinor solutions, the spin sum can be calculated:
$`P_U(y,y_{},n,𝒑_{\backslash y}){\displaystyle \underset{s}{}}U_s(y,n,𝒑_{\backslash y})\overline{U}_s(y_{},n,𝒑_{\backslash y})`$ (16)
$`=`$ $`{\displaystyle \frac{1}{2(E_n+m)}}[\{m(1+\sigma _z)+\text{/}p_{}\stackrel{~}{\text{/}p}_{}\gamma _5\}I_n(\xi )I_n(\xi _{})`$
$`+`$ $`\left\{m(1\sigma _z)+\text{/}p_{}+\stackrel{~}{\text{/}p}_{}\gamma _5\right\}I_{n1}(\xi )I_{n1}(\xi _{})`$
$`+`$ $`M_n\{(\gamma _1+i\gamma _2)I_n(\xi )I_{n1}(\xi _{})+(\gamma _1i\gamma _2)I_{n1}(\xi )I_n(\xi _{})\}],`$
using the notations $`\text{/}p_{}=\gamma _\mu p_{}^\mu `$, $`\stackrel{~}{\text{/}p}_{}=\gamma _\mu \stackrel{~}{p}_{}^\mu `$, where
$`p_{}^\mu =(p_0,0,0,p_z),\stackrel{~}{p}_{}^\mu =(p_z,0,0,p_0).`$ (17)
## 3 The fermion field operator
The fermion field operator can be written as
$`\psi (X)`$ $`=`$ $`{\displaystyle \underset{s=\pm }{}}{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{dp_xdp_z}{2\pi }\sqrt{\frac{E_n+m}{2E_n}}}`$ (18)
$`\times `$ $`\left[f_s(n,𝒑_{\backslash y})e^{ipX_{\backslash y}}U_s(y,n,𝒑_{\backslash y})+\widehat{f}_s^{}(n,𝒑_{\backslash y})e^{ipX_{\backslash y}}V_s(y,n,𝒑_{\backslash y})\right].`$
where $`f_s(n,𝒑_{\backslash y})`$is annihilation operator for the fermion, and $`\widehat{f}_s^{}(n,𝒑_{\backslash y})`$ is the creation operator for the antifermion in the $`n`$-th Landau level with given values of $`p_x`$ and $`p_z`$. The creation and annihilation operator satisfy the anticommutation relations
$`[f_s(n,𝒑_{\backslash y}),f_s^{}^{}(n^{},𝒑_{\backslash y}^{})]_+=\delta _{ss^{}}\delta _{nn^{}}\delta (p_xp_x^{})\delta (p_zp_z^{}),`$ (19)
etc. The prefactor appearing in Eq. (18) is determined from the fact that the field operator should satisfy the anticommutation rule
$`[\psi (X),\psi ^{}(X_{})]_+=\delta ^3(𝑿𝑿_{})`$ (20)
for $`X^0=X_{}^0`$. The one-fermion states are defined by
$`|n,𝒑_{\backslash y}={\displaystyle \frac{2\pi }{\sqrt{L_xL_z}}}f^{}(n,𝒑_{\backslash y})|0,`$ (21)
the normalization constant chosen so that these states are orthonormal in a box with sides $`L_x`$, $`L_y`$ and $`L_z`$. Then
$`<n,𝒑_{\backslash y}|\overline{\psi }_U(X)`$ $`=`$ $`\sqrt{{\displaystyle \frac{E_n+m}{2E_nL_xL_z}}}e^{ipX_{\backslash y}}\overline{U}_s(y,n,𝒑_{\backslash y})<0|,`$ (22)
where $`\psi _U`$ denotes the term in Eq. (18) that contains the $`U`$-spinors.
## 4 Inverse beta-decay
We now apply the above formalism for the inverse beta-decay process
$`\nu _e(𝒌)+n(𝑷)p(𝑷^{})+e(𝒑_{\backslash y}^{},n^{}).`$ (23)
Assume $`eBm_p^2`$, so that the magnetic field effects on the proton and neutron spinors can be ignored.
The interaction Lagrangian is
$`_{\mathrm{int}}=\sqrt{2}G_\beta \left[\overline{\psi }_{(e)}\gamma ^\mu L\psi _{(\nu _e)}\right]\left[\overline{\psi }_{(p)}\gamma _\mu (g_V+g_A\gamma _5)\psi _{(n)}\right],`$ (24)
where $`G_\beta =G_F\mathrm{cos}\theta _c`$, $`g_V=1`$ and $`g_A=1.26`$. For the $`S`$-matrix element, this gives
$`S_{fi}`$ $`=`$ $`\sqrt{2}G_\beta {\displaystyle d^4Xe(𝒑_{\backslash y}^{},n^{})\left|\overline{\psi }_{(e)}\gamma ^\mu L\psi _{(\nu _e)}\right|\nu _e(𝒌)}`$ (25)
$`\times p(P^{})\left|\overline{\psi }_{(p)}\gamma _\mu (g_V+g_A\gamma _5)\psi _{(n)}\right|n(P).`$
The hadronic part of the matrix element is obtained with usual rules, which gives
$`{\displaystyle \frac{e^{i(P^{}P)X}}{\sqrt{2P_0V}\sqrt{2P_0^{}V}}}\left[\overline{u}_{(p)}(𝑷^{})\gamma _\mu (g_V+g_A\gamma _5)u_{(n)}(𝑷)\right],`$ (26)
For the leptonic part, we need magnetic spinors for the electron. Using Eq. (22), we obtain this matrix element to be
$`{\displaystyle \frac{e^{ikX+ip^{}X_{\backslash y}}}{\sqrt{2\omega V}}}\sqrt{{\displaystyle \frac{E_n^{}+m}{2E_n^{}L_xL_z}}}\left[\overline{U}_{(e)}(y,n^{},𝒑_{\backslash y}^{})\gamma ^\mu Lu_{(\nu _e)}(𝒌)\right].`$ (27)
Putting these back into Eq. (25) and performing the integrations over all co-ordinates except $`y`$, we obtain
$`S_{fi}=(2\pi )^3\delta _{\backslash y}^3(P+kP^{}p^{})\left[{\displaystyle \frac{E_n^{}+m}{2\omega V\mathrm{\hspace{0.33em}2}P_0V\mathrm{\hspace{0.33em}2}P_0^{}V2E_n^{}L_xL_z}}\right]^{1/2}_{fi},`$ (28)
where
$`_{fi}`$ $`=`$ $`\sqrt{2G_\beta }\left[\overline{u}_{(p)}(𝑷^{})\gamma ^\mu (g_V+g_A\gamma _5)u_{(n)}(𝑷)\right]`$ (29)
$`\times `$ $`{\displaystyle 𝑑ye^{iq_yy}\left[\overline{U}_{(e)}(y,n^{},𝒑_{\backslash y}^{})\gamma _\mu Lu_{(\nu _e)}(𝒌)\right]},`$
using the shorthand
$`q_y=P_y+k_yP_y^{}.`$ (30)
Note that Eq. (28) does not have a $`\delta `$-function corresponding to the conservation of $`y`$-component of momentum, because the latter is not a good quantum number in this problem. The transition rate can now be written as
$`|S_{fi}|^2/T`$ $`=`$ $`{\displaystyle \frac{1}{16}}(2\pi )^3\delta _{\backslash y}^3(P+kP^{}p^{}){\displaystyle \frac{E_n^{}+m}{V^3\omega P_0P_0^{}E_n^{}}}\left|_{fi}\right|^2.`$ (31)
Putting in now the differential phase space for final particles, which is
$`{\displaystyle \frac{L_x}{2\pi }}dp_x^{}{\displaystyle \frac{L_z}{2\pi }}dp_z^{}{\displaystyle \frac{V}{(2\pi )^3}}d^3P^{}`$ (32)
where $`V=L_xL_yL_z`$, we can write the differential cross section as
$`d\sigma ={\displaystyle \frac{1}{64\pi ^2}}\delta _{\backslash y}^3(P+kP^{}p^{}){\displaystyle \frac{E_n^{}+m}{\omega P_0P_0^{}E_n^{}}}\left|_{fi}\right|^2{\displaystyle \frac{L_xL_z}{V}}dp_x^{}dp_z^{}d^3P^{}.`$ (33)
The square of the matrix element, averaged over initial neutron spin, is given by
$`\left|_{fi}\right|^2=G_\beta ^2\mathrm{}_{\mu \nu }H^{\mu \nu }.`$ (34)
Here, $`H^{\mu \nu }`$ is the hadronic part, which is
$`H^{\mu \nu }`$ $`=`$ $`4(g_V^2+g_A^2)(P^\mu P^\nu +P^\nu P^\mu g^{\mu \nu }PP^{})`$ (35)
$`+`$ $`4(g_V^2g_A^2)m_nm_pg^{\mu \nu }+8g_Vg_Ai\epsilon ^{\mu \nu \lambda \rho }P_\lambda ^{}P_\rho .`$
The leptonic part $`\mathrm{}_{\mu \nu }`$ contains magnetic spinors:
$`\mathrm{}_{\mu \nu }={\displaystyle 𝑑y𝑑y_{}e^{iq_y(yy_{})}\mathrm{Tr}\left[P_U(y_{},y,n^{},𝒑_{\backslash y}^{})\gamma _\mu \text{/}k\gamma _\nu L\right]}.`$ (36)
The integrations over $`y`$ and $`y_{}`$ can be exactly performed, yielding
$`\mathrm{}_{\mu \nu }={\displaystyle \frac{2\pi }{eB}}{\displaystyle \frac{1}{(E_n^{}+m)}}(\mathrm{\Lambda }_\mu k_\nu +\mathrm{\Lambda }_\nu k_\mu k\mathrm{\Lambda }g_{\mu \nu }i\epsilon _{\mu \nu \alpha \beta }\mathrm{\Lambda }^\alpha k^\beta ),`$ (37)
where
$`\mathrm{\Lambda }^\alpha `$ $`=`$ $`\left[I_n\left({\displaystyle \frac{q_y}{\sqrt{eB}}}\right)\right]^2(p_{}^\alpha \stackrel{~}{p}_{}^\alpha )+\left[I_{n1}\left({\displaystyle \frac{q_y}{\sqrt{eB}}}\right)\right]^2(p_{}^\alpha +\stackrel{~}{p}_{}^\alpha )`$ (38)
$`+2M_ng_2^\alpha I_n\left({\displaystyle \frac{q_y}{\sqrt{eB}}}\right)I_{n1}\left({\displaystyle \frac{q_y}{\sqrt{eB}}}\right).`$
We calculate the cross section in the rest frame of the neutron. The axes are chosen such that incoming neutrino is in the $`x`$-$`z`$ plane. Also, we assume that in the energy regime of interest, $`|𝑷^{}|m_p`$ so that the proton is non-relativistic. Then we obtain, for $`n^{}0`$,
$`\left|_{fi}\right|^2={\displaystyle \frac{16\pi G_\beta ^2}{eB}}{\displaystyle \frac{m_nm_p}{E_n^{}+m}}\left[(g_V^2+3g_A^2)\omega \mathrm{\Lambda }_0+(g_V^2g_A^2)k_z\mathrm{\Lambda }_z\right].`$ (39)
We now put it into Eq. (33). Integrations over $`P_x^{}`$ and $`P_z^{}`$ get rid of the $`\delta `$-functions. Integration over $`p_x^{}`$ gives a factor $`L_yeB`$.
$`d\sigma ={\displaystyle \frac{G_\beta ^2}{4\pi }}{\displaystyle \frac{\delta (Q+\omega E_n^{})}{\omega E_n^{}}}\left[(g_V^2+3g_A^2)\omega \mathrm{\Lambda }_0+(g_V^2g_A^2)k_z\mathrm{\Lambda }_z\right]dP_y^{}dp_z^{},`$ (40)
where $`Q=m_nm_p`$. Integration over $`P_y^{}`$ can now be performed using the orthogonality of the Hermite polynomials. This ensures that the $`M_n`$-term of Eq. (38) vanishes. The other two give
$`d\sigma _{n^{}>0}={\displaystyle \frac{eBG_\beta ^2}{2\pi }}\delta (Q+\omega E_n^{})\left[(g_V^2+3g_A^2)+(g_V^2g_A^2){\displaystyle \frac{k_zp_z^{}}{\omega E_n^{}}}\right]dp_z^{}.`$ (41)
The argument of the remaining $`\delta `$-function vanishes for two different values of $`p_z^{}`$ which are equal and opposite. The term proportional to $`g_V^2g_A^2`$ then cancels out and we obtain
$`\sigma _{n^{}>0}`$ $`=`$ $`{\displaystyle \frac{eBG_\beta ^2}{\pi }}(g_V^2+3g_A^2){\displaystyle \frac{Q+\omega }{\sqrt{(Q+\omega )^2m^2M_n^{}^2}}}.`$ (42)
For $`n^{}=0`$, however, even the second term of Eq. (38) vanishes, so we need to redo everything beginning from Eq. (39). The final result is:
$`\sigma _0`$ $`=`$ $`{\displaystyle \frac{eBG_\beta ^2}{2\pi }}\left[(g_V^2+3g_A^2)(g_V^2g_A^2){\displaystyle \frac{k_z}{\omega }}\right]{\displaystyle \frac{Q+\omega }{\sqrt{(Q+\omega )^2m^2}}}.`$ (43)
For any given energy of the initial states, there is a maximum possible value for $`n^{}`$, dictated by Eq. (2). The total cross section is obtained by performing the sum over all possible $`n^{}`$. We can then summarize our results as follows:
* $`\sigma _0`$ depends on the direction of neutrino momentum because of the term involving $`k_z`$.
* Total $`\sigma `$ always contains $`\sigma _0`$, and is hence asymmetric.
* The fractional amount of asymmetry depends on $`n_{\mathrm{max}}^{}`$. It is maximum if $`n_{\mathrm{max}}^{}=0`$, and decreases at smaller values of $`B`$ as more Landau levels contribute.
In Fig. 1, we show the cross section as a function of $`B`$ for neutrinos incident along and opposite to the direction of the magnetic field. For small values of $`B`$, the distinction between the two directions is not seen at the scale of the plot. However, when $`B`$ is so large that $`n_{\mathrm{max}}^{}=0`$, the asymmetry is large. It is, in fact, as large as 18% between these two extreme directions. It seems highly plausible that such a large asymmetry in the cross section would produce enough momentum asymmetry to explain the pulsar kicks. These calculations were not finished at the time of the conference and therefore could not be presented.
There is however another important issue to be settled before the momentum asymmetry is calculated. Neutrinos undergo many other reactions inside a dense star. The inverse beta decay has the largest cross section for the $`B=0`$ case of course, but there is no guarantee that other cross sections are not larger in strong magnetic fields. Our preliminary calculations for the neutrino-electron elastic scattering show that it can increase by several orders of magnitude in a magnetic field. If that is the case, then the opacities have to be calculated taking these processes into account. |
warning/0001/cond-mat0001163.html | ar5iv | text | # Untitled Document
Analytical Estimate of the Critical Velocity for
Vortex Pair Creation in Trapped Bose Condensates
by
M. Crescimanno, C. G. Koay, and R. Peterson
Physics Department
Berea College
Berea, Ky. 40404
and
R. Walsworth
Harvard-Smithsonian Center for Astrophysics
60 Garden Street
Cambridge, Ma. 02138
December 1999
ABSTRACT : We use a modified Thomas-Fermi approximation to estimate analytically the critical velocity for the formation of vortices in harmonically trapped BEC. We compare this analytical estimate to numerical calculations and to recent experiments on trapped alkali condensates.
I. Introduction : The experimental realization of Bose-Einstein condensation (BEC) in cold trapped alkali atoms has stimulated great interest in weakly interacting inhomogeneous quantum gases. The alkali condensates are excellent systems for the study of quantum fluids, because (i) they are dilute, and hence admit an accurate description by mean field theory; (ii) system parameters such as density, temperature, and trapping potential are under fine experimental control; and (iii) sensitive diagnostics (mostly optical) have been developed to probe condensate behavior. In particular, alkali BEC may serve as a powerful laboratory to study superfluidity – both long-established questions about its breakdown, and newer questions about the effects of finite size and spatial inhomogeneity.
It is well-known theoretically that collective excitations in a weakly interacting BEC determine a superfluid critical velocity, below which the condensate flow relative to a perturbing object or potential is without drag. This critical velocity is given by the Landau criterion :
$$v_c=\mathrm{min}\left(\frac{E(I)}{I}\right)$$
$`(1.1)`$
where $`E(I)`$ is the energy of a condensate collective excitation of linear momentum (or impulse) $`I`$. For liquid <sup>4</sup>He, the breakdown of superfluidity is observed experimentally to occur at a critical velocity that is much smaller than that due to the excitation of phonons or rotons . Beginning with Feynman more than 40 years ago , it has long been thought that the critical velocity for superfluid <sup>4</sup>He is set by the creation of complex vortex structures (e.g., pairs, rings, and loops) depending on system geometry, temperature, etc. However, it has not been possible to fully verify this basic hypothesis because of the strong interactions in a liquid, which complicate quantitative comparison of superfluid <sup>4</sup>He experiments with theory , as well as the limited ability to vary the liquid density. Alkali gas BEC may provide an experimental system to test quantitatively the link between the superfluid critical velocity and the creation of vortices.
In a landmark recent experiment , Ketterle and co-workers observed a critical velocity for the excitation of an alkali BEC when a perturbing potential (a blue-detuned laser beam) was moved through the trapped quantum fluid. These measurements are in qualitative agreement with both the well-known analytical calculation for a homogeneous weakly-interacting BEC in a channel of finite diameter (providing a $`v_c`$ that is a factor of $``$ 2 lower than experiment), and recent numerical calculations \[8-11\] based on the Gross-Pitaevskii equation (providing a $`v_c`$ that is a factor of $``$ 2 higher than experiment).
Detailed understanding of the breakdown of superfluidity in Bose condensates will likely require further experimentation (e.g., to vary the trapped alkali BEC geometry and density, and to create greater spatial symmetry in the BEC perturbation), as well as additional theoretical work – both numerical and analytical. To this end, we report in this paper an analytical calculation of the critical velocity for vortex pair production in a harmonically-trapped dilute Bose condensate. We employ a modification of the usual Thomas-Fermi (TF) approximation to the Bogoliubov mean-field theory , treating the BEC like a fluid with an exotic equation of state. We include leading kinetic energy terms caused by the two counter-rotating vortices, and use these kinetic energy terms as an effective potential for the BEC. We then compute, approximately, the energy $`E(I)`$ and impulse $`I`$ of the vortex pair, and determine the critical velocity from the Landau criterion.
As a step toward checking this modified TF limit for BEC vortices, we also compute the Bogoliubov many-body “wavefunction” ($`\psi `$) for a condensate with a single central vortex, and then calculate $`E`$, $`I`$, and $`v_c`$ for this case. Since our calculations are based on an effective hydrodynamic model of BEC, the vortices that we consider do not, strictly speaking, have quantized angular momentum in units of $`\mathrm{}`$. However, we find that the modified TF prediction agrees quite well with a vortex solution computed numerically.
We believe the advantages of the analytical method presented here are (i) its simplicity; and (ii) that it works well for the limit of large inter-particle interaction, which is the limit largely being explored by current experiments. In addition, the modified Thomas-Fermi analysis suggests the manner in which $`v_c`$, as a fraction of the speed of sound ($`c_B`$), decreases as the inter-particle interaction increases.
In section II below we introduce both the model used in our calculation, and the modified Thomas-Fermi approximation. In section III we apply this model and approximation to the creation of vortex pairs near the center of a harmonically-trapped BEC, and then compare the analytically estimated critical velocity with the recent measurements by the MIT group . In section IV we show that a modified TF analytical calculation of $`v_c`$ for a single central vortex in a trapped BEC agrees well with numerical simulation. Finally, in section V we summarize the implications of this approach.
II. Model of Trapped BEC and the Modified Thomas-Fermi Limit : We restrict our discussion to the dynamics of a harmonically-trapped single-component Bose condensate at $`T=0`$, which is well described by the Gross-Pitaevskii (GP) equation :
$$\frac{\mathrm{}^2}{2M}^2\psi ^{}+\frac{KR^2}{2}\psi ^{}+U|\psi ^{}|^2\psi ^{}=\mu ^{}\psi ^{}$$
$`(2.1)`$
where the field $`\psi ^{}`$, the diagonal part of the multiparticle wave function, serves as the condensate order parameter, and is referred to as “the wavefunction of the condensate”; $`M`$ is the atomic mass; $`\mu ^{}`$ is the BEC chemical potential; and $`U|\psi ^{}|^2`$ is an effective nonlinear potential arising from the atomic interactions (due to $`s`$-wave binary atomic collisions). Here $`U=\frac{4\pi \mathrm{}^2a_s}{M}`$, where $`a_s`$ is the atomic $`s`$-wave scattering length. For simplicity, we assume the trap has cylindrical symmetry and take $`R^2=x^2+y^2=|\stackrel{}{r}|^2`$, with force constant $`K`$ and no trapping potential along the symmetry ($`\widehat{z}`$) axis. Note that we solve the above GP equation subject to the constraint that the total number of atoms $`N`$ is fixed,
$$N=\mathrm{d}^2rdz|\psi ^{}|^2.$$
$`(2.2)`$
For ease of calculation we use scaled harmonic oscillator units (h.o.u.) in which the units of length, time, and energy are $`\sqrt{\frac{\mathrm{}}{2M\omega }}`$ , $`\frac{1}{\omega }`$, and $`\mathrm{}\omega `$, respectively, where $`\omega =\sqrt{\frac{K}{M}}`$ is the trap angular frequency. Also, we reduce the GP equation to two dimensions by assuming the wavefunction is separable: $`\psi ^{}=\psi (\stackrel{}{r})\mathrm{\Phi }(z)`$. Thus, adopting the notation of Jackson et al. , we find
$$^2\psi +\frac{R^2}{4}\psi +C|\psi |^2\psi =\mu \psi $$
$`(2.3)`$
where we have scaled $`\psi `$ so that
$$1=\mathrm{d}^2r|\psi |^2.$$
$`(2.4)`$
Solving Eq. (2.3) subject to Eq. (2.4) fixes $`\mu `$, the dimensionless chemical potential. The dimensionless interaction parameter is
$$C=\frac{2MUN}{\mathrm{}^2a_t\zeta }$$
$`(2.5)`$
where
$$a_t=\sqrt{\frac{\mathrm{}}{2M\omega }}$$
$`(2.6)`$
is the classical turning-point width of the condensate in the harmonic trap in the $`C0`$ limit (equal to the unit of length in h.o.u.), and $`\zeta `$ is the trap aspect ratio given approximately by $`\zeta =\frac{dzz\mathrm{\Phi }^{}\mathrm{\Phi }}{a_t|\mathrm{\Phi }(0)|^2}`$.
For large $`C`$ – which corresponds to most current experiments – the solution to the GP equation is well approximated by the Thomas-Fermi (TF) limit . In this limit one neglects the derivative term in Eq. (2.3), yielding
$$\psi (R)=\sqrt{\frac{\mu R^2/4}{C}}and\mu =\sqrt{\frac{C}{2\pi }},$$
$`(2.7)`$
where the second relation comes from the normalization condition, Eq. (2.4). (See also Eq. (3.4) and discussion below.) Eqs. (2.7) will be used extensively in this paper as the no-vortex solution that we compare to the vortex solutions. For reference, note that in the TF limit (i.e., the long wavelength limit), the local speed of (Bogoliubov) sound in the condensate is given in h.o.u. by $`c_B=\sqrt{2C|\psi |^2}`$ (see for example Ref. ).
The many-body wavefunction of a vortex in the condensate will have a spatially-dependent phase \[see Eq. (3.1) below\], and a spatially-dependent amplitude which we refer to as the ’envelope function’. Let $`\varphi (\stackrel{}{r})`$ and $`A(\stackrel{}{r})`$ be the phase and envelope functions (both real) parameterizing this wavefunction via $`\psi =Ae^{i\varphi }`$. Using this parameterization in Eq. (2.3), and equating real and imaginary parts, we find:
$$^2A+A(\stackrel{}{}\varphi )^2+\frac{R^2}{4}A+CA^3=\mu A$$
$`(2.8)`$
$$^2\varphi 2\stackrel{}{}A\stackrel{}{}\varphi =0.$$
$`(2.9)`$
One may think of Eq. (2.8) as an equation of hydrostatic equilibrium and Eq. (2.9) as a continuity equation. What we call the modified TF approximation consists of first solving the continuity equation \[Eq. (2.9)\] for $`\varphi (\stackrel{}{r})`$, and then using that solution in Eq. (2.8) to determine $`A(\stackrel{}{r})`$, neglecting the $`^2A`$ term. Thus, in this approximation the envelope function $`A(\stackrel{}{r})`$ is the solution of a simple algebraic equation.
Once we have normalized the vortex wavefunction via Eq. (2.4), we compute the energy and integrated impulse. The energy functional is
$$E=\mathrm{d}^2r\left[\stackrel{}{}\psi ^{}\stackrel{}{}\psi +\frac{R^2}{4}\psi ^{}\psi +\frac{C}{2}(\psi ^{}\psi )^2\right].$$
$`(2.10)`$
Using $`\psi =Ae^{i\varphi }`$ with $`A`$ and $`\varphi `$ real, one finds,
$$E=\mathrm{d}^2r\left[(\stackrel{}{}A)^2+(A\stackrel{}{}\varphi )^2+\frac{R^2A^2}{4}+\frac{C}{2}A^4\right].$$
$`(2.11)`$
The local momentum density is $`\stackrel{}{P}=\frac{i\mathrm{}}{2}\left((\stackrel{}{}\psi ^{})\psi \psi ^{}\stackrel{}{}\psi \right)`$, which is used to compute the total impulse:
$$I=\mathrm{d}^2r|\stackrel{}{P}|$$
$$=\mathrm{}\mathrm{d}^2rA^2|\stackrel{}{}\varphi |.$$
$`(2.12)`$
We can then estimate the critical velocity for vortex pair creation by applying the Landau criterion \[Eq. (1.1)\],
$$v_c=\frac{E_{vortex}E_{novortex}}{I_{vortex}I_{novortex}}.$$
$`(2.13)`$
In the next section we describe the details of an analytical computation of this ratio for a vortex pair near the center of the trap.
III. Analytical Calculation of the Vortex Pair Critical Velocity : We consider a dilute BEC containing a pair of counter-rotating vortices with opposite charge or vorticity $`\pm n`$. The vortices are assumed to be located symmetrically about the trap center, with the cores a distance $`d`$ apart. We apply the modified Thomas-Fermi (TF) method to the hydrostatic equilibrium expression \[Eq. (2.8)\] by assuming that $`^2A`$ is negligible. In addition, we adopt the ansatz that the two vortices are far enough from each other, but near enough to the trap center, that the total phase advance about the vortex pair is the simple sum of the phase advances of the two vortices viewed individually. This ansatz is akin to the analytical approximation developed by Feynman for homogeneous BEC, and also used by Fetter in a similar context . Practically, the ansatz is equivalent to requiring $`\frac{1}{\mu }<d^2<\mu `$ in dimensionless harmonic oscillator units (h.o.u.). Operationally it means that the wavefunction’s phase $`\varphi (\stackrel{}{r})`$ satisfies $`^2\varphi =0`$ everywhere outside the vortex cores, which implies from the continuity expression \[Eq. (2.9)\] that $`\stackrel{}{}A\stackrel{}{}\varphi `$ must be zero in this region. Our solution for $`A(\stackrel{}{r})`$ in this case satisfies the continuity equation near the vortex cores, and also far away from the vortices, since radial gradients of the phase vanish as $`\frac{1}{R^2}`$. In addition, the spatial integral of the continuity equation vanishes identically, and deviations due to the approximations described above are asymptotically bounded.
We have found that neglect of the $`^2A`$ term in Eq. (2.8) leads to the main systematic errors in the analytical calculation described here. However, the deviations are small for parameters typical of current experiments.
With the approach outlined above, the phase of the vortex pair wavefunction is
$$\varphi (\stackrel{}{r})=n\mathrm{atan}\left[\frac{\mathrm{sin}\theta \frac{d}{2R}}{\mathrm{cos}\theta }\right]n\mathrm{atan}\left[\frac{\mathrm{sin}\theta +\frac{d}{2R}}{\mathrm{cos}\theta }\right]$$
$`(3.1)`$
where $`R`$ and $`\theta `$ are the usual polar coordinates as measured from the trap center. Employing the modified TF method, we find the condensate wavefunction envelope to be
$$A^2(\stackrel{}{r})=\frac{1}{C}\left(\mu \frac{R^2}{4}\frac{d^2n^2}{R^2d^2\mathrm{cos}^2\theta +(R^2d^2/4)^2}\right)$$
$`(3.2)`$
where the last term is the leading kinetic energy contribution, due to the $`|\stackrel{}{}\varphi |^2`$ term in Eq. (2.8). Note that this kinetic energy contribution is simply $`\frac{1}{(r_1r_2)^2}`$, where $`r_1`$ and $`r_2`$ are the respective distances measured from the vortex cores to the point $`(R,\theta )`$.
The above calculation scheme for $`A(\stackrel{}{r})`$ breaks down very close to the vortex cores (small $`r_1`$ or $`r_2`$) and also at large $`R`$, since the right hand side of Eq. (2.2) becomes negative. Considering these regions to be excluded complicates the analytic evaluation of energy and impulse \[Eqs. (2.11) and (2.12)\] precisely where the TF approximation fails. To circumvent this calculational difficulty we adopt a regulated expression for $`A(\stackrel{}{r})`$:
$$A^2(\stackrel{}{r})=\frac{1}{C}\left(\mu \frac{R^2}{4}\frac{d^2n^2}{R^2d^2\mathrm{cos}^2\theta +(R^2d^2/4)^2+ϵ}\right)$$
$`(3.3)`$
where $`ϵ=\frac{d^2n^2}{\mu }`$. This regulated expression is justified by the observation that for vortex pairs not too far from the trap center ($`d^2<\mu `$), the contribution to $`A(\stackrel{}{r})`$ from the kinetic energy term $`|\stackrel{}{}\varphi |^2`$ is never larger than $`\mu `$. The regulation markedly affects the wavefunction envelope near the vortex cores, but has a small effect on the calculated energy, impulse, and critical velocities. (This is shown for a trap-centered, single vortex in section IV below.) In sum, the use of the regulated expression for $`A(\stackrel{}{r})`$ is an important practical step in our analytical calculation, because it allows us to perform the spatial integrals for $`E_{vortex}`$ and $`I_{vortex}`$ over the entire plane, rather than excluding regions near the vortex cores.
We begin the integrations by normalizing the condensate wavefunction for the cases of no vortices and a single vortex pair. Referring to Eq. (2.7), the normalization for the no vortex case yields
$$\mu \mu _0=\sqrt{\frac{C}{2\pi }},$$
$`(3.4)`$
where we have performed the integration in Eq. (2.4) out to a maximum radius, $`R_{TF}=2\sqrt{\mu }`$, i.e., the Thomas-Fermi condensate edge (the radius at which the condensate chemical potential is dominated by the trap potential in the Thomas-Fermi limit). Similarly, inserting Eq. (3.3) in Eq. (2.4), and noting that $`|\psi |^2=A^2`$, we determine the normalization condition for the vortex pair wavefunction:
$$\frac{C}{2\pi }=\mu ^2n^2\mathrm{ln}\left(\frac{1+\frac{\mu d^2}{n^2}}{1+\frac{d^2}{8\mu }}\right)$$
$`(3.5)`$
where we have again performed the spatial integration out to the Thomas-Fermi condensate edge, which for large $`C`$ is $`R_{TF}=2\sqrt{\mu }\frac{n^2d^2}{16\mu ^{5/2}}+\mathrm{}`$. Note that Eq. (3.5) indicates that $`\mu `$ increases with the vortex charge, $`n`$, as expected. (See Ref. for details of the calculation of these normalization conditions.)
Next, we compute the condensate energy. For the case of no vortex, we insert Eq. (2.7) in Eq. (2.10), and find
$$E_0=\frac{4\pi \mu _0^3}{3C}.$$
$`(3.6)`$
For a vortex pair of charge $`\pm n`$, we insert in Eq. (2.11) both the vortex pair’s phase given in Eq. (3.1) and the regulated expression for $`A(\stackrel{}{r})`$ given in Eq. (3.3). Performing the spatial integration again out to $`R_{TF}`$ (see for details), we find the energy for a vortex pair to be:
$$E_n=\frac{\pi }{2C}\left[\mu ^2R_{TF}^2\frac{R_{TF}^6}{48}\right]\frac{d^2n^2\pi }{8C}\left[\frac{8\mu ^2}{1+d^2\mu /n^2}+\mathrm{ln}\left(\frac{R_{TF}^4\mu ^2(1+d^2/2R_{TF}^2)}{n^4(1+d^2\mu /n^2)}\right)\right].$$
$`(3.7)`$
Using Eqs. (3.6) and (3.7), we compute the energy difference between a trapped condensate with a symmetric pair of unit-charged, counter-rotating ($`n=1`$) vortices, and the no vortex ground state ($`n=0`$). To leading order in the limit of large $`C`$, we find
$$E_1E_0=\frac{2\pi \mu _0}{C}\mathrm{ln}\left(d^2\mu _0+1\right)+\mathrm{}$$
$`(3.8)`$
Note that the energy difference vanishes in the $`d0`$ limit, as expected. Also, the energy difference varies as the natural logarithm of the distance between the vortices (at large separation), which resembles results from earlier studies of vortices in homogeneous BEC .
Next, we calculate the impulse $`I_n`$ for the condensate with a symmetric pair of counter-rotating vortices with charge $`\pm n`$. (With no vortices, the trapped condensate impulse is zero.) Using Eq. (3.1) we find
$$|\stackrel{}{}\varphi |=\frac{nd}{[R^2d^2\mathrm{cos}^2\varphi +(R^2d^2/4)^2]^{\frac{1}{2}}}.$$
$`(3.9)`$
We insert this formula and the regulated expression for $`A^2(\stackrel{}{r})`$ \[Eq. (3.3)\] into Eq. (2.12) to determine the vortex pair impulse. We find the largest contribution to the spatial impulse integral comes at large $`R`$, and is thus dependent on the condensate size (set roughly by the Thomas-Fermi condensate edge, $`R_{TF}`$) as well as the vortex pair charge magnitude ($`n`$) and separation ($`d`$). Evaluating the integral out to $`R_{TF}`$, and assuming large $`C`$, we find the vortex pair impulse to be
$$I_n=\frac{nd\mu _0\pi }{C}\mathrm{ln}\left(\frac{64\mu _0}{ed^2}\right)+\mathrm{}$$
$`(3.10)`$
to leading order (see for details). Note that like the energy, the impulse for vortex pair creation vanishes in the small $`d`$ limit, and is proportional to the vortex charge, as expected. Also, since we are using a macroscopic hydrodynamic model in our calculations, the quantization of angular momentum of the BEC wavefunction does not lead to a simple quantization condition on the integrated impulse.
Assembling the results for the energy and impulse leads directly to an estimate of the critical velocity at which symmetrically-placed, oppositely-charged vortex pairs, at relative separation $`d`$, can form near the center of a Bose condensate trapped harmonically in two dimensions. In the large $`C`$ limit by using Eq. (2.13) we find
$$v_c\frac{2}{d}\frac{\mathrm{ln}(d^2\mu _0)}{\mathrm{ln}(\frac{64\mu _0}{ed^2})}$$
$`(3.11)`$
where again, $`\mu _0=\sqrt{\frac{C}{2\pi }}`$. These expressions indicate a weak dependence (via $`\mu _0`$ and $`C`$ within the logarithms) of $`v_c`$ on the BEC non-linear self-energy. Recall that the derivation of Eq. (3.11) assumes that $`1/\mu <d^2<\mu `$ in dimensionless harmonic oscillator units (h.o.u.). The first inequality is the requirement that the vortices be at least several healing lengths away from each other, and the second inequality is the requirement that they are not too far from the trap center.
These conditions seem to be largely met in the recent MIT critical velocity experiment using a trapped sodium BEC , although this experiment and our model have different spatial symmetries (i.e., geometries). Onset of energy dissipation in the MIT condensate was observed at a critical velocity when moving a repulsive barrier (a blue-detuned laser beam) through the middle of the cold atom cloud. The laser beam was directed along the smaller, radial direction of the cigar-shaped condensate, which had Thomas-Fermi diameters of 45 $`\mu `$m and 150 $`\mu `$m in the radial and axial directions, respectively. The laser barrier was moved back-and-forth along the condensate’s axial direction (perpendicular to the axis of the laser beam) at a constant speed, and energy dissipation was determined from measured changes in the condensate fraction. Thus the experiment was explicitly a three-dimensional system with anisotropic trapping in the plane perpendicular to the laser beam axis.
Ignoring the different geometry of our two-dimensional, isotropically-trapped BEC model, we use the analytical calculation outlined above to estimate the critical velocity for vortex pair creation in the MIT condensate. We assume the vortex cores are parallel to the axis of the blue-detuned laser beam, and are separated by a distance $`d`$ equal to the laser beam diameter (13 $`\mu `$m). In the central region of the MIT condensate, the number of atoms per unit length parallel to the laser beam axis ($`n_L`$) was about 3 x 10<sup>9</sup> cm<sup>-1</sup>. Using 2.9 nm for the sodium $`s`$-wave scattering length ($`a_s`$), we find from Eq. (2.5) that $`C8\pi a_sn_L25,000`$ (in h.o.u.) for the MIT BEC experiment.
Using these values in Eq. (3.11), we estimate that the onset of vortex pair formation, and hence energy dissipation, occurs at a critical velocity $`v_c`$ 0.4 mm/s, a factor of about four below the measured value . (Note: this estimated $`v_c`$ varies by $`30\%`$ over the range of trap frequencies present in the MIT experiment – from 18 Hz in the axial direction to 65 Hz in the radial direction.) We defer further interpretation of this estimate of $`v_c`$ to the conclusion.
IV. A Test of the Modified Thomas-Fermi Limit : As a test of the analytical calculation of vortex critical velocity using the modified Thomas-Fermi (TF) approximation, we consider a simple, symmetric example: a single vortex of charge $`n`$ at the center of an isotropic BEC that is harmonically-trapped in two dimensions. In this case, the cylindrical symmetry of the system requires the phase of the condensate to advance proportional to the polar angle $`\theta `$, and hence Eq. (2.9) limits the envelope function $`A(\stackrel{}{r})`$ to be a function of $`R=|\stackrel{}{r}|`$ alone. Also, the single-valuedness of the condensate wavefunction constrains the phase to be $`\varphi =n\theta `$, where $`n`$ is an integer (the vortex charge). Therefore, Eq. (2.8) reduces to a second order ordinary differential equation for $`A(R)`$ that can be integrated numerically, without making the modified Thomas-Fermi approximation, which can then be used to test the modified TF analytical calculation. We employed an adaptive mesh relaxation algorithm to perform the numerical integration of Eq. (2.8) for both $`n=0`$ and $`n=1`$, using the value $`C=200,000`$ for the coefficient of the non-linear term throughout. The computed $`A(R)`$ for $`n=0`$ is shown in Figure 1 and compared to the corresponding $`A(R)`$ calculated analytically using the traditional TF approximation.
Fig. 1: Calculated envelope function $`A(R)`$ for a condensate
with no vortex ($`n=0`$). Numerical and TF analytical
calculations give identical results on the scale of this graph.
Figure 2 provides a comparison of the numerical and modified TF analytical solutions for $`A(R)`$ with a single central vortex of charge $`n=1`$.
Fig. 2: Comparison of numerical and modified TF envelope functions $`A(R)`$
for a condensate with a single central vortex of unit charge $`(n=1)`$.
(a) Full radial extent of $`A(R)`$: differences not perceptible on this scale.
(b) $`A(R)`$ near the trap center. Dashed line is modified TF calculation.
In the modified TF limit, the analytical envelope function for a single central vortex of charge $`n`$ is
$$A(R)=\sqrt{\frac{\mu \frac{R^2}{4}\frac{n^2}{R^2}}{C}},\varphi (\theta )=n\theta .$$
$`(4.1)`$
This expression for $`A(R)`$ shows that including the leading kinetic energy term due to the vortex in the analytical calculation is equivalent to an effective potential barrier near the vortex core caused by the angular momentum of the condensate. In this approximation the condensate extends over the annulus $`R_0<R<R_{max}`$, outside of which $`A`$ vanishes. From Eq. (4.1) we see that
$$R_0^2=2\mu 2\sqrt{\mu ^2n^2}andR_{max}^2=2\mu +2\sqrt{\mu ^2n^2}.$$
$`(4.2)`$
Hence in the modified TF limit, the vortex core radius (in h.o.u.) is effectively $`n/\sqrt{\mu }`$ at large $`\mu `$. On general grounds, we expect the radius of the vortex core to be of the order of the condensate healing length, $`\xi =\frac{1}{8\pi a_sn_0}`$, where $`a_s`$ is the $`s`$-wave scattering length and $`n_0`$ is the local density. Thus the modified Thomas-Fermi limit reproduces the expected scaling of vortex core size, since $`\xi \frac{1}{\sqrt{\mu }}`$ in dimensionless units. Note also that at large $`R`$ (ignoring the trap potential) the effect of the vortex on the condensate envelope falls off as $`1/R^2`$ in the modified TF approximation, which is identical to the asymptotic behavior of the trial vortex wavefunction in homogeneous BEC used by Fetter .
The modified TF analytical calculation of the energy of a single central vortex of charge $`n`$ is found by using Eq. (4.1) in Eq. (2.11) (ignoring the $`\stackrel{}{}A`$ term) and integrating over the annulus given in Eq. (4.2):
$$E_n=\frac{4\pi (\mu ^2n^2)^{\frac{3}{2}}}{3C}.$$
$`(4.3)`$
\[Compare with Eq. (3.6).\] Although not obvious from this equation, $`E_n`$ increases as a function of $`n`$ due to the dependence of $`\mu `$ on $`n`$ through the normalization condition $`\mathrm{d}^2r|\psi |^2=1`$. In particular, we find this normalization requires
$$\frac{C}{2\pi }=\mu \sqrt{\mu ^2n^2}\frac{n^2}{2}\mathrm{ln}\left(\frac{\mu +\sqrt{\mu ^2n^2}}{\mu \sqrt{\mu ^2n^2}}\right),$$
$`(4.4)`$
hence causing $`R_{max}`$ (Eq. (4.2)) to increase with $`n`$. Next, using Eq. (4.1) in the analytical calculation of the impulse \[Eq. (2.12)\] of a single central vortex, we find
$$I_n=\frac{8\pi n}{3C}(\mu n)^{\frac{3}{2}}.$$
$`(4.5)`$
For the no-vortex case ($`n=0`$) and $`C=200,000`$, numerical integration of Eq. (2.3) yields the total condensate energy: $`E_0=118.965`$. \[All numbers here are in harmonic oscillator units (h.o.u.).\] With the analytical method we find $`E_0=\frac{4\pi \mu _0^3}{3C}=118.94`$, where $`\mu _0=\sqrt{\frac{C}{2\pi }}=178.4`$. For a single central vortex with $`n=1`$ and $`C=200,000`$, the numerical solutions give $`E_1=118.991`$ and $`I_1=0.0996`$, whereas the modified TF analytical calculation yields $`E_1=118.97`$ and $`I_1=0.0989`$. Using these values, we find $`v_c=\frac{\mathrm{\Delta }E}{\mathrm{\Delta }I}=0.26`$ (numerical) and 0.30 (analytical). \[Recall that the impulse integral is identically zero for the no vortex solution.\] This reasonable agreement suggests that the modified Thomas-Fermi approximation in the large $`C`$ limit should enable analytical calculations of 10-20% accuracy for vortex critical velocities in inhomogeneous BEC, provided the calculational model and experiment have the same spatial symmetries.
IV: Conclusion : In this paper we report an analytical calculation of the critical velocity for creation of a pair of counter-rotating vortices in harmonically-trapped, dilute Bose condensates. Excitation of vortices may set the lowest limit for a superfluid critical velocity in BEC. For the analytical calculation presented here, we developed a modified Thomas-Fermi approximation to the standard mean-field theory . This approximation is appropriate to the limit of large interparticle interactions, and thus is relevant to most current experiments. The approximation consists of two steps, described above in Section II: (i) solve a continuity equation for the phase of the condensate wavefunction; and then (ii) insert this solution for the phase in an equation of hydrostatic equilibrium and solve for the condensate envelope function, ignoring the second order spatial derivative of the envelope function. In essence, we include contributions to the condensate kinetic energy that come from gradients of the phase (a signature of vortices), but assume that vortex-induced gradients in the condensate envelope function are negligible. Alternatively, one can think of the modified Thomas-Fermi approximation as equivalent to treating BEC like a fluid with an exotic equation of state, with the leading vortex kinetic energy terms acting as an effective potential. Once the envelope function is determined, we compute approximately the energy and integrated linear momentum (or impulse) of the vortex pair, and determine the critical velocity from the Landau criterion, $`v_c=min(energy/impulse)`$.
As described in Section III, we find qualitative agreement between our analytical calculation and the BEC excitation critical velocity recently measured by Ketterle and co-workers at MIT . This result is not surprising given the obvious difference in symmetry between our model and the MIT critical velocity experiment: we assume the vortex core axes to be parallel to the axis of cylindrical symmetry in a two-dimensional trap confining the condensate; whereas, the experimental excitation would likely create vortices with core axes perpendicular to the finite cylindrical axis in the actual, quasi-two-dimensional trap. In addition, vortices produced at velocities substantially below the observed critical velocity may collectively dissipate energy at a rate comparable to the decay rate of the unstirred condensate itself, making it difficult to identify $`v_c`$ experimentally (i.e., the MIT experiment may only place an upper limit on $`v_c`$). We expect that detailed understanding of the role of vortex creation in the breakdown of superfluidity in BEC will require further experimental development to enable more symmetric excitation of the condensate, as well as theoretical treatments (both numerical and analytical) of more realistic systems. Nevertheless, as shown in Section IV, the modified Thomas-Fermi analytical method is in reasonable agreement with numerical calculations based on the full GP equation of the critical velocity for a vortex with high symmetry: specifically, a single vortex in the center of a two-dimensional harmonic trap.
We note two important issues that are not included in the analysis presented in this paper: (i) the breakdown of the hydrodynamic approximation at the edges of the trapped condensate; and (ii) the dynamics of vortex excitation in the presence of nominally smooth trapping and perturbing potentials. Further theoretical and experimental work is necessary on these problems.
We conclude by emphasizing that the modified Thomas-Fermi approximation presented in this paper is straightforward to implement. In addition to enabling an analytical calculation of vortex critical velocity (for more details see Ref. ) this approximation should have other applications in the physics of trapped BEC.
We thank E. Timmermans for helpful discussions.
References
1. See, for example, I. M. Khalatnikov, Introduction to the Theory of Superfluidity (Benjamin, New York, 1965).
2. A nice discussion of the breakdown of superfluidity in liquid <sup>4</sup>He, including many important references, is given in: R.K. Pathria, Statistical Mechanics (Pergamon Press, Oxford, 1972).
3. R.P. Feynman, in Progress in Low Temperature Physics (North-Holland, Amsterdam, 1955), Vol. 1, p. 17.
4. A. Amar, Y. Sasaki, R.L. Lozes, J.C. Davis, and R.E. Packard, Phys. Rev. Lett. 68, 2624 (1992); O. Avenel and E. Varoquaux, Phys. Rev. Lett. 55, 2704 (1985).
5. P.W. Anderson, Rev. Mod. Phys. 38, 298 (1966); E.R. Huggins, Phys. Rev. A 1, 332 (1970); L.J. Campbell, J. Low Temp. Phys. 8, 105 (1972).
6. C. Raman, M. Köhl, R. Onofrio, D.S. Durfee, C.E. Kuklewicz, Z. Hadzibabic, and W. Ketterle, Phys. Rev. Lett. 83, 2502 (1999).
7. A.L. Fetter, Phys. Rev. Lett. 10, 507 (1963); M.P. Kawatra and R.K. Pathria, Phys. Rev. 151, 132 (1966).
8. T. Frisch, Y. Pomeau, and S. Rica, Phys. Rev. Lett. 69, 1644 (1992).
9. C. Huepe and M.-E. Brachet, C.R. Acad. Sci. Paris 325, 195 (1997).
10. B. Jackson, J.F. McCann, and C. S. Adams, Phys. Rev. Lett. 80, 3903 (1998).
11. T. Winiecki, J.F. McCann, and C. S. Adams, Phys. Rev. Lett. 82, 5186 (1999).
12. V.L. Ginzburg and L.P. Pitaevskii, Sov. Phys. JETP 7, 858 (1958); E.P. Gross, J. Math. Phys. 4, 195 (1963).
13. See discussion in the review article: F. Dalfovo, S. Giorgini, L.P. Pitaevskii, and S. Stringari, Rev. Mod. Phys. 71, 463 (1999).
14. A.L. Fetter, Phys. Rev. 138, A429 (1965).
15. M. Crescimanno and R. Walsworth, to appear. |
warning/0001/cond-mat0001441.html | ar5iv | text | # Magnon dispersions in quantum Heisenberg ferrimagnetic chains at zero temperature
## I Introduction
In the last decade a large variety of quasi-one-dimensional (1D) mixed-spin compounds with ferrimagnetic properties has been synthesized. Most of them are molecular magnets containing two different transition-metal magnetic ions which are alternatively distributed on the lattice (Fig. 1). For example, two families of such compounds read ACu(pba)(H<sub>2</sub>O)$`{}_{3}{}^{}`$nH<sub>2</sub>O and ACu(pbaOH)(H<sub>2</sub>O)$`{}_{3}{}^{}`$nH<sub>2</sub>O, where pba = 1,3-propylenebis (oxamato), pbaOH = 2-hydroxo-1,3-propylenebis (oxamato), and A=Ni, Fe, Co, and Mn. Published experimental work implies that the magnetic properties of these mixed-spin materials are basically described by a quantum Heisenberg spin model with antiferromagnetically coupled nearest-neighbor localized spins:
$``$ $`=`$ $`J{\displaystyle \underset{n=1}{\overset{N}{}}}\left[𝐒_1(n)+𝐒_1(n+1)\right]𝐒_2(n)`$ (1)
$``$ $`\mu _BH{\displaystyle \underset{n=1}{\overset{N}{}}}\left[g_1S_1^z(n)+g_2S_2^z(n)\right],J>0.`$ (2)
Here the integers $`n`$ number the $`N`$ elementary cells, each of them containing two site spins with quantum numbers $`S_1>S_2`$. The $`g`$-factors, related to the spins $`S_1`$ and $`S_2`$, are denoted by $`g_1`$ and $`g_2`$, respectively. $`\mu _B`$ is the Bohr magneton, and $`H`$ is the external uniform magnetic field applied along the $`z`$ direction. As an example, the following values for the parameters in Eq. (1) have been extracted from magnetic measurements on the recently synthesized quasi-1D bimetallic compound NiCu(pba)(D<sub>2</sub>O)$`{}_{3}{}^{}`$2D<sub>2</sub>O: $`(S_1,S_2)(S_{Ni},S_{Cu})=(1,\frac{1}{2})`$, $`J/k_B=121K`$, $`g_1g_{Ni}=2.22`$, $`g_2g_{Cu}=2.09`$. In view of the recent developments in the physics of uniform quasi-1D systems (concerning, in particular, the behavior of antiferromagnetic chains and ladders in external magnetic fields: see, e.g., Ref. and references therein), ferrimagnetic chains and ladders open an interesting new area. The point is that the presence of two or more different quantum spins in the elementary cell considerably increases the number of situations of interest: the topology of spin arrangements plays an essential role in the structure of the ground state and low-energy excitations. For example, the diamond lattice in Fig. 1 represents another interesting class of Heisenberg ferrimagnetic chains constructed of one kind of antiferromagnetically coupled site spins but with a different number of lattice sites in the $`𝒜`$ and $``$ sublattices. It is remarkable that only in the last few years interest in the physical properties of 1D quantum Heisenberg ferrimagnets has considerably increased: most of the early efforts were concentrated on the chemistry of molecular magnets and a relatively small amount of work was devoted to physical properties.
Since interactions in the models presented in Fig. 1 set up bipartite lattices, the Lieb-Mattis theorem is applicable. In particular, for the mixed-spin system it predicts the existence of a ferrimagnetic ground state with the total-spin quantum number $`S_g=(S_1S_2)N`$. Thus the model (1) exhibits long-range ordered magnetic ground state characterized by the magnetization density $`M_0=(S_1S_2)/2a_0`$, $`a_0`$ being the lattice constant. Such a magnetic phase may be referred to as a quantized unsaturated ferromagnetic phase: it is characterized by both the quantized ferromagnetic order parameter $`𝐌`$ (quantized in integral or half-integral multiples of the number of elementary cells, $`M=M_0N`$) and the macroscopic sublattice magnetizations $`𝐌_A=_{n=1}^N𝐒_1(n)`$ and $`𝐌_B=_{n=1}^N𝐒_2(n)`$. What makes such a state and the related transitions to magnetically disordered states interesting is the fact that the ferromagnetic order parameter is a conserved quantity: in Heisenberg systems this is expected to imply strong constraints on the critical behavior. Consequences of the spontaneous symmetry breaking $`SO(3)SO(2)`$ (related to the establishment of a ferrimagnetic ground state) for the structure of low-energy excitations are dictated by the nonrelativistic version of Goldstone’s theorem: in the absence of long-range forces, a spontaneous symmetry breaking leads to low-energy excitations whose energy tends to zero for wave-vectors $`k0`$. In contrast to the relativistic version, the theorem does neither specify the exact form of the dispersion relation for small $`k`$, nor does it determine the number of different Goldstone modes: these features are not fixed by symmetry considerations alone – rather, they depend on the specific properties of the system.
Turning to the case of Heisenberg ferromagnets, in a hydrodynamic description the quadratic form of the magnon energies
$$E_k=\frac{\rho _s}{M_0}k^2+𝒪(k^4)$$
(3)
results entirely from the symmetry of the ferromagnetic state and the fact that the order parameter is itself a constant of the motion. Here $`M_0`$ and $`\rho _s`$ are, respectively, the magnetization density and the spin-stiffness constant of the Heisenberg ferromagnet. This form of the Goldstone modes (which is just the Landau-Lifshitz formula) can rigorously be obtained by simple sum-rule arguments (see, e.g., Ref. ). Similar arguments are applicable to Heisenberg ferrimagnets as well. An alternative approach, relying on the suggested conformal invariance of the model (1), also predicts the quadratic form of Eq. (3). Due to the doubling of elementary cell in the ferrimagnetic ground state, there appears a second spin-wave branch in the ferrimagnetic case (optical magnons) with a finite energy gap $`\mathrm{\Delta }`$ at $`k=0`$: $`\mathrm{\Delta }M_0/\chi _{||}`$, where $`\chi _{||}`$ is the magnetic susceptibility parallel to $`𝐌`$. At intermediate temperatures, it is the optical magnon branch which produces a number of specific thermodynamic properties characteristic especially for the ferrimagnetic systems.
The existence of a macroscopic magnetic ground state also opens an interesting and rare opportunity to apply the spin-wave theory (SWT) to the low-dimensional quantum spin system (1). Indeed, recently published analysis based on the linear spin-wave approximation qualitatively confirmed the expected structure of low-energy excitations in Heisenberg ferrimagnets. Moreover, the second-order SWT was shown to produce precise quantitative results for the ground-state energy $`E_0`$ and the on-site magnetizations $`m_1=M_A/N`$ and $`m_2=M_B/N`$ even for the extreme quantum system $`(S_1,S_2)=(1,\frac{1}{2})`$: the second-order SWT results differ by less than $`0.017\%`$ for $`E_0`$, and by less than $`0.18\%`$ for $`m_1`$, from the density matrix renormalization group (DMRG) estimates. In this respect, an interesting question is to what extend the spin-wave approach can effectively be used to describe the properties of the 1D model (1) at zero temperature. A purpose of the present paper is to demonstrate, through an explicit study of the magnon-dispersion perturbation series, that the spin-wave approach can produce precise quantitative results for the magnon dispersions as well.
The paper is organized as follows. In Section 2, using the Dayson-Maleev boson representation of spin operators, the original spin Hamiltonian (1) is transformed to an equivalent boson Hamiltonian. The choice of an appropriate zeroth-order quadratic Hamiltonian for the perturbation series is also discussed here. In Section 3 we study the second-order corrections for the magnon energies and the related corrections for the spin-stiffness constant and the optical magnon gap. Third-order self-energy diagrams and the respective corrections for the optical magnon gap are also considered here. In the last Section the results are summarized and a comparison with available numerical estimates is made.
## II Dyson-Maleev formalism
To develop a spin-wave theory (see, e.g., Refs. and references therein), one firstly transforms the original spin Hamiltonian (1) to an equivalent boson Hamiltonian. We adopt the following Dyson-Maleev representation of spin operators:
$`S_1^+(i)`$ $`=`$ $`\sqrt{2S_1}\left(a_i{\displaystyle \frac{1}{2S_1}}a_i^{}a_ia_i\right),`$ (4)
$`S_1^{}(i)`$ $`=`$ $`\sqrt{2S_1}a_i^{},S_1^z(i)=S_1a_i^{}a_i,`$ (5)
$`S_2^+(j)`$ $`=`$ $`\sqrt{2S_2}\left(b_j^{}{\displaystyle \frac{1}{2S_2}}b_j^{}b_j^{}b_j\right),`$ (6)
$`S_2^{}(j)`$ $`=`$ $`\sqrt{2S_2}b_j,S_2^z(j)=b_j^{}b_jS_2,`$ (7)
where $`S_\alpha ^\pm =S_\alpha ^x\pm ıS_\alpha ^y`$, $`\alpha =1,2`$. $`a_i`$ and $`b_j`$ are boson operators defined on the lattice sites $`i𝒜`$ and $`j`$, respectively.
Substituting the latter expressions in Eq. (1), one finds the following boson representation of $``$ in terms of the Fourier transforms $`a_k`$ and $`b_k`$ of $`a_i`$ and $`b_j`$:
$$_B=2\sigma S^2NJ+_0^{^{}}+V_{DM}^{^{}},$$
(8)
where
$$_0^{^{}}=2SJ\underset{k}{}\left[a_k^{}a_k+\sigma b_k^{}b_k+\sqrt{\sigma }\gamma _k\left(a_k^{}b_k^{}+a_kb_k\right)\right],$$
(9)
and
$`V`$ $`{}_{DM}{}^{^{}}={\displaystyle \frac{J}{N}}{\displaystyle \underset{14}{}}\delta _{12}^{34}\times `$ (10)
$`(`$ $`2\gamma _{14}a_3^{}a_2b_1^{}b_4+\sqrt{\sigma }\gamma _{1+24}a_3^{}b_2^{}b_1^{}b_4+{\displaystyle \frac{1}{\sqrt{\sigma }}}\gamma _4a_3^{}a_2a_1b_4).`$ (11)
$`\gamma _k=\mathrm{cos}(ka_0)`$ is the lattice structure factor and $`\delta _{12}^{34}\delta (1+234)`$ is the Kronecker $`\delta `$-function. We have used the convention $`(k_1,k_2,k_3,k_4)(1,2,3,4)`$ and the notations $`\sigma S_1/S_2>1`$ and $`S_2S`$. Here and in what follows the external magnetic field is $`H=0`$.
In the above expressions $`_0^\mathrm{`}=𝒪(S)`$ is the quadratic boson Hamiltonian of the linear spin-wave theory (LSWT) and $`V_{DM}^{^{}}=𝒪(1)`$ is the quartic Dyson-Maleev boson interaction. Omitting completely the boson interaction $`V_{DM}^{^{}}`$, a number of authors has recently used the quadratic LSWT Hamiltonian $`_0^\mathrm{`}`$ for a qualitative description of the 1D ferrimagnetic model (1) at zero temperature. $`_0^\mathrm{`}`$ is easily diagonalized by use of the following Bogoliubov transformation to the quasiparticle boson operators $`\alpha _k`$ and $`\beta _k`$:
$`a_k`$ $`=`$ $`u_k(\alpha _kx_k\beta _k^{}),b_k=u_k(\beta _kx_k\alpha _k^{}),`$ (12)
$`u_k`$ $`=`$ $`\sqrt{{\displaystyle \frac{1+\epsilon _k}{2\epsilon _k}}},x_k={\displaystyle \frac{\eta _k}{1+\epsilon _k}},`$ (13)
where $`\epsilon _k=\sqrt{1\eta _k^2}`$ and $`\eta _k=2\sqrt{\sigma }\gamma _k/(1+\sigma )`$.
As a matter of fact, for the 1D ferrimagnetic model (1) the quadratic Hamiltonian (9) is not the most appropriate choice of a starting zeroth-order approximation for perturbation series. Since we are also interested in systems with small site spins, it is more appropriate from the very beginning to redefine $`_0^{^{}}`$ by adding the first-order $`1/S`$ corrections to magnon dispersions. The latter corrections come entirely from a normal ordering of the quasiparticle boson operators $`\alpha _k`$ and $`\beta _k`$ in the quartic interaction $`V_{DM}^{^{}}`$. It is important that these corrections (similar to Oguchi’s corrections in antiferromagnets) renormalize the magnon excitation spectra (and the ground-state energy) without changing their basic structure, i.e., the number of Goldstone modes. Unlike the case of Heisenberg antiferromagnets, where Oguchi’s corrections are numerically small even for site spins $`S=\frac{1}{2}`$, in quantum 1D ferrimagnets the magnon dispersions are significantly renormalized (see below). As a result of the normal-ordering procedure, the boson Hamiltonian (8) is recast to the following basic form:
$$_B=E_0+_0+\lambda V,V=V_2+V_{DM},\lambda =1.$$
(14)
The ground-state energy $`E_0`$ calculated up to first order in $`1/S`$ reads
$`{\displaystyle \frac{E_0}{2NJ}}`$ $`=`$ $`\sigma S^2+S\left[2\sqrt{\sigma }a_1+(\sigma +1)a_2\right]`$ (15)
$``$ $`a_1^2a_2^2{\displaystyle \frac{\sigma +1}{\sqrt{\sigma }}}a_1a_2,`$ (16)
where
$$a_1=\frac{\sqrt{\sigma }}{2(\sigma +1)}\frac{1}{N}\underset{k}{}\frac{\gamma _k^2}{\epsilon _k},a_2=\frac{1}{2}+\frac{1}{2N}\underset{k}{}\frac{1}{\epsilon _k}.$$
(17)
Expressed in terms of quasiparticle operators, the corrected LSWT Hamiltonian reads
$$_0=2SJ\underset{k}{}\left[\omega _k^{(\alpha )}\alpha _k^{}\alpha _k+\omega _k^{(\beta )}\beta _k^{}\beta _k\right],$$
(18)
where
$$\omega _k^{(\alpha ,\beta )}=\left(1\frac{a_1}{S\sqrt{\sigma }}\right)\left(\frac{\sigma +1}{2}\epsilon _k\frac{\sigma 1}{2}\right)a_2\frac{1\gamma _k^2}{\epsilon _kS}.$$
(19)
The normal-ordered quasiparticle interaction $`V`$ contains a quadratic term,
$$V_2=J\underset{k}{}\left[V_k^{(+)}\alpha _k^{}\beta _k^{}+V_k^{()}\alpha _k\beta _k\right],$$
(20)
where
$$V_k^{(+,)}=\frac{a_2(\sigma 1)^2}{\sqrt{\sigma }(\sigma +1)}\left(1\pm \frac{\sigma +1}{\sigma 1}\epsilon _k\right)\frac{\gamma _k}{\epsilon _k},$$
(21)
and the quartic normal-ordered Dyson-Maleev interaction $`V_{DM}`$ containing nine vertex functions: $`V^{(i)}=V_{12;34}^{(i)}`$, $`i=1,\mathrm{},9`$. Explicit expressions for $`V_{DM}`$ and the related vertex functions are presented in Appendix A.
The quadratic Hamiltonian (18) describes two branches of spin-wave excitations (magnons) defined with the dispersion relations $`E_k^{(\alpha ,\beta )}=2SJ\omega _k^{(\alpha ,\beta )}`$, Eq. (19). The $`\alpha `$ excitations are gapless acoustical magnons in the subspace with $`S_{tot}^z=(S_1S_2)N1`$, whereas the $`\beta `$ excitations are gapful optical magnons in the subspace $`S_{tot}^z=(S_1S_2)N+1`$: $`𝐒_{tot}=_{n=1}^N\left[𝐒_1(n)+𝐒_2(n)\right]`$. It is easy to show that the effect of the external magnetic field in Eq. (1) on the dispersions $`E_k^{(\alpha ,\beta )}`$ reduces to the simple substitution $`E_k^{(\alpha ,\beta )}E_k^{(\alpha ,\beta )}\pm g\mu _BH`$ provided that $`g_1=g_2g`$.
The spin stiffness constant $`\rho _s`$ (related to the acoustical branch) plays a basic role (together with the magnetization density $`M_0`$) in the low-temperature thermodynamics of the model (1) and can be obtained from the Landau-Lifshitz relation (3) and Eq. (19):
$$\rho _s^{(0)}=Ja_0S_1S_2\left[1\frac{a_1}{S\sqrt{\sigma }}a_2\frac{\sigma +1}{S\sigma }\right].$$
(22)
In the same free-quasiparticle approximation, the spectral gap of optical magnons at $`k=0`$ reads
$$\mathrm{\Delta }_0=2J(S_1S_2)\left(1\frac{a_1}{S\sqrt{\sigma }}\right).$$
(23)
In the case $`(S_1,S_2)=(1,\frac{1}{2})`$ Eqs. (22) and (23) give $`\rho _s^{(0)}/Ja_0S_1S_2=0.761`$ and $`\mathrm{\Delta }_0=1.676J`$. On the other hand, the LSWT Hamiltonian $`_0^{^{}}`$ produces the parameters of the related classical system: $`\rho _s^{(0)}/Ja_0S_1S_2=1`$ and $`\mathrm{\Delta }_0=J`$. The numerical estimate for the gap $`\mathrm{\Delta }=1.759J`$ clearly demonstrates the importance of the $`1/S`$ corrections in Eq. (19).
Next, let us consider the macroscopic sublattice magnetizations $`m_1=S_1^z(i)=S_1a_i^{}a_i`$ and $`m_2=S_2^z(j)=S_2+b_j^{}b_j`$ which are finite in the ferrimagnetic state. In the case of small site spins and for small magnetization densities $`M_0`$, $`m_1`$ and $`m_2`$ are strongly reduced as compared to their classical values (respectively, $`S_1`$ and $`S_2`$):
$$m_1=S_1a_2,m_1+m_2=S_1S_2.$$
(24)
As an example, for $`(S_1,S_2)=(1,\frac{1}{2})`$ Eq. (24) predicts a $`42\%`$ reduction of the small spin $`S_2=\frac{1}{2}`$. However, the off-diagonal quadratic interaction $`V_2`$ in Eq. (14) produces important first-order corrections for the sublattice magnetizations $`m_1`$ and $`m_2`$. Thus, to first order in $`1/S`$ Eq. (24) should be replaced by the more precise expression
$$m_1=S_1a_2\frac{\sqrt{\sigma }}{2S(\sigma +1)^2}\frac{1}{N}\underset{k}{}\left(V_k^++V_k^{}\right)\frac{\gamma _k}{\epsilon _k^2}.$$
(25)
Now for the ferrimagnetic chain $`(1,\frac{1}{2})`$ the last expression gives the result $`m_1=0.816`$ which differs by only $`3\%`$ from the DMRG numerical estimate $`0.79248`$.
Summarizing, it may be stated that the free-quasiparticle approximation based on the Hamiltonian (18) gives a good qualitative description of the ground-state properties of the model (1). Further improvement of the SWT results may be achieved by considering the role of quasiparticle interactions.
## III Role of the quasiparticle interactions
Since the first-order $`1/S`$ corrections for the ground state energy $`E_0`$ and the dispersions $`E_k^{(\alpha ,\beta )}`$ have already been taken into account by the normal-ordering procedure, corrections to Eqs. (15) and (19) arise only up from the second order of the perturbation series in $`V`$.
### A Second-order corrections for the magnon energies
The second-order corrections for the dispersion of acoustical magnons $`\omega _k^{(\alpha )}`$, Eq. (19), are connected with the self-energy diagrams in Fig. 2. The respective analytic expressions read
$$\delta \omega _k^{(\alpha )}(a)=\frac{1}{(2S)^2}\frac{V_k^{(+)}V_k^{()}}{\omega _k^{(\alpha )}+\omega _k^{(\beta )}},$$
(26)
$$\delta \omega _k^{(\alpha )}(bc)=\frac{1}{(2S)^2}\frac{2}{N}\underset{p}{}\frac{V_p^{(+)}V_{kp;pk}^{(2)}+V_p^{()}V_{kp;pk}^{(3)}}{\omega _p^{(\alpha )}+\omega _p^{(\beta )}},$$
(27)
$$\delta \omega _k^{(\alpha )}(d)=\frac{1}{(2S)^2}\frac{2}{N^2}\underset{24}{}\delta _{k2}^{34}\frac{V_{43;2k}^{(8)}V_{k2;34}^{(7)}}{\omega _k^{(\alpha )}+\omega _2^{(\alpha )}+\omega _3^{(\beta )}+\omega _4^{(\beta )}},$$
(28)
$$\delta \omega _k^{(\alpha )}(e)=\frac{1}{(2S)^2}\frac{2}{N^2}\underset{24}{}\delta _{k2}^{34}\frac{V_{43;2k}^{(3)}V_{k2;34}^{(2)}}{\omega _k^{(\alpha )}+\omega _2^{(\beta )}+\omega _3^{(\alpha )}+\omega _4^{(\alpha )}}.$$
(29)
Note that since the vertex functions $`V_k^{()}`$, $`V_{kp;pk}^{(2)}`$, $`V_{kp;pk}^{(3)}`$, $`V_{43;2k}^{(8)}`$, and $`V_{43;2k}^{(3)}`$ vanish at the zone center $`k=0`$ (see Appendix A), the gapless structure of $`\omega _k^{(\alpha )}`$ is preserved by each of the second-order corrections, Eqs. (26) to (29).
The second-order corrections for the dispersion of optical magnons $`\omega _k^{(\beta )}`$, Eq. (19), come from similar diagrams (see Fig. 3). The first diagram gives $`\delta \omega _k^{(\beta )}(a)=\delta \omega _k^{(\alpha )}(a)`$, whereas the other four diagrams give the following contributions:
$$\delta \omega _k^{(\beta )}(bc)=\frac{1}{(2S)^2}\frac{2}{N}\underset{p}{}\frac{V_p^{(+)}V_{kp;pk}^{(5)}+V_p^{()}V_{kp;pk}^{(6)}}{\omega _p^{(\alpha )}+\omega _p^{(\beta )}},$$
(30)
$$\delta \omega _k^{(\beta )}(d)=\frac{1}{(2S)^2}\frac{2}{N^2}\underset{24}{}\delta _{k2}^{34}\frac{V_{43;2k}^{(7)}V_{k2;34}^{(8)}}{\omega _k^{(\beta )}+\omega _2^{(\beta )}+\omega _3^{(\alpha )}+\omega _4^{(\alpha )}},$$
(31)
$$\delta \omega _k^{(\beta )}(e)=\frac{1}{(2S)^2}\frac{2}{N^2}\underset{24}{}\delta _{k2}^{34}\frac{V_{43;2k}^{(5)}V_{k2;34}^{(6)}}{\omega _k^{(\beta )}+\omega _2^{(\alpha )}+\omega _3^{(\beta )}+\omega _4^{(\beta )}}.$$
(32)
Note that in the present perturbation scheme (using $`_0`$ as a zeroth-order Hamiltonian), in Eqs. (26) to (32) there appear the renormalized dispersions $`\omega _k^{(\alpha ,\beta )}`$. The standard $`𝒪(1/S^2)`$ corrections to the magnon dispersions can easily be obtained by substituting in Eqs. (26) to (32) the bare excitation energies (i.e., the functions $`\omega _k^{(\alpha ,\beta )}`$ without the $`1/S`$ corrections). Since we are also interested in the extreme quantum systems which are composed of small site spins and have small magnetization densities $`M_0`$ \[such as $`(1,\frac{1}{2})`$ and $`(\frac{3}{2},1)`$\], it may be expected that the adopted perturbation scheme, where the quasiparticle interaction $`V`$ is treated as a small perturbation, is more appropriate. Such a viewpoint is similar to Oguchi’s treatment of the Heisenberg antiferromagnet and is supported by the following observations: (i) the higher-order corrections to the principle approximation for $`\omega _k^{(\alpha ,\beta )}`$ are numerically small (see below), and (ii) the third-order series in $`V`$ gives a somewhat better result for the gap $`\mathrm{\Delta }`$ in the extreme quantum system $`(1,\frac{1}{2})`$. As a matter of fact, noticeable deviations from the standard $`1/S`$ expansions appear only in the above mentioned extreme quantum cases.
The dispersion functions $`E_k^{(\alpha ,\beta )}`$ for the system $`(1,\frac{1}{2})`$, calcutated up to second order in $`V`$, are presented in Fig. 4. For the optical magnon branch we find an excellent agreement with the exact-diagonalization results in the whole Brillouin zone. Recently, a very successful description of the optical magnon branch of the system $`(1,\frac{1}{2})`$ has also been achieved through the variational matrix product approach.. As to the acoustical branch, the agreement with the available quantum Monte Carlo results is not so satisfactory, especially near the zone boundary at $`k=\pi /2a_0`$. It is worth noticing that in the above region the calculated second-order corrections to $`\omega _k^{(\alpha )}`$, Eq. (19), are very small (about $`0.5\%`$ of the principal approximation). In principle, it is not excluded that the spin-wave expansion partially fails to describe the acoustical branch in the system $`(1,\frac{1}{2})`$, as the latter is characterized by the minimal magnetization density $`M_0=1/4a_0`$ (see below). To clearify the problem, further studies are needed in this direction.
The above second-order corrections to $`\omega _k^{(\alpha ,\beta )}`$ can be used to find the coefficient $`r_2`$ in the perturbation series for the spin-stiffness constant $`\rho _s`$
$$\frac{\rho _s}{Ja_0S_1S_2}=1+\frac{r_1}{2S}+\frac{r_2}{(2S)^2}+𝒪\left(\frac{1}{S^3}\right)$$
(33)
and the coefficient $`\delta _2`$ in the series for the optical magnon gap
$$\frac{\mathrm{\Delta }}{2J(S_1S_2)}=1+\frac{\delta _1}{2S}+\frac{\delta _2}{(2S)^2}+\frac{\delta _3}{(2S)^3}+𝒪\left(\frac{1}{S^4}\right).$$
(34)
Here $`r_1=2a_1/\sqrt{\sigma }2a_2(\sigma +1)/\sigma `$ and $`\delta _1=2a_1/\sqrt{\sigma }`$ are obtained from Eqs. (22) and (23), respectively. The results for $`r_2`$ and $`\delta _2`$ are presented in Table 1.
### B Third-order corrections for the optical magnon gap
The third-order corrections for the optical magnon gap $`\mathrm{\Delta }`$ are connected with the self-energy diagrams in Fig. 5. These can be obtained by drawing all connected irreducible self-energy diagrams containing three vertex functions and only oppositely-oriented ($`\alpha `$, $`\beta `$) pairs of internal magnon lines. The diagrams containing magnon lines closing on themselves vanish since the vertices have already been normal ordered. In addition, the diagrams which have internal $`\alpha `$ lines carrying the momentum $`k`$ give zero contributions as well, since all vertex functions containing in-going $`\alpha `$ lines with momentum $`k`$ vanish at $`k=0`$ (see Appendix A).
The analytic expressions related to the diagrams in Fig. 5 can easily be obtained by means of standard diagrammatic rules (see, e.g., Ref. ). For example, the diagrams $`a_1`$, $`d_1`$, and $`e_1`$ in Fig. 5 give the following contributions for the dispersion of optical magnons:
$$\delta \omega _k^{(\beta )}(a_1)=\frac{1}{(2S)^3}\frac{2}{N}\underset{p}{}\frac{V_p^{(+)}V_{pk;pk}^{(4)}V_p^{()}}{\left[\omega _p^{(\alpha )}+\omega _p^{(\beta )}\right]^2},$$
(35)
$$\delta \omega _k^{(\beta )}(d_1)=\frac{1}{(2S)^3}\frac{2}{N^3}\underset{15}{}\delta _{12}^{34}\delta _{12}^{k5}\frac{V_{12;34}^{(7)}V_{34;5k}^{(9)}V_{k5;12}^{(8)}}{\left[\omega _1^{(\alpha )}+\omega _2^{(\alpha )}+\omega _3^{(\beta )}+\omega _4^{(\beta )}\right]\left[\omega _1^{(\alpha )}+\omega _2^{(\alpha )}+\omega _k^{(\beta )}+\omega _5^{(\beta )}\right]},$$
(36)
$$\delta \omega _k^{(\beta )}(e_1)=\frac{1}{(2S)^3}\frac{4}{N^2}\underset{24}{}\delta _{k2}^{34}\frac{V_4^{(+)}V_{34;2k}^{(5)}V_{2k;43}^{(4)}}{\left[\omega _4^{(\alpha )}+\omega _4^{(\beta )}\right]\left[\omega _k^{(\beta )}+\omega _2^{(\alpha )}+\omega _3^{(\beta )}+\omega _5^{(\beta )}\right]}.$$
(37)
The results for $`\delta _3`$ in systems with different site spins $`(S_1,S_2)`$ are collected in Table 1. It is seen that the third-order corrections for $`\mathrm{\Delta }`$ are numerically small (as compared to the second-order correction) even for the extreme quantum system $`(S_1,S_2)=(1,\frac{1}{2})`$.
## IV Summary of the results and discussion
The spin-wave results for the spin-stiffness constant $`\rho _s`$ and the magnon gap $`\mathrm{\Delta }`$ for a number of combinations $`(S_1,S_2)`$ are summarized in Table 1. The results for $`\mathrm{\Delta }`$ are compared with available DMRG estimates. We find an excellent agreement with the numerical estimates for a number of systems, the largest deviation (about 0.88$`\%`$) being connected with the system $`(1,\frac{1}{2})`$.
Since in the extreme quantum cases the constructed perturbation expansions basically rely on the suggested smallness of the quasiparticle interaction $`V`$, it is instructive to check the self-consistency of the theory by writing the series for $`\rho _s`$ and $`\mathrm{\Delta }`$ in terms of the formal parameter $`\lambda =1`$, Eq. (14). Using the results from Table 1, the series for the system $`(S_1,S_2)=(1,\frac{1}{2})`$ read
$$\frac{\rho _s}{Ja_0S_1S_2}=0.7609\lambda ^0+0.0283\lambda ^2+𝒪\left(\lambda ^3\right),$$
(38)
$$\frac{\mathrm{\Delta }}{2J(S_1S_2)}=1.6756\lambda ^0+0.1095\lambda ^20.0107\lambda ^3+𝒪\left(\lambda ^4\right).$$
(39)
The above expansions explicitly demonstrate that the quasiparticle interaction $`V`$ in Eq. (14) introduces numerically small corrections to the zeroth-order principle approximation based on the quadratic quasiparticle Hamiltonian $`_0=𝒪(\lambda ^0)`$, Eq. (18). Of course, the smallness of corrections by itself does not ensure a good quality of the spin-wave expansion. The main weakness of the spin-wave theory is connected with the assumption that the long-range order is well established: it includes only transverse spin fluctuations, whereas the longitudinal spin fluctuations are completely neglected. In this respect, a typical example arises when spin-wave expansions are used for magnetic systems near the order-disorder critical point where the transverse and longitudinal fluctuations should be treated on equal ground. In spite of the fact that the corrections are small, the spin-wave series give unsatisfactory quantitative results. On the other hand, as the distance from the critical point increases the spin-wave description becomes more and more reliable. The indicated discrepancy for the acoustical branch in the extreme quantum system $`(1,\frac{1}{2})`$ (although the second-order corrections to the principle approximation are numerically small) probably reflects the discussed weakness of the SWT. As a matter of fact, taking the parameter $`M_0`$ as a measure of the distance from the disordered phase, such a behavior of the spin-wave series in the $`(S_1,S_2)`$ family of 1D quantum Heisenberg ferrimagnets can also be indicated for the magnon gap $`\mathrm{\Delta }`$ (see Table 1) and the parameters $`E_0`$ and $`m_1`$ (see Ref. ): the largest deviations from the DMRG results appear in the cases with minimal magnetization density, $`M_0=1/4a_0`$. However, even in the extreme quantum system $`(1,\frac{1}{2})`$ the discrepancies are small and the spin-wave expansion produces precise quantitative results. Concerning the dispersion of acoustical branch $`\omega _k^{(\alpha )}`$ (and the related spin-stiffness constant $`\rho _s`$), more numerical results are needed to make a statement about the accuracy of spin-wave description in this case. We believe, however, that at least for the systems with $`M_0>1/4a_0`$ the reported results for $`\rho _s`$ closely approximate the true spin-stiffness constants at zero temperature.
###### Acknowledgements.
This work was supported by the Deutsche Forschungsgemeinschaft (Grant No. 436BUL 113/106/0) and the Bulgarian Science Foundation (Grant No. F817/98).
## A Dyson-Maleev vertices for the ferrimagnetic model
Using the symmetric form adopted in Ref. , the normal-ordered Dyson-Maleev quasiparticle interaction $`V_{DM}`$, Eq. (14), reads
$`V_{DM}`$ $`=`$ $`{\displaystyle \frac{J}{2N}}{\displaystyle \underset{14}{}}\delta _{12}^{34}[V_{12;34}^{(1)}\alpha _1^{}\alpha _2^{}\alpha _3\alpha _4+2V_{12;34}^{(2)}\alpha _1^{}\beta _2\alpha _3\alpha _4+2V_{12;34}^{(3)}\alpha _1^{}\alpha _2^{}\beta _3^{}\alpha _4`$ (A1)
$`+`$ $`4V_{12;34}^{(4)}\alpha _1^{}\alpha _3\beta _4^{}\beta _2+2V_{12;34}^{(5)}\beta _4^{}\alpha _3\beta _2\beta _1+2V_{12;34}^{(6)}\beta _4^{}\beta _3^{}\alpha _2^{}\beta _1`$ (A2)
$`+`$ $`V_{12;34}^{(7)}\alpha _1^{}\alpha _2^{}\beta _3^{}\beta _4^{}+V_{12;34}^{(8)}\beta _1\beta _2\alpha _3\alpha _4+V_{12;34}^{(9)}\beta _4^{}\beta _3^{}\beta _2\beta _1].`$ (A3)
The explicit form of the symmetric vertex functions for the ferrimagnetic Heisenberg model (1) is
$$V_{12;34}^{(i)}=u_1u_2u_3u_4\overline{V}_{12;34}^{(i)},i=1,2,\mathrm{},9,$$
(A4)
where
$`\overline{V}_{12;34}^{(1)}=`$ $`+`$ $`\gamma _{13}x_1x_3+\gamma _{14}x_1x_4+\gamma _{23}x_2x_3+\gamma _{24}x_2x_4`$ (A5)
$``$ $`\sigma ^{1/2}(\gamma _{134}x_1x_3x_4+\gamma _{234}x_2x_3x_4)\sigma ^{1/2}(\gamma _1x_1+\gamma _2x_2),`$ (A6)
$`\overline{V}_{12;34}^{(2)}=`$ $``$ $`\gamma _{13}x_1x_2x_3\gamma _{14}x_1x_2x_4\gamma _{23}x_3\gamma _{24}x_4`$ (A7)
$`+`$ $`\sigma ^{1/2}(\gamma _{134}x_1x_2x_3x_4+\gamma _{234}x_3x_4)+\sigma ^{1/2}(\gamma _1x_1x_2+\gamma _2),`$ (A8)
$`\overline{V}_{12;34}^{(3)}=`$ $``$ $`\gamma _{13}x_1\gamma _{14}x_1x_3x_4\gamma _{23}x_2\gamma _{24}x_2x_3x_4`$ (A9)
$`+`$ $`\sigma ^{1/2}(\gamma _{134}x_1x_4+\gamma _{234}x_2x_4)+\sigma ^{1/2}(\gamma _1x_1x_3+\gamma _2x_2x_3),`$ (A10)
$`\overline{V}_{12;34}^{(4)}=`$ $`+`$ $`\gamma _{13}x_1x_2x_3x_4+\gamma _{14}x_1x_2+\gamma _{23}x_3x_4+\gamma _{24}`$ (A11)
$``$ $`\sigma ^{1/2}(\gamma _{134}x_1x_2x_3+\gamma _{234}x_3)\sigma ^{1/2}(\gamma _1x_1x_2x_4+\gamma _2x_4),`$ (A12)
$`\overline{V}_{12;34}^{(5)}=`$ $``$ $`\gamma _{13}x_2x_3x_4\gamma _{14}x_2\gamma _{23}x_1x_3x_4\gamma _{24}x_1`$ (A13)
$`+`$ $`\sigma ^{1/2}(\gamma _{134}x_2x_3+\gamma _{234}x_1x_3)+\sigma ^{1/2}(\gamma _1x_2x_4+\gamma _2x_1x_4),`$ (A14)
$`\overline{V}_{12;34}^{(6)}=`$ $``$ $`\gamma _{13}x_4\gamma _{14}x_3\gamma _{23}x_1x_2x_4\gamma _{24}x_1x_2x_3`$ (A15)
$`+`$ $`\sigma ^{1/2}(\gamma _{134}+\gamma _{234}x_1x_2)+\sigma ^{1/2}(\gamma _1x_3x_4+\gamma _2x_1x_2x_3x_4),`$ (A16)
$`\overline{V}_{12;34}^{(7)}=`$ $`+`$ $`\gamma _{13}x_1x_4+\gamma _{14}x_1x_3+\gamma _{23}x_2x_4+\gamma _{24}x_2x_3,`$ (A17)
$``$ $`\sigma ^{1/2}(\gamma _{134}x_1+\gamma _{234}x_2)\sigma ^{1/2}(\gamma _1x_1x_3x_4+\gamma _2x_2x_3x_4),`$ (A18)
$`\overline{V}_{12;34}^{(8)}=`$ $`+`$ $`\gamma _{13}x_2x_3+\gamma _{14}x_2x_4+\gamma _{23}x_1x_3+\gamma _{24}x_1x_4`$ (A19)
$``$ $`\sigma ^{1/2}(\gamma _{134}x_2x_3x_4+\gamma _{234}x_1x_3x_4)\sigma ^{1/2}(\gamma _1x_2+\gamma _2x_1),`$ (A20)
$`\overline{V}_{12;34}^{(9)}=`$ $`+`$ $`\gamma _{13}x_2x_4+\gamma _{14}x_2x_3+\gamma _{23}x_1x_4+\gamma _{24}x_1x_3`$ (A21)
$``$ $`\sigma ^{1/2}(\gamma _{134}x_2+\gamma _{234}x_1)\sigma ^{1/2}(\gamma _1x_2x_3x_4+\gamma _2x_1x_3x_4).`$ (A22) |
warning/0001/cond-mat0001035.html | ar5iv | text | # Distribution of the magnetic field and current density in superconducting films of finite thickness
## 1 Introduction
Of active interest recently is theoretical and experimental investigation of mixed static and dynamic states in superconducting films in a perpendicular magnetic field. Theoretical calculations of various magnetic characteristics of films in such a geometry, using the microscopic theory or the Ginzburg-Landau equations does not seem possible because of the mathematical intricacies involved in their solution.Therefore, in practical calculations the following two approaches are mostly employed. In the first method the Maxwell equation is analysed jointly with the London equation, which yields an integral-differential equation for distribution of the current density integrated over the film thickness $`d`$. This equation was derived on the assumption that the film is thin ($`d\lambda `$, $`\lambda `$ is the London penetration depth), which naturally limits its applicability range.
The other approach is based on the theory of complex variable functions (TCVF) , used in transformation of integral equations . In this case the phenomenological dependences $`𝐁(𝐇)`$ and $`𝐉(𝐄)`$ are employed as an additional condition instead of the London equation. It should be noted that neither of these methods has ever focussed on investigating the effects related to finiteness of the London penetration depth $`\lambda `$. Besides, the equality of the results obtained through solving of the integral equation for thin films and by the use of the TCVF methods for thick films leads us to believe that for films of arbitrary thickness there should exist one equation describing distribution of the current density and the magnetic field across the film width.
The present paper deals with a study on the distribution of the magnetic field and current density in the Meissner and mixed states for films placed in a perpendicular magnetic field. It is shown that for the finite-thickness films the Maxwell-London equation describes distribution of the vector potential (current density) averaged over the film thickness $`d`$, provided the latter, is much smaller than the film width $`W`$: $`d/W1`$. For thin films, $`d\lambda `$, this equation is valid practically over the entire width of the film, while in the case $`d\lambda `$ it holds everywhere in the film, except for the areas near the edges, which measure as $`W/2<|Y|W/2d/2`$.
The paper is organized as follows. Section 2 describes derivation of an one-dimensional equaton for distribution of the film-thickness-averaged vector potential across a sample width, based on the analysis of 2D Maxwell-London equation. In subsections 2.1 and 2.2 the 2D and 1D equations are numerically studied and compared for thin and thick films respectively. An approximation dependence for $`\overline{A}(y)`$ is obtained through numerical solution of these equations. The distribution of the vector potential (or the local current density) and of the local magneti field over a superconductor cross-section is described. Section 3 deals with estimation of the field for the first vortex entry into thin and thick superconducting films (subsection 3.1 and 3.2 respectivly). In subsection 3.3 we discuss the influence of surface defects and a layered structure of superconductors on a barrier suppression field value. Section 4 is concerned with the structure of a mixed state in thin and thick superconducting films of an arbitrary width in the absence of bulk pinning. An approximation formula is found for the integral (over thickness) current density, which is then used as a basis for constructing the magnetization curves for these superconductors. Section 5 sums up the main results obtained in this work.
## 2 The structure of the Meissner state
Assume an infinite (in the $`X`$ direction) superconducting film of width $`W`$ and thickness $`d`$ in a perpendicular magnetic field; the geometry is shown in Fig.1. Let us first consider the Meissner state. The Maxwell equation has the form (in a gauge $`𝐀=0`$)
$$\mathrm{\Delta }𝐀=\frac{4\pi }{c}𝐣,$$
$`(1)`$
where $`\mathrm{\Delta }`$ is the 2D Laplacian operator. It also follows from the symmetry of the problem that only the $`x`$ components of the vector potential $`𝐀=(A_x,0,0)`$ and of the current density $`𝐣=(j_x,0,0)`$ are not zero. The boundary conditions are $`\frac{A_x}{Y}|_{Y\pm \mathrm{}}=H_{\mathrm{}}`$, $`\frac{A_x}{Z}|_{Z\pm \mathrm{}}=0`$, where $`H_{\mathrm{}}`$ is the field far from film $`𝐇=(0,0,H_{\mathrm{}})`$. It should be noted that by the magnetic field $`H`$ in a film here we imply a microscopic field averaged over scales much larger than the atomic one but much smaller than $`\lambda `$.
In this case it is convenient to change over from the differential equation (1) to its integral analog. Using the Green function of the 2D Laplacian operator, we rewrite Eq.(1) as
$$A_x(Y,Z)=A_x^0(Y)\frac{2}{c}_{W/2}^{W/2}_{d/2}^{d/2}(\mathrm{ln}|𝐑𝐑^{}|+C)j_x(Y^{},Z^{})𝑑Y^{}𝑑Z^{},$$
$`(2)`$
where $`C`$ is the constant generic for the 2D Green function, $`A_x^0(Y)`$ is the vector potential of an external field: $`A_x^0(Y)=H_{\mathrm{}}Y`$ and the origin of coordinates is chosen in the film centre.
Employing London equation $`𝐣=𝐀/4\pi \lambda ^2`$ and introducing dimensionless coordinates $`y=2Y/W`$, $`z=2Z/d`$, Eq.(2) reads
$$\begin{array}{c}\text{ }A_x(y,z)=A_x^0(y)+\frac{Wd}{16\pi \lambda ^2}_1^1_1^1\mathrm{ln}\left((yy^{})^2+\left(\frac{d}{W}\right)^2(zz^{})^2\right)A_x(y^{},z^{})𝑑y^{}𝑑z^{}\text{ }\hfill \\ \text{ }+\frac{Wd}{16\pi \lambda ^2}\stackrel{~}{C}_1^1_1^1A_x(y^{},z^{})𝑑y^{}𝑑z^{},\text{(3)}\text{ }\hfill \end{array}$$
where $`\stackrel{~}{C}=C+2\mathrm{ln}(W/2)`$. The latter integral in (3) is directly proportional to the total current. In a magnetic field (without a transport current) the total current is equal to zero due to the symmetry, so the last term in the right-hand side of Eq. (3) vanishes. We now average Eq.(3) over the film thickness, which yields the following expression (hereafter by $`\overline{A}(y)`$ we imply 0.5$`_1^1A_x(y,z)𝑑z`$)
$$\begin{array}{c}\text{ }\overline{A}(y)=H_{\mathrm{}}yW/2+\frac{Wd}{4\pi \lambda ^2}_1^1\mathrm{ln}|yy^{}|\overline{A}(y^{})𝑑y^{}\text{ }\hfill \\ \text{ }+\frac{Wd}{32\pi \lambda ^2}_1^1_1^1_1^1\mathrm{ln}(1+b^2/(yy^{})^2)A_x(y^{},z^{})𝑑y^{}𝑑z𝑑z^{},\text{(4)}\text{ }\hfill \end{array}$$
where the integral kernel of Eq.(3) is written in the form
$$\mathrm{ln}\left((yy^{})^2+\left(\frac{d}{W}\right)^2(zz^{})^2\right)=2\mathrm{ln}|yy^{}|+\mathrm{ln}\left(1+\frac{b^2}{(yy^{})^2}\right)$$
and the designation $`b=(zz^{})d/W1`$ is introduced (as in this case of superconducting films $`d/W1`$).
The function $`\mathrm{ln}(1+b^2/(yy^{})^2)`$ is not zero only in a small region around point $`y^{}`$: $`|yy^{}||b|`$. In this case integration over $`y^{}`$ in the second integral of Eq.(4) can be done only over this small region. For the same reason we can expand the function $`A_x(y^{},z^{})`$ into the Taylor series in terms of $`y^{}`$ near point $`y`$:
$$A_x(y^{},z^{})=A_x(y,z^{})+\frac{A_x(y,z^{})}{y}(y^{}y)+.$$
$`(5)`$
Note that expansion (5) (in the above specified limit) is valid for thin ($`d<\lambda `$) films over the entire sample width. In thick ($`d>\lambda `$) films the validity of this expansion is violated in the near-edge regions with dimensions of order $`d/W`$. More details on the applicability of Eq.(5) are provided in the end of this Section.
After the series-expansion of function $`A_x(y^{},z^{})`$ (5) it is now possible to calculate the last term of Eq.(4)
$$_{yb}^{y+b}_1^1_1^1\mathrm{ln}\left(1+\left(\frac{d}{W}\right)^2(zz^{})^2/(yy^{})^2\right)\left(A_x(y,z^{})+\frac{A_x(y,z^{})}{y}(y^{}y)+\mathrm{}\right)𝑑y^{}𝑑z𝑑z^{},$$
$`(6)`$
(note, that the upper (lower) limit of the integration over $`y^{}`$ will change by $`1(1)`$, when point $`y`$ becomes close to the film edge, i.e., when $`1|b|||<1`$; this takes place because the integration in (4) is carried out only over a sample volume) and to show that (6) is equal to zero in a wide parameters range. Indeed, integration of Eq.(6), first in terms of $`y^{}`$ and then in $`z`$ and $`z^{}`$, provides a direct evidence (bearing in mind that function $`A_x(y^{},z^{})`$ is even in $`z^{}`$) that integral (6) is zero for all values of $`y`$ satisfying the inequality $`|y|1|b|`$. In the region $`1<|y|1|b|`$ integral (6) leads to appearance of nonzero terms that for thin films are small due to the presence of the corrections of $`(d/\lambda )^n`$ ($`n>1`$) type. They have to be taken into account when we are interested in, for example, the distribution of the derivative $`d\overline{A}/dy`$ near the film edges (since disregard for these terms will cause a logarithmic divergence of the first derivative). For thick films, allowance for the near-edge regions of a superconductor in integral (6) cannot largely affect the $`\overline{A}(y)`$ distribution off the film edges because of smallness of $`|b|`$.
Thus, the 2D equation (3) is reduced to a 1D equation for the film-thickness-averaged vector potential $`\overline{A}(y)`$, which is valid in the region $`|y|1|b|`$
$$\overline{A}(y)=H_{\mathrm{}}yW/2+\frac{Wd}{4\pi \lambda ^2}_1^1\mathrm{ln}|yy^{}|\overline{A}(y^{})𝑑y^{}$$
$`(7)`$
In the following sections the results of a numerical solution of Eqs. (7) and (3) are provided for thin and thick films.
### 2.1 Thin films ($`d<\lambda `$)
It turns out that in films satisfying the condition $`d/\lambda 1/4`$ the difference between the solutions of Eqs (7) and (3) (averaged over thickness) is about $`1\%`$ far from film edge and less than $`4\%`$ in a narrow near-edge region. An appreciable error in the near-edge region (which is slightly growing towards the film edge with a larger numerical step) depends on the presence of the small corrections in (7) that were ignored.
Besides, the vector potential in this case is practically independent of $`z`$ (but not the derivative $`A_x(y,z)/z`$). Thus, at $`d=\lambda `$ the relation $`A_x(y,1)/A_x(y,0)1.07`$, i.e., the difference is about $`7\%`$. With a lower $`d/\lambda `$ ratio the relation $`A_x(y,1)/A_x(y,0)`$ tends to unit.
We have derived an asymptotic expression for the vector potential distribution, satisfying Eq.(7) (and, hence, (3) averaged over $`z`$) with a sufficently high accuracy (not less than $`3\%`$ at a film edge and far from edge region, and not less than $`6\%`$ in the near-edge region, see Fig.2):
$$\overline{A}(y)=\frac{\lambda _{eff}H_{\mathrm{}}y}{\sqrt{\alpha (1y^2)+\beta }},$$
$`(8)`$
where $`\lambda _{eff.}=\lambda ^2/d`$, $`\beta 2\lambda _{eff}/\pi W+4(\lambda _{eff}/W)^2`$, and the dependence $`\alpha (W/\lambda _{eff})`$ is shown in Fig.3 together with the approximation (9)
$$\alpha 0.25\frac{0.63}{(W/\lambda _{eff})^{0.5}}+\frac{1.2}{(W/\lambda _{eff})^{0.8}}.$$
$`(9)`$
At $`W<\lambda _{eff}`$ the dependence $`\overline{A}(y)`$ becomes almost linear. Formula (8) with $`\alpha =0.25`$, $`\beta =0`$ was derived analytically in by solving Eq.(7) (to be more exact, a simplified version of Eq.(7) in which the left-hand part is omitted, which corresponds to the condition $`W\lambda _{eff}1`$).
The resultant dependence $`\overline{A}(y)`$ allows to calculate the field concentration parameter $`\gamma =\overline{H}_z^{edge}/H_{\mathrm{}}`$ (where $`\overline{H}_z^{edge}`$ is the edge field averaged over superconductor thickness) for films of such type:
$$\gamma =\frac{\overline{H}_z^{edge}}{H_{\mathrm{}}}=\frac{2\lambda _{eff}}{W\sqrt{\beta }}\left(\sqrt{1+\frac{\alpha }{\beta }}\right)$$
$`(10)`$
At $`W\lambda _{eff}`$ Eq.(10) transforms into
$$\gamma =\frac{\pi \sqrt{2\pi }}{10}\sqrt{\frac{W}{\lambda _{eff}}},$$
where the coefficient preceding $`\sqrt{W/\lambda _{eff}}`$ is a quantity of order unit. The difference between Eq.(10) and the numerically obtained expression for $`\gamma `$ may reach $`30\%`$: for wide films, ($`W\lambda _{eff}`$), Eq.(10) yields an overestimated result, see insert in Fig.4. This takes place because, unlike the $`\overline{A}(y)`$ function itself, the first derivative (8) with respect to $`y`$ (magnetic field) adequately satisfies the numerical solution everywhere except for the narrow region near a film edge (see Fig.4). Due to the logarithmic divergence of the magnetic field at a film edge, which follows from the solution of Eq.(7), the approximation expression for $`\overline{H}_z^{edge}`$ was compared with the numerical solution of Eq.(3). Fig.4 also shows the interpolation function $`1+0.66\sqrt{W/\lambda _{eff}}`$ for the numerical analysis data (see solid line in the insert). Thus, the difference between (10) and the numerical result in the wide film limit $`W/\lambda _{eff}1`$ comes to about $`17\%`$.
### 2.2 Strip of finite thickness ($`d\lambda `$)
A comparative numerical analysis of the solutions to Eqs (3) integrated over a superconductor thickness and (7) was also carried out for the case $`\lambda dW`$. It was found out that the solutions coincide (to the accuracy of about $`3\%`$) in the region $`|y|<1d/W`$ and ( quite surprisingly) in points $`|y|=1`$. In the near-edge region we observed a difference in the solutions of Eqs.(3) and (7).
An approximation equation has also been derived for the dependence $`\overline{A}(y)`$. It turned out to be exactly the same as the dependence (8) with the selection $`\alpha 0.25`$, $`\beta 0.64\lambda ^2/dW`$. One can easily see that the obtained values for $`\alpha `$ and $`\beta `$ practically coincide, to a numerical error, with those obtained for thin films at $`\lambda ^2/Wd1`$. Fig.5 shows the dependence $`\overline{A}(y)`$ derived from the solution to equation (3), and also Eq.(8). It is seen from the latter that maximum deviation of the solution to Eq.(3) from (8) (and, hence, from the solution to (7)) occurs in the region $`y>1d/W`$, but in point $`y=1`$, however, both solutions coincide again. Thus, expression (8) provides an adequate description of the $`\overline{A}(y)`$ distribution (the difference from the numerical solution of Eq.(3) does not exceed $`3\%`$) in the region $`|y|1d/W`$ and in points $`|y|=1`$. This confirms the above statement that Eq.(7) describes well the distribution of the $`\overline{A}(y)`$ dependence in a film depth. Besides, Eq.(7) is apparently valid immediately at the edge of a superconductor as well, which, in our opinion, is quite surprising fact.
Fig.6 shows the distribution of the current density over a film cross-section and the distribution of the absolute value of local magnetic field ($`|H|=\sqrt{H_z^2+H_y^2}`$) both inside and outside a film with dimensions $`W=100\lambda `$, $`d=10\lambda `$. As seen from Fig.6a,c, the magnetic field reaches its maximum at the corners (side edges) of a superconductor. At the equator $`(y=1,z=0)`$ the magnetic field is less intensive, but not appreciably smaller than the field at the corners (for comparison, at the corners $`|H|4.1H_{\mathrm{}}`$, while in the middle of a side face $`|H|2.9H_{\mathrm{}}`$ for the given parameters of film). It is easily seen that towards the film interior (along the $`y`$-axis) the magnetic field is practically uniform through the entire sample thickness, except for the near-surface areas with the thickness of order $`\lambda `$ where magnetic lines abruptly change direction. A similar behaviour is demonstrated by a current density (see figs.6 b,d). It is proved numerically that both the current density and the magnetic field fall off towards a sample centre by a law similar to the exponential one, not only from the side faces but also from the top and bottom ones. So, in thick films $`\lambda dW`$ screening currents flow only in the near-surface layers of thickness about $`\lambda `$. On the same scale there is a decrease of a local magnetic field in superconducting samples of such type.
The numerical solution of Eq.(3) also provides a possibility to determine the field at a film edge. Unfortunately, unlike with thin films, Eq.(8) differs largely from the numerical result for the near-edge region (this discrepancy may reach $`30\%`$). Therefore, the use of (8) in calculations of a thickness-averaged $`z`$-component of the magnetic field at a film edge certainly leads to a considerable error.
In Fig.7 the obtained numerical dependences $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ (the $`z`$-component of magnetic field, averaged over thickness) and $`H_z(1,0)/H_{\mathrm{}}`$ (the $`z`$-component of magnetic field on equator) on $`\sqrt{W/d}`$ are presented. It is clearly seen that with a good accuracy the dependences are linear even for the $`W/d`$ values which are close to unit. Note that the coefficient of proportionality between $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ $`\sqrt{W/d}`$ is equal to $`1.03`$, which practically is the same as its estimate ($`1`$) found in . Generally speaking, the value of this coefficient depends on a film thickness (or, rather, the ratio $`d/\lambda `$). Thus, the insert in Fig.7 illustrates relationships $`\overline{H}_z^{edge}/H_{\mathrm{}}`$, $`H_z(1,0)/H_{\mathrm{}}`$ for various values of film thickness and shape parameter $`W/d=5`$. It is seen that only the quantity $`H_z(1,0)`$ is practically independent of the ratio $`d/\lambda `$. The strongest dependence on film thickness is exhibited by $`H_z(1,1)`$ (it grows with the increase of the ratio $`d/\lambda `$). This results in a slight increase of the average field $`\overline{H}_z^{edge}`$ with a growth of film thickness (given the same $`W/d`$ ratio). However, for very thick films, $`d\lambda `$, $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ is supposed to practically cease to be dependent on $`d/\lambda `$. Indeed, in the limit of interest the equator field is independent of $`d/\lambda `$, while the corner field $`H_z(1,1)`$ increases as $`\sqrt{W/\lambda }`$ (which was derived from the expression (11c,d) given below)). The sharpest variation of the magnetic field intensity occurs at a distance of order $`\lambda `$ from the top/bottom surfaces. Correspondingly, the contribution of those regions in $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ will be of order $`\lambda \sqrt{W/\lambda }/d=\sqrt{W/d}\sqrt{\lambda /d}`$ and will become negligible with the increase of the ratio $`d/\lambda `$.
Besides the approximation expression for $`\overline{A}(y)`$, we have found numerical estimates for the vector potential in points ($`1,1`$) (on a corner); ($`1,0`$) (on the equator), and also the distribution of the vector potential (current density) over the upper/lower surfaces:
$$A_x(1,0)H\sqrt{\frac{W}{d}}\lambda ,$$
$`(11a)`$
$$A_x(1,1)A_x(W/2,0)\left(1+\frac{1}{\sqrt{16\pi }}\frac{d}{\lambda }\right)^{1/2},$$
$`(11b)`$
$$A_x(y,\pm 1)=\frac{\lambda _{eff}H_{\mathrm{}}y}{\sqrt{\alpha (1y^2)+\beta }},$$
$`(11c)`$
where
$$\alpha \frac{0.25}{(1+(d/\lambda )^2/2\pi )},\beta \frac{2\lambda _{eff}}{\pi W(1+(d/\lambda )/\sqrt{16\pi })}.$$
$`(11d)`$
It is easily seen that at $`d\lambda `$ the vector potential in points $`(\pm 1,\pm 1)`$ will largely exceed its value in points $`(\pm 1,0)`$. Using expression (11c) which is similar to (8) with renormalized parameters $`\alpha `$, $`\beta `$, we can find the points (lines) on the upper and lower surfaces, where the vector potential will coincide in absolute value with that on the equator. Simple calculations show that these lines are at a distance $`d/\pi `$ from the side surfaces of the strip having $`\lambda dW`$.
These results allow to confirm the assumption (see subsection 2.1) on the possibility of expanding $`A_x(y^{},z^{})`$ in the limits $`(yb,y+b)`$. Indeed, in the case of thin films the vector potential (or current density in a mixed state; see Sec. 4) varies on scales much larger than $`\lambda `$ and, hence, than $`d`$ (see (8)). For thick films, the vector potential (current density) far from edges is finite only in the surface layers of thickness of order $`\lambda `$. At the same time, the scale of variation for $`A_x(y,z)`$ along $`y`$ off sample edges is macroscopic (see expressions (11)). Therefore, expansion (6) is also possible off the edges. Near the edges, however, $`A_x(y,z)0`$ over entire thickness, and the scale of $`A(y,z)`$ variation (see fig.6b,d) in this region is $`\lambda `$ (in the $`y`$ direction). Hence, expansion (6) is not valid in the limits $`yby^{}y+b`$ near the edges ($`|y|1`$) of thick superconducting film.
## 3 The conditions for vortex entry in superconducting films
Using expressions (8,11) it is possible to estimate the edge barrier suppression field <sub>s</sub> or the first-vortex entry field into superconducting films.
### 3.1 Thin film
The vortices may enter into a thin film in the Meissner state provided the condition $`|\overline{A}(\pm 1)|=A_{crit}`$ is met; here $`A_{crit}\mathrm{\Phi }_0/2\pi \xi `$ ($`\mathrm{\Phi }_0`$ is the quantum of a magnetic flux, $`\xi `$ is the coherence length). The resultant expression for $`H_s`$ is
$$H_s\frac{\mathrm{\Phi }_0}{2\pi \xi \lambda _{eff}}\sqrt{\beta },$$
$`(12)`$
with $`\beta `$ being the same as in the expression (8). Dependence (12) in the limit $`W\lambda _{eff}`$ and $`W\lambda _{eff}`$ coincides to a factor of order one with the expression for the Meissner state breakdown field obtained in the limiting cases of wide and narrow thin films in . From (10,12) we can easily find the value of the magnetic field (or, rather, the thickness-averaged z-component) at a film edge $`\overline{H}_z^{edge}`$, when vortices start penetrating in it. For example, for wide films $`\overline{H}_z^{edge}`$ is
$$\overline{H}_z^{edge}\frac{\mathrm{\Phi }_0}{10\xi \lambda _{eff.}}.$$
To an accuracy of the factor of order one the above expression coincides with the field in the core of a Pearl vortex which is equal to $`\mathrm{\Phi }_0/4\pi \xi \lambda _{eff}`$ .
### 3.2 Thick film
The main difference between thick and thin films is that the vector potential for the former largely depends on $`z`$. However, we should apparently anticipate that the conditions of vortex entry here will be qualitatively similar to those for thin films. Indeed, as shown in , after the vector potential has reached its critical value at a superconductor edge in the Meissner state (in the mixed state it is the gauge-invariant potential $`𝚷=\mathrm{\Phi }_0\phi /2\pi 𝐀`$ that should reach a critical value), the order parameter $`\mathrm{\Psi }=|\mathrm{\Psi }|e^{i\phi }`$ is strongly suppressed and vortex formation begins. The above papers dealt with bulk superconductors and thin-film samples, which, due to the symmetry of the problem or problem geometry, could be assumed homogeneous along the $`z`$-axis.
It should be expected that in thick films the order parameter will be suppressed in the regions where the vector potential $`A_x(y,z)`$ reaches its critical value.
First, the condition $`|A_x(y,z)|=A_{crit}`$ will be satisfied at the side edges $`(\pm 1,\pm 1)`$ of a superconductor (as the magnetic field is increasing from zero). It means that the order parameter will be suppressed in these points first. With a further increase of the magnetic field this situation may develop by two scenarios:
1. Suppression of the order parameter results in tilted vortices that start to form at the corners of a superconductor cross-section. When the vector potential at the equator reaches the critical value, two tilted vortices fragments (from top and from bottom) will join each other to form one rectilinear vortex. In the absence of pinning the latter is able to penetrate into the sample centre driven by the Lorentz force. Similar vortex entry scheme was discussed in .
2. In the course of further magnetic field increase, the order parameter becomes suppressed in the region of side edges. This causes areas with a suppressed order parameter to appear near the side edges, which would allow to regard a film cross-section as a rectangular with rounded-off edges. It should be emphasized that the geometrical sizes of sample remain unchanged in this situation, only its physical properties vary in the regions near side edges. This scenario allows to explain the physical mechanism behind the formation of the ”geometrical rounds-off” near the corners of a rectangular cross-section sample, which were considered in . However, unlike , our approach is based on the assumption that vortices will start entering deep inside a superconductor when the condition $`|A_x(y,z)|=A_{crit}`$ is fulfilled at the equator. By that moment the effective ”round-off” radius will reach a value of order $`d/2`$.
Which of these two scenarios is practically feasible can be found out only through experiment. Theoretically this question can be answered by numerical solution of a problem on a vortex entry in a 3D sample within the nonstationary theory of Ginzburg-Landau.
The feature common for the above two schemes is, actually, the assumption that vortices enter deep a thick film when the condition $`|A_x(1,0)|=A_{crit}`$ is met. This allows to estimate the field of vortex entry inside a sample. Using Eq.(11a) (regardless of possible variation due to the penetration of tilted vortices or the existence of areas with a suppressed order parameter), we now derive the expression for field $`H_s`$, which is equivalent (12) (with $`\beta =2\lambda _{eff}/\pi W`$). This similarity is due to the fact that the $`A_x(\pm 1,0)`$ value is defined practically by the same expression for both thin and thick films, provided parameter $`\lambda ^2/dW`$ is the same.
By analogy with thin films, one can find a value of the field at a film edge when the vortices start to penetrate deep into a sample. Yet, as opposed to thin films, field $`H_z^{edge}`$ largely depends on the $`z`$ coordinate in the sample region. Estimations of the equator field yields
$$H_z^{eq}=H_z(1,0)\frac{\mathrm{\Phi }_0}{2\pi \xi \lambda },$$
which is practically equivalent to the $`\overline{H}_z^{edge}`$ (see fig.7). Note also that $`H_z^{eq}(H_s)`$ to the factor of order unit is equal to the thermodynamical field $`H_c`$, or the surface barrier suppression field for bulk superconductors in the absence of edge defects.
### 3.3 The influence of surface defects and anisotropy
The resultant values for field $`H_s`$ are characteristics of isotropic superconductors with ideal surface. As was established in , surface defects can considerably decrease the value of $`H_s`$. For example, in for the case $`\kappa 1`$ ($`\kappa `$ is the Ginzburg-Landau parameter) maximum suppression of the entry field $`g=H_s/H_{en}`$ (where $`H_{en}`$ is the field of vortices entry in a superconductor with surface defects) was estimated as $`g\sqrt{\kappa /\pi }`$. Thus, for $`\kappa =100`$ we will have $`g5.6`$.
A strong influence on the value of $`H_s`$ may be produced also by anisotropy or, rather, layered structure of superconductors. If the layers are not Josephson-coupled (or are weakly coupled), a superconductor should be regarded as a stack of superconducting layers. This geometry can be simulated, if we multiply the integrand in equation (3) (and, hence, in integral (6)) by the step periodic function $`z`$ which is equal to one in the superconducting layer and is zero in the interlayer space.
Let a period in such a structure be much smaller than $`\lambda `$ (which is practically always fulfilled for $`HTSC`$), a layer thickness be $`l`$ and an interlayer separation be $`m`$. Then we can assume the distribution of $`A_x(z)`$ to be a smooth function $`z`$, similar to the dependence $`A_x(z)`$ for a homogeneous superconductor. In this case, the integral in Eq.(3) for a layered superconductor will be $`(l+m)/l`$ times smaller than that for an isotropic superconductor. In other words, we can replace this integral for a layered superconductor by the integral for an isotropic case by introducing an effective penetration depth $`\lambda ^{}=\lambda \sqrt{(l+m)/l}`$. In this way we immediately obtain the distributions of $`\overline{A}(y)`$, $`A_x(y,1)`$, and also the values for $`A_x(y,z)`$ at the side edges and the equator with allowance for the layered structure of superconductor.
One particular effect of the anisotropy is that the value of $`A_x(1,0)`$ will be $`\sqrt{(l+m)/l}`$ times larger than that for an isotropic superconductor, all other conditions being equal. This, in its turn, will lead to a $`\sqrt{(l+m)/l}`$ times smaller field of vortex entry in a superconductor. For example, at $`l=3\dot{A}`$ and $`m=12\dot{A}`$, typical for $`BiBaCuO`$, one finds $`\sqrt{(l+m)/l}2.2`$.
Thus, the two above-mentioned factors, i.e., surface defects and layered structure of superconductors may cause a considerable (10-fold and more) decrease of the vortex entry field in layered superconductors with surface defects, as compared to the vortex entry field in isotropic perfect-surface superconducting films.
Another conclusion following from the fact of a layered structure in such systems is that the scale of a local magnetic field penetration, for example, for thick films, will be $`\lambda ^{}=\lambda \sqrt{(l+m)/l}`$, and for thin films the parameter $`\lambda _{eff}`$ has to be changed for $`\lambda _{eff}^{}=\lambda ^{}{}_{}{}^{2}/d`$. At the same time, the thickness-averaged current density, unlike the thickness-averaged vector potential, will practically remain unchanged across the entire film width, except for the regions lying close to sample edges. This can be accounted for by the fact that the expression for current density $`j(y)=cA(y)/4\pi \lambda ^2`$ will include $`\lambda `$ only at a superconductor edge (see (8)). Likewise, all other quantities that obviously do not include $`\lambda `$ (for example, a degree of the magnetic field concentration at the equator of thick film) will remain the same.
## 4 The structure of a mixed state
Let us now discuss the parameters of a mixed state arising in a superconducting film in fields larger than $`H_s`$. Here we neglect the presence of bulk pinning, which is justified for soft superconductors. This problem was studied earlier in . In the authors considered a case of narrow thin films $`Wd/\lambda ^21`$, deals with wide thin films $`Wd/\lambda ^21`$, and is a study on thick films that formally obey the condition $`Wd/\lambda ^21`$. We will analyse a general case to show that it embraces either of the above two situations. Besides, the resultant distribution of current density will be finite across the entire film width, as opposed to the results of the above cited works.
Consider a superconducting film in a mixed state, placed in a perpendicular magnetic field. The current density and the vector potential in the London limit are related as $`𝐣=(𝐀\mathrm{\Phi }_0\phi /2\pi )c/4\pi \lambda ^2`$, where $`\phi `$ is the order parameter phase. The Maxwell equation will have the form (1), in which $`\mathrm{\Delta }`$ is now a 3D Laplacian operator. Using the Green function for $`\mathrm{\Delta }`$, we can write this equation in an integral form:
$$𝐀(𝐫)=𝐀^0(𝐫)+\frac{1}{c}\frac{𝐣(𝐫^{})}{|𝐫𝐫^{}|}𝑑x^{}𝑑y^{}𝑑z^{}.$$
$`(13)`$
We now subtract function $`\phi (𝐫)`$ from the left- and right-hand parts of Eq.(13) and take curl from these parts (using the property $`\times \phi (𝐫)=2\pi \delta (𝐫𝐫^{})`$, where $`𝐫^{}=(x^{},y^{})`$ are the vortex coordinates). Next, we do the averaging over coordinates $`x,y`$ on scales much larger than an intervortex distance. After these operations, the distribution of a current density becomes uniform along the $`x`$-axis and we can perform integration over $`x^{}`$ in (13). Then we average the obtained equation over a film thickness and use the results of the integral (6) calculations. This will yields an equation for the sheet current density $`i(y)=_{d/2}^{d/2}j_x(y,z)𝑑z`$,
$$\frac{8\pi \lambda _{eff}}{cW}\frac{di(y)}{dy}+\frac{2}{c}_1^1\frac{i(y^{})}{yy^{}}𝑑y^{}=H_{\mathrm{}}+n(y)\mathrm{\Phi }_0,$$
$`(14)`$
where $`n(y)`$ is the density of vortices, the distance being measured in units of $`W/2`$. For the first time this equation was derived in for thin films $`d\lambda `$. Just like Eq.(7), (14) is valid at $`|y|1|b|`$ for thick films, and across the width for thin films (excluding an extremely narrow near-edge region of size $`d\lambda `$). Besides, it should be expected by analogy with the Meissner state that Eq.(14) for thick films will also be valid directly at a sample edge.
We analysed Eq.(14) numerically for different values of parameter $`W/\lambda _{eff}`$, using the condition that current density is zero in the region where vortices exist, and takes finite value in vortex-free regions . Besides, we set a boundary condition $`j(\pm 1)=\pm j_s`$ on a current density (in increasing magnetic field), which allows for an edge barrier. The value of the current density $`j_s`$ of order of the Ginzburg-Landau depairing current density for ideal-surface superconductors . In result, we have obtained the approximation expression for $`i(y)`$
$$i(y)=\{\begin{array}{cc}0\hfill & 0<|y|<a,\hfill \\ \frac{cH_{\mathrm{}}(z+1)}{4\pi \sqrt{\alpha (1z^2)+\beta \frac{(|y|+a)^2}{(1+a)^2}}}\mathrm{sign}(y)\hfill & a<|y|<1,\hfill \end{array}$$
$`(15)`$
where
$$z=\frac{2(y^2a^2)}{(1a^2)}1,$$
$$\beta \frac{8}{\pi }\frac{1}{1a^2}\frac{\lambda _{eff}}{W}+\frac{16}{(1a)^2}\left(\frac{\lambda _{eff}}{W}\right)^2,$$
$`a(H_{\mathrm{}})`$ is the half-width of the vortex-filled region and parameter $`\alpha `$ is defined by expression (9) in which W has been replaced by $`W(1a)`$.
Expressions (15), being not derived, represents a rather useful interpolation for the distribution of the sheet current density for a film in a mixed state. We would like to emphasize again that in the thin-film case $`W/\lambda _{eff}`$ can be both smaller (narrow films) and larger (wide films) than unit, whereas for thick films this ratio is always much larger than 1.
Fig.8 shows the dependence $`i(y)`$ for different values of a magnetic field. The difference of approximation (15) from the numerical result does not exceed $`4\%`$ in the vortex-free zone ($`a<y<1`$). Note, that in the near-edge region and close to the boundary of the vortex-filled region deviation may come to about $`9\%`$.
The dependence $`a(H_{\mathrm{}})`$ (in increasing magnetic field) is to be found from the following expression:
$$\frac{8}{\pi }\frac{1}{1a^2}\frac{\lambda _{eff}}{W}+\frac{16}{(1a)^2}\left(\frac{\lambda _{eff}}{W}\right)^2=\left(\frac{H_{\mathrm{}}}{H_s}\right)^2\left(\frac{8}{\pi }\frac{\lambda _{eff}}{W}+16\left(\frac{\lambda _{eff}}{W}\right)^2\right).$$
For $`W/\lambda _{eff}1`$ we have
$$a(H_{\mathrm{}})=\sqrt{1(H_s/H_{\mathrm{}})^2},H_{\mathrm{}}H_s,$$
or
$$a(H_{\mathrm{}})=1\sqrt{\frac{2\lambda _{eff}}{\pi W}}\frac{H_s}{H_{\mathrm{}}},H_{\mathrm{}}H_s,$$
while for $`W/\lambda _{eff}1`$ we have
$$a(H_{\mathrm{}})=1H_s/H_{\mathrm{}},$$
for all values of $`H_{\mathrm{}}`$.
Using dependence (15), we can find distribution of the $`z`$-averaged magnetic field across a film width:
$$\overline{H}_z(y)=H_{\mathrm{}}+\frac{2}{c}_1^1\frac{i(y^{})}{yy^{}}𝑑y^{}.$$
$`(16)`$
Fig. 9 shows the dependences $`\overline{H}_z(y)`$ for a film with $`W/\lambda _{eff}=200`$ and $`a=0.6`$ ($`H_{\mathrm{}}1.3H_s`$), obtained numerically and by means of expression (15,16). It is seen that these dependences practically coincide across the entire width of a film, except for the near-edge regions and the boundary of the vortex-filled zone. The difference in the magnetic field value at $`y=0.6`$ should be attributed to the inaccuracy of approximation (more precisely, its first derivative at the boundary of the vortex-filled area). Dependence $`\overline{H}_z(y)`$ shown in the insert to Fig. 9 was obtained theoretically in . It is seen that, as opposed to this analytical dependence, a non-zero magnetic field does exist outside the vortex-filled region also, and it is quite strong ($`>0.1H_{\mathrm{}}`$) for a film with the given parameters. Another distinguishing feature is the occurrence of a jump from zero to some finite value for the dependence $`n(y)=\overline{H}_z(y)/\mathrm{\Phi }_0`$ at $`y=a`$. The reason of the vortex density discontinuity is explained by a non-zero magnetic field in the region $`(a,1)`$ and the condition of a magnetic field continuity at $`y=a`$.
Using Eq.(15), we can estimate the dependence of $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ on the parameters of a film and an increasing external magnetic field (for thin films; see subsection 2.2)
$$\frac{\overline{H}_z^{edge}}{H_{\mathrm{}}}=\frac{2\lambda _{eff}}{W}\frac{1}{\sqrt{\beta }}\left(1+\frac{1}{\beta }\left(2\alpha \frac{\beta (1a)}{2\sqrt{1+a^2/2}}\right)\right)\frac{4}{1a^2}.$$
$`(17)`$
It is nicely seen that in the limit $`a1`$($`H_{\mathrm{}}\mathrm{}`$) the ratio $`\overline{H}_z^{edge}/H_{\mathrm{}}`$ tends to 1.
Knowing the dependence $`i(y)`$, we can find the magnetization curves of superconducting films for different values of $`W/\lambda _{eff}`$. Fig.10 illustrates the obtained results. One can see that with a increasing parameter $`W/\lambda _{eff}`$ the magnetization curves become similar to the dependence $`M(H)`$ derived theoretically in for wide thin films. As $`W/\lambda _{eff}`$ decreases, the magnetization curves tend to the dependence which is valid for thin narrow films . Thus, even at $`W/\lambda _{eff}=1`$ the dependence $`M(H)`$ for a narrow film and the $`M(H)`$ obtained numerically by the use of Eq.(15) practically coincide. So, expression (15) allows to obtain magnetization curves for such superconductors at arbitrary ratio $`W/\lambda _{eff}`$. Note that the magnetization curves in this case, i.e., at any value of parameter $`W/\lambda _{eff}`$ lie between two curves - 1 and 3 as shown in Fig. 10 (in dimensionless units).
## 5 Conclusion
It is shown in the present paper that the Maxwell-London equation used so far only for thin films is also valid for samples of finite thickness. This equation is shown to define the distribution of a thickness-averaged vector potential and/or current density (in the mixed state case) across a sample width. For thin films the equation holds practically everywhere in a film, whereas in the thick film case its applicability is restricted only within a narrow bands near the edges $`W/2d/2|Y|W/2`$.
An approximation expression is found, describing distribution of vector potential $`\overline{A}`$ (or current density $`\overline{j}(y)`$) across the width of a film in the Meissner state, which applies to both thin and thick films. For thick films analytical expressions have been derived, defining the value of the vector potential (local current density) at the equator ($`Y=1,Z=0`$), side edges $`(Y=1,Z=1)`$, and also on the top and bottom surfaces ($`Y,Z=\pm d/2`$) of a sample. Besides, analytical approximation expressions have been found for the magnetic field at the equator and for the thickness-averaged edge magnetic field.
The vector potential distribution data were used to evaluate the field of the the first vortex entry (barrier suupresion field) for superconductors of such geometry. It’s described by an universal expression (12) valid for both thin and thick films. It is shown that besides surface defects the layered structure of superconductor may result in a significant (up to $`100\%`$) suppression of the vortex entry field. Thus, mutual influence both surface defects and layered structure may lead to suppression of $`H_s`$ by factor ten (and even greater).
The study of a mixed state yields an interpolation expression for distribution of a sheet current density $`i(y)`$ in superconducting films without bulk pinning. This result allows to for the first time estimate the dependence on the magnetic field concentration $`\gamma =\overline{H}_z^{edge}/H_{\mathrm{}}`$ on the parameters of superconductor and external magnetic field $`H_{\mathrm{}}`$. Besides, these data were used to calculate the magnetization curves for film superconductors at any values of parameter $`Wd/\lambda ^2`$.
## 6 Acknowledgements
Authors are obliged to Prof. J.R.Clem, Dr. G.M. Maksimova for helpful discussions of the results obtained. This work is supported by the Science Ministry of Russia (Project No. 98-012), and in part Basic Foundation for Fundamental Research (Project No. 97-02-17437). Partial support of the International Center for Advanced Studies (INCAS) through Grant No. 99-2-03 is gratefully acknowledged.
Figure captions
fig.1
Geometry of the problem.
fig.2
Distribution of averaged vector potential, for different ratios $`W/\lambda _{eff}`$: curves 1-5 for $`W/\lambda _{eff}=1,5,10,50,200`$, respectively. Dotted lines are numerical results, solid lines are approximation (8).
fig.3
Dependence of the parameter $`\alpha `$ on $`W/\lambda _{eff}`$. Circles are numerical results, solid curve 1 is the approximation (9).
fig.4
Distribution of $`\overline{H}_z`$ inside the film for $`W/\lambda _{eff}=200`$. Dots are numerical result, solid line is expression obtained from approximation (8). Insert shows dependence $`\overline{H}_z^{edge}(W/\lambda _{eff})`$: circles are numerical result, dotted line is the expression (10), solid line is the fitting function $`1+0.66\sqrt{W/\lambda _{eff}}`$.
fig.5
Distribution of averaged vector potential at thick film ($`d=10\lambda `$) for various widths: $`W=50\lambda `$ (1), $`W=100\lambda `$ (2), $`W=200\lambda `$ (3). Dots are numerical result, solid lines are expression (8) with $`\alpha =0.25`$, $`\beta =2\lambda ^2/\pi dW`$.
fig.6
Contour lines of the intensity of magnetic field (a,c) and current density (b,d) of thick ($`W=100\lambda `$, $`d=10\lambda `$) film in applied perpendicular magnetic field. The step for magnetic field is $`0.41H_{\mathrm{}}`$, for current density is $`0.1j(1,1)`$. Maximum values of magnetic field ($`H=4.1H_{\mathrm{}}`$) and current density ($`j=1`$) are reached at the corners of the film.
fig.7
Dependences $`\overline{H}_z^{edge}`$ (circles) and $`H_z(1,0)`$ (dots) on the parameter $`\sqrt{W/d}`$. Solid line 1 is the fitting function $`1/3+1.03\sqrt{W/d}`$, dotted line 2 is the fitting function $`1/3+0.92\sqrt{W/d}`$. Insert shows the dependences of $`\overline{H}_z^{edge}`$(circles), $`H_z(1,0)`$(dots) and $`H_z(1,1)`$(stars) are shown on the film thickness for $`W/d=5`$.
fig.8
Distribution of the sheet current $`i(y)`$ for film with parameter $`W/\lambda _{eff}=200`$ and for different values of $`a`$: 0.0 (1), 0.4 (2), 0.8 (3). Dotted lines are numerical results, solid lines are expression (15).
fig.9
Distribution of the averaged $`z`$-component of magnetic field for $`W/\lambda _{eff}=200`$ and $`a=0.6`$. Solid line is obtained with help of expression (15,16), dotted line is numerical result. Insert shows detailed distribution of the field; dashed line is the function $`H\sqrt{a^2y^2}/\sqrt{1y^2}`$ from .
fig.10
Magnetization curves of superconducting films for different ratio $`W/\lambda _{eff}`$: curve 1 for $`W/\lambda _{eff}=\mathrm{}`$, curve 2 for $`W/\lambda _{eff}=200`$, curve 3 for $`W/\lambda _{eff}=1`$. Curve 3 practically coincides with the magnetization curve for narrow $`W\lambda _{eff}`$ films . |
warning/0001/nucl-th0001011.html | ar5iv | text | # EQUATION OF STATE OF NUCLEONIC MATTER11footnote 1Supported by BMBF and GSI Darmstadt
## 1 Introduction
Relativistic heavy-ion collisions (RHIC) provide a unique tool to study nuclear matter at high densities and temperatures, reminiscent of the early big-bang of the universe, but with better statistics and under controlled conditions. These reactions also provide constraints on the interior of neutron stars, where the nuclear equation-of-state (EoS) plays an essential role for the possible existence of an inner quark core or an extended mixed phase of quarks and hadrons . However, since in a RHIC the system initially is far away from thermal and chemical equilibrium, both particle production and collective motion depend on various quantities such as the stiffness of the EoS, the momentum dependence of the interaction or mean-field potentials (MDI), in-medium modifications of the $`NN`$ cross section $`\sigma _{NN}`$, the initial momentum distribution of the nucleons as well as the number of hadronic degress of freedom accounted for in the transport simulation . It is thus necessary not to focus on a single observable alone but to investigate the dynamical evolution of the RHIC within a single model that is able to describe all relevant single-particle as well as collective quantities.
Whereas the experimental and theoretical studies of collective nuclear flow have been restricted to the 1-2 A GeV energy regime in the past , more recently both the directed transverse flow (sideward flow) and the flow tensor (elliptic flow) have been measured and reported by the BNL-E877 collaboration for heavy-ion ($`Au+Au`$) collisions at AGS energies in the energy range of 1 A GeV $`E_{inc}11A`$ GeV. In this energy range the directed transverse flow first grows, saturates at around 2 A GeV, and then decreases experimentally with energy showing no minimum as expected from hydrodynamical calculations including a first order phase transition in the EoS . Whether this decrease in directed flow or the change of sign in elliptic flow is indicative of a phase transition is a question of high current interest.
In this contribution the collective behaviour of nuclear matter in a heavy-ion collision is reviewed in the energy range from 150 A MeV to 11 A GeV for various systems using the transport model . For energies above 1 A GeV it has been, furthermore, complemented by the string dynamics from the HSD transport approach which has been tested extensively for $`p+A`$ and $`A+A`$ collisions from SIS to SPS energies .
## 2 The extended RBUU-Model
To describe the heavy-ion collision data at energies starting from the SIS at GSI to the SPS regime at CERN, relativistic transport models have been extensively used . For a general derivation of transport theories the reader is refered to Ref. and to Ref. for a recent review. Among these transport models the Relativistic Boltzmann-Uehling-Uhlenbeck (RBUU) approach incorporates the relativistic mean-field (RMF) theory, which is applicable also to various nuclear structure problems as well as for neutron star studies . Since it is based on an effective hadronic Lagrangian density it allows to evaluate directly the nuclear EoS at zero and finite nuclear temperature $`T`$ as well as the scalar and vector mean fields $`U_S`$ and $`U_\mu `$, that determine the in-medium particle properties, for arbitrary configurations of nucleons in phase space . Here we essentially base the studies on the Lagrangian (parameter set NL3) from Ref. since this Lagrangian has been applied widely in the analysis of heavy-ion collisions by various groups .
We recall that the most simple versions of RMF theories assume the scalar and vector fields to be represented by point-like meson-baryon couplings. These couplings lead to a linearly growing Schrödinger-equivalent potential in nuclear matter as a function of the kinetic energy $`E_{kin}`$, which naturally explains the energy dependence of the nucleon optical potential at low energies ($`200`$ MeV). However, a simple RMF does not describe the nucleon optical potential at higher energies, where it deviates substantially from a linear function and saturates at $`E_{kin}`$ 1 GeV . Since the energy dependence of sideward flow is controlled in part by the nucleon optical potential, the simple RMF cannot be applied to high-energy heavy-ion collisions. In order to remedy this aspect, the more sophisticated RBUU approaches invoke an explicit momentum dependence of the coupling constant, i.e. a form factor for the meson-baryon couplings .
Further important ingredients at AGS energies are the resonance/string degrees of freedom which are excited during the reaction in high energy baryon-baryon or meson-baryon collisions. While at SIS energies particle production mainly occurs through baryon resonance production and their decay, the string phenomenology is found to work well at SPS energies ($``$ 200 A GeV) . One of the characteristic features of the AGS energy regime is the competition between these two particle production mechanisms which might be separated by some energy scale $`\sqrt{s_{sw}}`$ . Due to this complexity there are various ways to implement elementary cross sections in transport models . One of the extremes is to parameterize all possible cross sections directly for multi-pion production, $`NNNNn\pi (n3)`$ only through $`N,\mathrm{\Delta },\pi `$ degrees of freedom; the other extreme is to fully apply string phenomenology in this energy region without employing any resonances. Although it is possible to reproduce the elementary cross sections from $`NN`$ and $`\pi N`$ collisions and the inclusive final hadron spectra in heavy-ion collisions within these different models, we expect that differences should appear in the dynamical evolution of the system, e.g. in the thermodynamical properties and in collective flow . For example, if thermal equilibrium is achieved at a given energy density, models with a larger number of degrees of freedom including strings will give smaller temperature and pressure .
We have employed here the combination of a resonance production model and the Lund string model as incorporated in the Hadron-String-Dynamics (HSD) approach . In the practical implementation for $`NN`$ or $`MN`$ collisions at invariant energies lower (higher) than a threshold energy $`\sqrt{s_{sw}}`$ resonances (strings) are assumed to be excited (see below).
### 2.1 The optical potential
In this presentation the scalar and vector mean fields $`U_S`$ and $`U_\mu `$ are calculated on the basis of the same Lagrangian density as considered in Ref. , which contains nucleon, $`\sigma `$ and $`\omega `$ meson fields and nonlinear self-interactions of the scalar field (cf. NL3 parameter set ). The scalar and vector form factors at the vertices are taken into account in the form
$$f_s(p)=\frac{\mathrm{\Lambda }_s^2\frac{1}{2}p^2}{\mathrm{\Lambda }_s^2+p^2}\text{and}f_v(p)=\frac{\mathrm{\Lambda }_v^2\frac{1}{6}p^2}{\mathrm{\Lambda }_v^2+p^2},$$
(1)
where the cut-off parameters $`\mathrm{\Lambda }_s=1.0`$ GeV and $`\mathrm{\Lambda }_v=0.9`$ GeV are obtained by fitting the Schrödinger equivalent potential,
$$U_{sep}(E_{kin})=U_s+U_0+\frac{1}{2M}(U_s^2U_0^2)+\frac{U_0}{M}E_{kin},$$
(2)
to Dirac phenomenology for intermediate energy proton-nucleus scattering .
Fig. 1 The Schrödinger equivalent potential (2) at density $`\rho _0`$ as a function of the nucleon kinetic energy $`E_{kin}`$. The solid curve (RBUU) results from the momentum-dependent potentials discussed in the text. The data points are from Hama et al. .
The resulting Schrödinger equivalent potential (2) is shown in Fig. 1 as a function of the nucleon kinetic energy with respect to the nuclear matter at rest in comparison to the data from Hama et al. (open squares). The experimental increase of the Schrödinger equivalent potential up to $`E_{kin}=1`$ GeV is decribed quite well; then the potential decreases and is set to zero above 3.5 GeV.
For the transition rate in the collision term of the transport model we employ in-medium cross sections as in Ref. that are parameterized in line with the corresponding experimental data for $`\sqrt{s}\sqrt{s_{sw}}`$. For higher invariant collision energies we adopt the Lund string formation and fragmentation model as incorporated in the HSD transport approach which has been used extensively for the description of particle production in nucleus-nucleus collisions from SIS to SPS energies . In the present relativistic transport approach (RBUU) as in Ref. we explicitely propagate nucleons and $`\mathrm{\Delta }`$’s as well as all baryon resonances up to a mass of 2 GeV with their isospin degrees of freedom . Furthermore, $`\pi ,\eta `$, $`\rho `$, $`\omega `$, $`K,\overline{K}`$ and $`\sigma `$ mesons are propagated, too, where the $`\sigma `$ is a short lived effective resonance that describes s-wave $`\pi \pi `$ scattering. For more details we refer the reader to Refs. concerning the low energy cross sections and to Refs. with respect to the implementation of the string dynamics.
### 2.2 Transverse mass spectra of protons
In Fig. 2 we show the dependence of the calculated proton transverse mass spectra in a central collision of Au + Au at 11.6 $`A`$ GeV for b $`<3.5`$ fm for $`\sqrt{s_{sw}}`$ = 2.6 GeV (dotted histograms) and 3.5 GeV (solid histograms) in comparison to the experimental data of the E802 collaboration . A cascade calculation (crosses) is shown additionally for $`\sqrt{s_{sw}}`$ = 3.5 GeV to demonstrate the effect of the mean-field potentials which lead to a reduction of the transverse mass spectra below 0.3 GeV. As expected, the transverse mass spectrum is softer for smaller $`\sqrt{s_{sw}}`$ due to the larger number of degrees of freedom in the string model relative to the resonance model.
Fig. 2: The transverse mass spectra of protons for $`Au+Au`$ collisions at $`b<3.5`$ fm. The solid line and the dot- dashed line with crosses are results for $`\sqrt{s_{sw}}`$=3.5 GeV with and without nuclear potentials, respectively. The dotted line RBUU(2.6) is for $`\sqrt{s_{sw}}`$=2.6 GeV. The data points are taken from the E802 collaboration .
We note that strings may be regarded as hadronic excitations in the continuum of lifetime $`t_F`$ 0.8 fm/c (in their rest frame) that take over a significant part of the incident collision energy by their invariant mass. They decay dominantly to light baryons and mesons and only to a low extent to heavy baryon resonances. Thus the number of particles for fixed system time is larger for string excitations than for the resonance model where several hadrons propagate as a single heavy resonance which might be regarded as a cluster of a nucleon + $`n`$ pions. As a consequence the translational energies are suppressed in string excitations and, as a result, the temperature as well as the pressure are smaller when exciting strings.
From the above comparison with the experimental transverse mass spectra for protons we find $`\sqrt{s_{sw}}`$ 3.5 GeV, which implies that binary final baryon channels should dominate up to $`\sqrt{s}`$ 3.5 GeV which corresponds to a proton laboratory energy of about 4.6 GeV for $`pp`$ collisions.
## 3 SIS energies
### 3.1 Transverse flow
Within the RBUU model described above we now calculate the transverse flow for various systems and beam energies and analyse the dependence on different quantities. The flow $`F`$ is defined as the slope of the transverse momentum distribution at midrapidity,
$$F=\frac{dp_x}{dy}_{|y=y_0},$$
(3)
which is essentially generated by the participating matter in the ’fireball’ . The latter finding explains why flow does not clearly distinguish between an EoS with and without momentum-dependent forces. Since the fireball contains the stopped matter, the relative momenta in the fireball (besides the unordered thermal motion) are small. Only when applying additional cuts, e.g. on high transverse momenta , i.e. by selecting particles escaping early from the fireball, or selecting mainly participant or spectator particles by appropriate $`\mathrm{\Theta }_{cm}`$-Cuts , a difference between the momentum-dependent and momentum-independent EoS can be established.
The FOPI data on proton flow indicate a decrease of sidewards flow above 1 A GeV incident energy following the well known logarithmic increase at low energies. Using standard potential parameterizations, both nonrelativistic and relativistic , this behavior cannot be understood within conventional transport models. In the latter the optical potential stays constant or even increases at high momenta and therefore the repulsion generated from the momentum-dependent forces in a HIC gives rise to a significant contribution to the flow signal. However, since the nucleon-nucleus optical potential is only known up to 1 GeV experimentally , it was proposed that this decrease of flow above 1 A GeV might indicate a decrease of the optical potential at high relative momenta or at high baryon density .
In Ref. the transverse momentum of the baryons has been disentangled into a collisional part, a mean-field part and a part originating from the Fermi-motion of the particles,
$$p_t=p_t^{\mathrm{coll}}+p_t^{\mathrm{MF}}+p_t^{\mathrm{Fermi}},$$
(4)
where $`p_t^{\mathrm{Fermi}}`$ practically does not contribute at midrapidity. Thus disentangling the flow signal into a collisional and a potential part, it turns out that $``$50 % of the flow stems from the particle - particle collisions while roughly another 50 % are generated by the potential repulsion at 1 A GeV. Fig. 3 shows these two contributions for flow for the system $`Ni+Ni`$ at b=4 fm in comparison to the experimental data using different EoS denoted by ’hard’, ’medium’ and ’soft’. Both the collisional ”background” and the potential part rise up to 1 A GeV incident energy and remain constant above, whereas the data indicate a decrease above 1 A GeV. As seen from Fig. 3 the $`Ni+Ni`$ data are described reasonably by both a ’soft’ and ’medium’ EoS, while a cascade calculation fails substantially.
Fig. 3: Transverse flow for a $`Ni+Ni`$ collision at b=4 fm as calculated within the RBUU model in cascade mode (dashed lower line) and for equations of state with different compressibilities $`K`$ in comparison to data from EOS, Plastic Ball and FOPI as compiled by Ref. .
A note of caution has to be added here: The flow $`F`$ (3) not only depends on the baryon self-energies $`U_S`$ and $`U_\mu `$ but also on the number of (resonance) degrees of freedom above about 1 A GeV as first pointed out by Hombach et al. . This observation will become crucial at AGS energies.
### 3.2 Radial flow
Radial flow has been discovered when analyzing the flow pattern of very central events of RHIC. In contrast to transverse flow up to about 70 % of the incident energy (stored in the hot compressed fireball) is released as ordered radial expansion of the nuclear matter. Thus the hope is to extract information especially on the compressibility of the EoS via the magnitude of the radial flow.
Experimentally the radial flow is characterized or fitted in terms of the Siemens-Rasmussen formula
$$\frac{d^3N}{dEd^2\mathrm{\Omega }}pe^{\gamma E/T}\left\{\frac{\mathrm{sinh}\alpha }{\alpha }(\gamma E+T)T\mathrm{cosh}\alpha \right\}$$
(5)
with $`\gamma =(1\beta ^2)^{1/2}`$ and $`\alpha =\gamma \beta p/T`$, while $`\beta `$ denotes the flow velocity and $`T`$ characterizes some temperature. We follow the same strategy in our RBUU calculations and apply a least square fit to the RBUU nucleon spectra using Eq. (5).
Fig. 4: The temperature (l.h.s.) and radial flow velocity (r.h.s.) for central $`Au+Au`$ collisions evaluated via Eq. (5) from the RBUU calculations in comparison to the experimental data from Refs. . The symbol ’s’ denotes a soft EoS without momentum dependent forces, ’h’ a hard EoS and ’smd’, ’mmd’ and ’hmd’ correspond to a soft, medium and hard momentum dependent EoS, respectively.
The results of the RBUU calculations for central $`Au+Au`$ collisions are shown in Fig. 4 as a function of the bombarding energy in comparison to the data from . We find that the ’temperature’ $`T`$ is systematically underpredicted in all schemes investigated (soft and hard EoS, with and without momentum dependent forces), and that the flow velocities are not correctly reproduced, being too low at low energy and crossing the experimental data around 800 A MeV. This might indicate a strong binding from the potential which gives not enough repulsion at high densities and overcompensates the collisional pressure from the fireball. However, the nucleon spectra resulting from the RBUU calculation show a strong non-thermal component at low incident energies and are thus in contradiction to the physical picture behind Eq. (5) which assumes an isentropic expansion of a thermal equilibrated source. In Ref. the degree of equilibration in a HIC has been investigated as a function of the incident energy and the system mass and it has been found that even the most massive systems like $`Au+Au`$ do not equilibrate at low energies.
## 4 AGS energies
### 4.1 Sidewards flow
We now turn to the AGS energy regime from 1 – 11 AGeV. The calculations are performed for the impact parameter $`b=6fm`$ for $`Au+Au`$ systems, since for this impact parameter we get the maximum flow which corresponds to the multiplicity bins $`M3`$ and $`M4`$ as defined by the Plastic Ball collaboration at BEVALAC/SIS energies. In Fig. 5 (l.h.s.) the transverse flow (3) is displayed in comparison to the data from Refs. for $`Au+Au`$ systems. The solid line (RBUU with $`\sqrt{s_{sw}}`$ = 3.5 GeV) is obtained with the scalar and vector self energies as discussed above, Eq. (1). The dotted line (CASCADE with $`\sqrt{s_{sw}}`$ = 3.5 GeV) corresponds to cascade calculations for reference in order to show the effect of the mean field relative to that from collisions. We observe that the solid line (RBUU, cf. Fig. 1) is in good agreement with the flow data at all energies; above bombarding energies of 6 A GeV the results are practically indentical to the cascade calculations showing the potential effects to cancel out.
We note that the sideward flow shows a maximum around 2 A GeV for $`Au+Au`$ and decreases continuously at higher beam energy ($``$ 2 A GeV) without showing any explicit minimum as in Ref. . This is due to the fact that the repulsive force caused by the vector mean field decreases at high beam energies (cf. Fig. 1) such that in the initial phase of the collision there are no longer strong gradients of the potential within the reaction plane. In subsequent collisions, which are important for $`Au+Au`$ due to the system size, the kinetic energy of the particles relative to the local rest frame is then in a range ($`E_{kin}`$ 1 GeV) where the Schrödinger equivalent potential (at density $`\rho _0`$) is determined by the experimental data . We thus conclude that for the sideward flow data up to 11 A GeV one needs a considerably strong vector potential at low energy and that one has to reduce the vector mean field at high beam energy in line with Fig. 1. In other words, there is only a weak repulsive force at high relative momenta or high densities.
Fig. 5: (l.h.s.) The sideward flow $`F(y)`$ as a function of the beam energy per nucleon for $`Au+Au`$ collisions at $`b=6`$ fm from the RBUU calculations. The solid line results for the parameter set RBUU, the dotted line for a cascade calculation with $`\sqrt{s_{sw}}`$ = 3.5 GeV. The data points are from the FOPI and EoS Collaborations . (r.h.s.) The elliptic flow $`v_2`$ of protons versus the beam energy per nucleon for $`Au+Au`$ collisions at $`b=6`$ fm from the RBUU calculations. The solid line results for the parameter set RBUU, the dotted line for a cascade calculation with $`\sqrt{s_{sw}}`$ = 3.5 GeV. The data points are from the EoS Collaboration .
Another aspect of the decreasing sideward flow can be related to the dynamical change in the resonance/string degrees of freedom as already discussed above. For instance, for $`\sqrt{s_{sw}}=2.6`$ GeV the calculated flow turns out to be smaller than the data above 1.5 A GeV and approaches the cascade limit already for $``$ 3-4 A GeV. This is due to the fact that in strings the incident energy is stored to a larger extent in their masses and the translational energy is reduced accordingly.
### 4.2 Elliptic flow
Apart from the in-plane flow of protons the out-of-plane collective flow provides additional information and constraints on the nuclear potentials involved. In this respect the elliptic flow for protons
$$v_2=(P_x^2P_y^2)/(P_x^2+P_y^2)$$
(6)
for $`|y/y_{cm}|0.2`$ is shown in Fig. 5 (r.h.s.) as a function of incident energy for $`Au+Au`$ collisions at $`b`$ = 6 fm. The solid line (RBUU with $`\sqrt{s_{sw}}`$ = 3.5 GeV) is obtained with the same mean fields as discussed before while the dotted line (CASCADE with $`\sqrt{s_{sw}}`$ = 3.5 GeV) stands again for the cascade results. The flow parameter $`v_2`$ changes its sign from negative at low energies ($`5A`$ GeV) to positive elliptic flow at high energies ($`5A`$ GeV).
This can be understood as follows: At low energies the squeeze-out of nuclear matter leads to a negative elliptic flow since projectile and target spectators distort the collective expansion of the ’fireball’ in the reaction plane. At high energies the projectile and target spectators do not hinder anymore the in-plane expansion of the ’fireball’ due to their high velocity ($`c`$); the elliptic flow then is positive. The competition between squeeze-out and in-plane elliptic flow at AGS energies depends on the nature of the nuclear force as pointed out already by Danielewicz et al. . We note, however, that in our calculation with the momentum-dependent potential (Fig. 1) we can describe both the sideward as well as elliptic flow data simultaneously without incorporating any phase transition in the EOS as in Ref. .
In the cascade calculation the elliptic flow from squeeze-out is weaker due to the lack of a nuclear force which demonstrates the relative role of the momentum-dependent nuclear forces on the $`v_2`$ observable below bombarding energies of about 5 A GeV.
## 5 Summary
In this contribution we have explored the dependence of transverse and radial flow signals on various model inputs - that are related to the nuclear EoS - using the coupled channel RBUU model. We find that the mass distribution of the resonances included in the model plays an important role for the description of transverse flow above 1 A GeV. For the radial flow we have concentrated on the difference between the results for the flow temperature $`T`$ and flow velocity $`\beta `$ when using different EoS. However, no sizeable sensitivity to the compressibility of the EoS could be established.
On the other hand, we found that in order to reach a consistent understanding of the nucleon optical potential up to 1 GeV, the transverse mass distributions of protons at AGS energies as well as the excitation function of sidewards and elliptic flow up to 11 A GeV, the strength of the vector potential has to be reduced in the RBUU model at high relative momenta and/or densities. Otherwise, too much flow is generated in the early stages of the reaction and cannot be reduced at later phases where the Schrödinger equivalent potential is experimentally known. This constrains the parameterizations of the explicit momentum dependence of the vector and scalar mean fields $`U_\mu `$ and $`U_S`$ at high relative momenta.
In addition, we have shown the relative role of resonance and string degrees of freedom at AGS energies. By reducing the number of degrees of freedom via high mass resonances one can build up a higher pressure and/or temperature of the ’fireball’ which shows up in the transverse mass spectra of protons as well as in the sidewards flow . A possible transition from resonance to string degrees of freedom is indicated by the RBUU calculations at invariant baryon-baryon collision energies of $`\sqrt{s}`$ 3.5 GeV which corresponds to a proton laboratory energy of about 4.6 GeV. Due to Fermi motion of the nucleons in $`Au+Au`$ collisions the transition from resonance to string degrees of freedom becomes smooth and starts from about 3 A GeV; at 11 A GeV practically all initial baryon-baryon collisions end up in strings, i.e. hadronic excitations in the continuum that decay to hadrons on a time scale of about 0.8 fm/c in their rest frame. This initial high-density string matter (up to 10 $`\rho _0`$ at 11 $`A`$ GeV) should not be interpreted as hadronic matter since it implies roughly 5 constituent quarks per fm<sup>3</sup>, which is more than the average quark density in a nucleon.
It is interesting to note that at roughly 4 A GeV in central $`Au+Au`$ collisions the ratio $`K^+/\pi ^+`$ is enhanced experimentally relative to transport calculations . One might speculate that a restoration of chiral symmetry could be responsible for the softer collective response as well as enhanced strangeness fraction. |
warning/0001/hep-th0001177.html | ar5iv | text | # Holography for degenerate boundaries
## I Introduction
Following the now famous conjecture of Maldacena , , , relating supergravity on anti-de Sitter spacetimes to a conformal field theory in one less dimension there has been a great deal of interest in supergravity compactifications on anti-de Sitter spaces. One of the most interesting features of the AdS/CFT correspondence is that it provides an example of the holographic principle . In the context of the Maldacena conjecture, holography was first discussed in detail in whilst discussions of holography in cosmology have appeared in , .
However, the holographic description of the anti-de Sitter bulk theory relies heavily on special features of the boundary, namely that the conformal boundary in the sense of Penrose is a non-degenerate manifold of codimension one. Although most of the physical negative curvature solutions that one is interested in considering, such as black holes, satisfy these properties, several classes of negatively curved spacetimes do not.
In particular, coset spaces such as $`SO(3,1)/SO(3)`$, Bergman type metrics and products of hyperbolic spaces have conformal boundaries of codimension greater than one, which we will refer to as degenerate boundaries. Typically, if $`1/ϵ`$ characterizes the radius of the $`(d+1)`$-dimensional spacetime, then the volume of a hypersurface of constant $`ϵ`$ behaves as $`ϵ^\alpha `$ with $`\alpha >d`$; the induced hypersurface does not blow up uniformly as $`ϵ0`$. First steps towards understanding the bulk boundary correspondence in these cases were taken in and the purpose of this paper is to take this correspondence further and deal with some of the unresolved issues of .
In particular, although the authors of were able to analyse the correspondence between scalar fields in the bulk and boundary scalar operators for the Bergman metric, they did not find such an interpretation for coset spaces which we shall do here. Furthermore, to properly formulate the bulk/boundary correspondence, one needs to understand how the bulk partition function gives rise to the partition function for the boundary conformal field theory.
When the conformal boundary is non-degenerate, there is a well understood way of defining the bulk (Euclidean) action without introducing a background , , . One introduces local boundary counterterms into the action, which remove all divergences and leave a finite action corresponding precisely to the partition function of the conformal field theory.
This procedure must break down when the boundary is degenerate, since there is no conformal frame in which the hypersurface radius $`ϵ`$ does not appear in the metric. This means that one cannot hope to eliminate all radius dependence from the partition function; put differently, as one takes the limit $`ϵ0`$, the partition function for the conformal field theory must diverge, since the geometry is becoming highly degenerate.
So the question is whether one can make sense of a correspondence between bulk and boundary partition functions. We will show that one can, provided that one takes the correspondence to be at finite radius. That is, an IR cutoff in the bulk theory will appear as a regulation of the conformal field theory target geometry. This is a novel manifestation of the IR/UV correspondence .
In several recent papers the relationship between the Randall/Sundrum scenario and AdS/CFT correspondence has been explored , , . In this context, one can view the five-dimensional negative curvature spacetime to be a construction from a symmetric non-degenerate four-dimensional brane world. In , , the brane world lives at finite radius $`R`$, and the induced action on the brane includes Einstein gravity plus corrections. For our “degenerate” brane worlds, however, the induced action on the brane does not include Einstein gravity, even when the brane is four-dimensional.
The plan of this paper is as follows. In §II, we discuss classes of metrics which have degenerate conformal boundaries. In §III, we discuss the interpretation of the gravitational bulk action for such spacetimes in terms of a dual conformal field theory. In §IV, we analyse how correlation functions for scalar operators in the conformal field theory may be derived from bulk massive scalar fields.
## II Classes of metrics with degenerate boundaries
Suppose that $`M`$ is a complete Einstein manifold of negative curvature and dimension $`d+1`$ which has a conformal boundary which a $`d`$-manifold $`N`$. This means that the metric of $`M`$ can be written near the boundary as
$$ds^2=\frac{du^2}{u^2}+\frac{1}{u^2}g_{ij}(u,x)dx^idx^j,$$
(1)
where $`u`$ is a smooth function with a first order zero on $`N`$ which is positive on $`M`$. Usually we assume that $`g_0=g(0,x)`$ is a non-degenerate metric on $`N`$, independently of how we take the limit $`u0`$. However, more general negative curvature manifolds can have conformal boundaries which are degenerate in the sense that $`g(ϵ,x)`$ is divergent as $`ϵ0`$. The vielbeins for such a metric will not all be finite as $`ϵ0`$; at least one will tend to zero or diverge.
Let us refer to $`N_ϵ`$ as the regulated boundary; then $`N_ϵ`$ will have a natural conformal structure. We will show in the following section that $`N_ϵ`$ must have negative curvature of the order of the $`(d+1)`$-dimensional cosmological constant when $`N`$ is degenerate. We expect there to be a correspondence between conformal field theory on $`N_ϵ`$ with a UV cutoff $`1/ϵ`$ and quantum gravity on $`M`$ with an IR cutoff $`ϵ`$. To investigate this correspondence we will as usual consider the relation between bulk and boundary partition functions. Before we do so, however, it will be helpful to give several examples of spaces which have degenerate boundaries; we will be using these explicit examples to illustrate general arguments in the following sections.
We will consider here two generic classes of negative curvature manifolds which have degenerate conformal boundaries. Let us normalise the $`(d+1)`$-dimensional Einstein action such that
$$I_{\mathrm{bulk}}=\frac{1}{16\pi G_{d+1}}_Md^{d+1}x\sqrt{g}(R_g+d(d1)l^2)\frac{1}{8\pi G_{d+1}}_Nd^dx\sqrt{\gamma }K,$$
(2)
where $`R_g`$ is the Ricci scalar and $`K`$ is the trace of the extrinsic curvature of the boundary $`N`$ embedded in $`M`$.
An example of the first class of degenerate boundary manifolds was considered in ; it is included in the family of Einstein-Kähler metrics of the form
$`ds^2`$ $`=`$ $`(l^2r^2+{\displaystyle \frac{1}{4}}{\displaystyle \frac{k}{r^4}})^1dr^2+r^2(l^2r^2+{\displaystyle \frac{1}{4}}{\displaystyle \frac{k}{r^4}})(d\psi +2\mathrm{cos}\theta d\varphi )^2`$ (4)
$`+r^2(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),`$
which was constructed in . Regularity requires that the periodicity of $`\psi `$ is
$$\beta =\frac{8\pi }{k}=\frac{4\pi }{U^{}(r_+)r_+},$$
(5)
where the horizon location is $`r_+`$ and $`U(r)`$ is the metric function. The boundary admits the conformal structure
$$ds^2=(l^2r^2+\frac{1}{4}\frac{k}{r^4})(d\psi +2\mathrm{cos}\theta d\varphi )^2+(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$
(6)
which is a squashed three-sphere of the form
$$ds^2=l_3^2(d\stackrel{~}{\psi }+\mathrm{cos}\theta d\varphi )^2+(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$
(7)
and becomes degenerate as we take the limit $`r\mathrm{}`$. The $`k=0`$ metric is the Bergman metric which is the group manifold $`SU(2,1)`$ discussed in ; it will be convenient to use here an alternative form of the Bergman metric
$$ds^2=\frac{2}{l^2}[d\rho ^2+\frac{1}{4}\mathrm{sinh}^2\rho \mathrm{cosh}^2\rho (d\psi +\mathrm{cos}\theta d\varphi )^2+\frac{1}{4}\mathrm{sinh}^2\rho (d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2)],$$
(8)
where the prefactor ensures that the curvature behaves as $`R_g=12l^2`$. Although the boundary of the Bergman metric becomes degenerate in the infinite limit, there is a very natural choice of conformal boundary since this space, like others in the same family, is constructed by radially extending a $`U(1)`$ bundle over an Einstein space.
Much that one says about the Bergman metric can also be extended to other radial extensions of bundles over compact spaces. Such spaces have been considered in the past in the context of compactifications and a number of other examples are known. For example, in eight dimensions one could consider an Einstein hyper-Kähler manifold of the form
$$ds^2=d\rho ^2+\frac{1}{4}\mathrm{sinh}^2\rho [d\theta ^2+\frac{1}{4}\mathrm{sin}^2\theta \omega _i^2+\frac{1}{4}\mathrm{cosh}^2\rho (\nu _i+\mathrm{cos}\theta \omega _i)^2],$$
(9)
where $`\nu _i,\omega _i`$ are left-invariant forms satisfying $`SU(2)`$ algebras. This metric is obtained by analytically continuing the standard Fubini-Study Einstein metric on $`HP^2`$, the quaternionic projective plane. Hypersurfaces of constant $`\rho `$ are squashed seven-spheres which become infinitely squashed as one takes the boundary to infinity. Analysis of the AdS/CFT correspondence in this case would be very similar to that for the Bergman metric. One would also expect that one could find an Einstein metric which is a radial extension of a twisted $`S^2`$ bundle over $`S^2`$. Hypersurfaces of constant radius should correspond to off-shell extensions of the Page metric .
We should mention that these metrics do not have a nice physical Lorentzian interpretation. The existence of a topologically non-trivial degenerate conformal boundary is related to the non-trivial fibration of $`U(1)`$ coordinates which would have to be analytically continued to give a Lorentzian metric. Analytically continuing the Bergman metric leads to a Lorentzian metric which is complex and whose physical interpretation is unclear. Furthermore, not only are these metrics not supersymmetric (the Bergman metric is certainly not supersymmetric since there is no supergroup which has $`SU(2,1)`$ as its bosonic part) but also many such metrics will not even admit a spin structure.
Given the problems with the Lorentzian interpretation, we shouldn’t be surprised if the thermodynamic quantities in the conformal field theory take unphysical values. It is well known that a generic quantum field theory within a causality violating background will exhibit pathologies such as negative entropy which one interprets as reflecting the unphysical nature of the background and we will see similar phenomena here. However, although such backgrounds are not of great physical interest in string theory they provide interesting examples of a more general AdS/CFT correspondence. There may be other backgrounds in string theory which are physically interesting and exhibit similar degenerate behaviour.
The second category of manifolds in which we are interested is characterized by the existence of more than one “radial” coordinate. The simplest example which we will consider in the most detail is the product of hyperbolic manifolds of dimensions $`d_1`$ and $`d_2`$ respectively:
$$ds^2=\frac{(d_11)}{(d_1+d_21)l^2z_1^2}(dz_1^2+d𝐱_1^2)+\frac{(d_21)}{(d_1+d_21)l^2z_2^2}(dz_2^2+d𝐱_2^2).$$
(10)
“Infinity” in this metric is obtained by taking either $`z_1`$ or $`z_2`$ to zero; however, for a holographic principle to be formulated we need to define a boundary of codimension one. One way to do this would be to effectively divide the boundary into two parts: consider a hypersurface of constant $`z_1=ϵ_1<<1`$ and a hypersurface of constant $`z_2=ϵ_2<<1`$ which are glued together along the hypersurface of codimension two $`z_1=ϵ_1;z_2=ϵ_2`$. However, since one would need to be careful about boundary conditions for fields along this join it is easier to work with the conformally equivalent surface defined by setting
$$z_1=u\mathrm{cos}\theta ,z_2=u\mathrm{sin}\theta ,$$
(11)
so that for example when $`d_1=d_2=2`$ the metric becomes
$`ds^2`$ $`=`$ $`{\displaystyle \frac{1}{3l^2}}\{{\displaystyle \frac{2du^2}{u^2}}+{\displaystyle \frac{2dud\theta }{u}}({\displaystyle \frac{\mathrm{cos}\theta }{\mathrm{sin}\theta }}{\displaystyle \frac{\mathrm{sin}\theta }{\mathrm{cos}\theta }})+{\displaystyle \frac{(\mathrm{sin}^4\theta +\mathrm{cos}^4\theta )}{\mathrm{sin}^2\theta \mathrm{cos}^2\theta }}d\theta ^2`$ (13)
$`+{\displaystyle \frac{dx_1^2}{u^2\mathrm{cos}^2\theta }}+{\displaystyle \frac{dx_2^2}{u^2\mathrm{sin}^2\theta }}\}.`$
With this choice of coordinates the induced metric on constant $`u`$ hypersurfaces is
$$ds^2=\frac{d\theta ^2}{2\mathrm{sin}^2\theta \mathrm{cos}^2\theta }+\frac{dx_1^2}{u^2\mathrm{cos}^2\theta }+\frac{dx_2^2}{u^2\mathrm{sin}^2\theta },$$
(14)
which is degenerate as $`u0`$ and further degenerates when $`\theta 0,\pi /2`$. Notice that the metric is non-singular in this conformal frame, with the limits in $`\theta `$ corresponding to non-singular tubes in the Euclidean metric.
As well as products of negative curvature manifolds, various coset spaces exhibit a similar degenerate behaviour. In particular, one can analytically continue manifolds which were considered in the context of supergravity compactifications, reviewed in . The coset space $`SO(3,1)/SO(2)`$ possesses an Einstein metric
$$ds^2=\frac{1}{9}(d\psi +i\mathrm{cosh}\rho _1d\varphi _1+i\mathrm{cosh}\rho _2d\varphi _2)^2+\frac{1}{6}\underset{i}{}(d\rho _i^2+\mathrm{sinh}^2\rho _id\varphi _i^2),$$
(15)
where we have taken $`l^2=\frac{1}{4}`$. The Euclidean metric is not real, but one can find a Lorentzian section (discussed in ) which is; however, our Lorentzian theory will still exhibit pathologies since there are closed timelike curves.
The most useful choice of boundary in this case is probably to take
$$e^{\rho _1}=\frac{1}{u\mathrm{cos}\theta };e^{\rho _2}=\frac{1}{u\mathrm{sin}\theta },$$
(16)
in analogy to the above. Then as $`u0`$ the leading order induced metric on a constant $`u`$ surface is
$`ds^2`$ $`=`$ $`{\displaystyle \frac{d\theta ^2}{12\mathrm{sin}^2\theta \mathrm{cos}^2\theta }}+{\displaystyle \frac{d\varphi _1^2}{72u^2\mathrm{cos}^2\theta }}+{\displaystyle \frac{d\varphi _2^2}{72u^2\mathrm{sin}^2\theta }}+{\displaystyle \frac{1}{9}}d\psi ^2`$ (18)
$`+{\displaystyle \frac{i}{9u}}d\psi ({\displaystyle \frac{d\varphi _1}{\mathrm{cos}\theta }}+{\displaystyle \frac{d\varphi _2}{\mathrm{sin}\theta }}){\displaystyle \frac{1}{18u^2\mathrm{sin}\theta \mathrm{cos}\theta }}d\varphi _1d\varphi _2,`$
which has a $`so(2)^3`$ symmetry group and is degenerate as $`\theta 0,\pi /2`$.
All of the examples so far are symmetric negative curvature manifolds with degenerate boundaries. One could also start with a symmetric manifold which is degenerate as some parameter $`u`$ is taken to zero. One would then re-interpret $`u`$ as an effective radius and radially extend the hypersurface to construct an Einstein manifold in one higher dimension. This point of view was discussed in in the context of negative curvature spacetimes with non-degenerate boundaries. It will become clearer in the following section how one would radially extend the hypersurface in the degenerate case.
The conformal symmetry group will be larger than in most of the examples given above, and so one will be able to fix more quantities in the conformal field theory. However, calculating bulk quantities will be correspondingly more difficult, since the bulk symmetry group will in general be smaller. This will be particularly relevant when trying to derive, for example, scalar correlation functions from bulk actions for massive scalar fields.
## III Regularisation of the Euclidean action
One of the most interesting developments arising from the AdS/CFT correspondence has been the use of counterterms in the Euclidean action to define the action independently of background for negative (and in certain limits zero) curvature manifolds , , . Consider an Einstein manifold which satisfies the field equations derived from (2), whose metric near the boundary can be written in the form
$$ds^2=\frac{dx^2}{l^2x^2}+\frac{1}{x^2}\gamma _{ij}dx^idx^j,$$
(19)
where $`\gamma `$ is finite and non-degenerate on the boundary itself. In this section we will consider only four-dimensional spacetimes, since this is the dimensionality of the explicit examples which we will use. Following a theorem by Fefferman and Graham , the conformal metric $`\gamma _{ij}`$ can be written , as
$$\gamma _{ij}=\gamma _{ij}^0+x^2\gamma _{ij}^2+x^4\gamma _{ij}^4+\mathrm{},$$
(20)
where $`\gamma ^2`$ is defined in terms of the curvature of $`\gamma ^0`$ as
$$\gamma _{ij}^2=\frac{1}{4l^2}(R_{ij}^0\frac{1}{4}R^0\gamma _{ij}^0),$$
(21)
and $`\gamma ^4`$ depends on fourth derivatives of $`\gamma ^0`$. $`\gamma ^0`$ is independent of $`x`$ when the conformal boundary is non-degenerate. Note that the choice of bulk conformal frame - or in other words, the magnitude of the cosmological constant - combined with the existence of a non-degenerate conformal boundary effectively fixes the coefficient of the $`dx^2`$ term in (19) to be $`1/l^2`$. We emphasise this point since it will be important in what follows.
We can then formally expand the Einstein action as
$`I_{\mathrm{bulk}}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi G_4}}{\displaystyle _{N_ϵ}}d^4x\sqrt{g}(R(g)+6l^2){\displaystyle \frac{1}{8\pi G_4}}{\displaystyle _{N_ϵ}}d^3x\sqrt{\gamma }K;`$ (22)
$`=`$ $`{\displaystyle \frac{1}{16\pi G_4}}{\displaystyle _{N_ϵ}}d^3x\sqrt{\gamma ^0}({\displaystyle \frac{4l}{ϵ^3}}+{\displaystyle \frac{16l}{ϵ}}\mathrm{tr}(\gamma ^0)^1\gamma ^2+\mathrm{});`$ (23)
$`=`$ $`{\displaystyle \frac{l}{4\pi G_4}}{\displaystyle _{N_ϵ}}d^3x\sqrt{\stackrel{~}{\gamma }}(1+{\displaystyle \frac{1}{4l}}R(\stackrel{~}{\gamma })+\mathrm{}),`$ (24)
where the inverse radius of the hypersurface over which we integrate, $`ϵ1`$, is an IR regularisation parameter and $`x^2\stackrel{~}{\gamma }=\gamma `$. The ellipses indicate non-divergent terms which we have omitted. The second equality is obtained by using the expansion of the boundary metric (20), (19) and integrating the bulk action explicitly. The key point is that since there are only a finite number of divergent terms one can introduce a local counterterm action $`I_{\mathrm{ct}}`$ dependent only on $`\stackrel{~}{\gamma }_{ij}`$ and its covariant derivatives
$$I_{\mathrm{ct}}=\frac{l}{4\pi G_4}_{N_ϵ}d^3x\sqrt{\stackrel{~}{\gamma }}(1+\frac{1}{4l}R(\stackrel{~}{\gamma })).$$
(25)
Then the content of the AdS/CFT conjecture is that we make the identification that
$$I_{\mathrm{bulk}}=W_{\mathrm{cft}}+I_{\mathrm{ct}},$$
(26)
where we take the boundary to be the true conformal boundary and $`W_{\mathrm{cft}}`$ is the (finite) partition function for the conformal field theory. The purpose of this section is to show how and why this analysis breaks down when the boundary becomes degenerate.
### A Degenerate boundaries and counterterm regularisation
For the manifolds with degenerate boundaries considered here, the expansion of $`\gamma _{ij}`$ given in (20) breaks down; its derivation in fact relies on the existence of a non-degenerate conformal boundary of codimension one , . We are going to consider a more general form for the expansion of the metric near the conformal boundary
$$ds^2=\frac{dx^2}{l^2x^2}+x^\delta \gamma _{ij}dx^idx^j.$$
(27)
We will assume that there is a well-defined expansion for the boundary metric of the form
$$\gamma _{ij}=\gamma _{ij}^0+\gamma _{ij}^2+\mathrm{}.,,$$
(28)
but $`\gamma ^0`$ will now depend on $`x`$. There is a preferred frame in which its determinant is independent of $`x`$: this choice of $`\gamma ^0`$ is natural if one requires that both the determinant and the inverse metric are well-defined as $`x0`$. We assume from here on that this choice of normalisation for $`\gamma ^0`$ is imposed in the metric (27), which along with the normalisation of the coefficient of $`dx^2`$ in (28), effectively determines the choice of the coordinate $`x`$ given an Einstein metric satisfying (2). $`\gamma ^2`$ will be subleading in the sense that its determinant behaves as a positive power of $`x`$ as one takes the limit $`x0`$. As we will see later on in this section, the explicit metrics given in §II all admit an expansion of this form.
Suppose that $`\gamma ^0`$ effectively degenerates to a $`p`$-dimensional metric in the limit that $`x0`$, where $`p`$ will be determined by the number of independent vielbeins. Note that the $`p`$-dimensional metric does not in general have a non-zero determinant nor will the associated vielbeins be closed. In fact, the Bergman metric degenerates to a metric with zero determinant whose associated single vielbein is not closed. Then a typical degenerate boundary metric $`\gamma ^0`$ might be written as
$$\gamma _{ij}^0=x^1h_{ij}^{(p)}+x^{\frac{p}{3p}}h_{ij}^{(3p)},$$
(29)
where $`h^q`$ is of dimension $`q`$ in the sense defined above. Now $`\gamma ^2`$ is defined by requiring that the metric (27) satisfies the Einstein equations expanded out in powers of $`x`$. By analysing the field equations, however, we find that although $`\gamma ^2`$ is well-defined, it is not a covariant tensor: it cannot be written in terms of $`\gamma ^0`$ and its curvature invariants. The definition of $`\gamma ^2`$ for (29) is not particularly illuminating since it is not generic; it involves second derivatives of $`h^p`$ and $`h^{3p}`$ as one would expect.
Using the asymptotic form for the metric, the Einstein action for a manifold with a degenerate boundary (27) can then be written as
$`I_{\mathrm{bulk}}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi G_4}}{\displaystyle _{N_ϵ}}d^4x\sqrt{g}(R(g)+6l^2){\displaystyle \frac{1}{8\pi G_4}}{\displaystyle _{N_ϵ}}d^3x\sqrt{\gamma }K,`$ (30)
$`=`$ $`{\displaystyle \frac{l}{16\pi G_4}}{\displaystyle _{N_ϵ}}d^3x\sqrt{\stackrel{~}{\gamma }}({\displaystyle \frac{4}{\delta }}3\delta )+I_{\mathrm{nl}},`$ (31)
where the second equality follows from explicitly substituting the metric (27) into the action and integrating. $`\stackrel{~}{\gamma }=ϵ^\delta \gamma `$ is the metric induced on a codimension one hypersurface of constant $`ϵ`$ and the integral is taken over a hypersurface of constant $`ϵ1`$.
The second part of the action, $`I_{\mathrm{nl}}`$, includes non-local terms and cannot be expressed covariantly in terms of the boundary metric. This term in the action is not in general finite, but diverges as one takes the limit $`ϵ0`$. However, the leading order divergence of the bulk action (which behaves as $`ϵ^{3\delta /2}`$) can be removed by subtracting the first term in (31); this follows from the condition that $`\gamma ^2`$ is subleading to $`\gamma ^0`$. One will be left with a leading order divergent term in $`I_{\mathrm{nl}}`$ which behaves as $`ϵ^a`$ with $`a<3\delta /2`$. Note that the first term in the action (31) agrees with that for non-degenerate boundaries when one takes $`\delta =2`$.
One implication of the above is that one cannot introduce local counterterms to remove the divergence of the bulk action as $`ϵ0`$. Suppose we tried to take a counterterm action of the form
$$I_{ct}=\frac{1}{16\pi G_4}_{N_ϵ}d^3x\sqrt{\stackrel{~}{\gamma }}[a_0[\delta ]+a_1R(\stackrel{~}{\gamma })+a_2(R(\stackrel{~}{\gamma })^2+b_2R^{ij}(\stackrel{~}{\gamma })R_{ij}(\stackrel{~}{\gamma })+\mathrm{}.).$$
(32)
Provided we pick the first coefficient $`a_0[\delta ]`$ according to (31) we can remove the leading order divergence - but there is no generic way to define the other coefficients. In fact, even choosing $`a_0[\delta ]`$ in this way really represents a fine-tuning which we are not allowed to do. One more general grounds, we can see that this series cannot be convergent in $`ϵ`$ without adjusting the coefficients to each solution. Since the curvature invariants of hypersurfaces of constant $`ϵ`$ are of order $`l`$ for degenerate boundaries (compared to invariants of order $`ϵ`$ to positive powers for non-degenerate boundaries), there is no small expansion parameter and no reason for the series to converge.
This behaviour of the curvature invariants follows from the Gauss-Codacci condition for the induced hypersurface
$$R(\stackrel{~}{\gamma })=(K^2K_{ab}K^{ab}6l^2),$$
(33)
where $`K_{ab}`$ is the extrinsic curvature of the hypersurface and $`K`$ is its trace as before. For a metric which can be written in the form (20) with $`\gamma ^0`$ non-degenerate, then
$$K^2=9l^2+𝒪(ϵ^2);K_{ab}K^{ab}=3l^2+𝒪(ϵ^2),$$
(34)
and so the curvature invariants of the hypersurface behave as positive powers of $`ϵ`$, which is really the basis of the counterterm subtraction procedure . However, if $`\gamma ^0`$ is degenerate and, for example, of effective dimension $`p`$, then
$$K^2=\frac{(3+p)^2}{4}l^2+𝒪(ϵ^2);K_{ab}K^{ab}=\frac{3(p+1)}{4}l^2+𝒪(ϵ^2),$$
(35)
and so, as previously mentioned, the curvature of the hypersurface is negative and of order $`l`$. It is a generic feature of spaces with degenerate conformal boundary that the induced metric on the boundary has finite negative curvature, rather than an infinite curvature radius as is usual. We should perhaps mention here that some of the analysis of the AdS/CFT correspondence relies on non-negative curvature of the CFT background spacetime. In particular, the discussion in relies specifically on a conformal boundary of positive scalar curvature. Interesting issues that arise even in the non-degenerate case when the boundary has negative curvature are discussed in .
Of course after a little reflection we should not be surprised that local counterterms cannot remove the divergence of the bulk action. Since $`ϵ`$ appears explicitly in the conformal field theory background geometry, we cannot expect the partition function to be independent of this parameter. Furthermore we should probably expect the partition function to diverge as the geometry becomes degenerate.
As a simple example let us consider a generic conformal field theory on a $`d`$-dimensional background
$$ds^2=u^{d1}d\tau ^2+u^1h_{ij}dx^idx^j,$$
(36)
where $`\tau `$ is the trivially fibered imaginary time coordinate with period $`\beta `$ and $`h`$ is a non-degenerate metric. We suppose that $`u`$ corresponds to a radial parameter in the bulk theory, and this metric is conformal to that induced on hypersurfaces of constant $`u`$. As $`u0`$, the metric will become degenerate, although in this (preferred) conformal frame the determinant remains regular. Suppose we now conformally rescale the metric such that
$$\stackrel{~}{ds}^2=d\stackrel{~}{\tau }^2+h_{ij}dx^idx^j,$$
(37)
where we have defined a new imaginary time coordinate $`\stackrel{~}{\tau }=u^{\frac{d}{2}}\tau `$. In this conformal frame it is trivial to write down the main dependence of the partition function since the effective temperature is high in the degenerate limit: $`\stackrel{~}{\beta }0`$ as $`u0`$. This means that the partition function for the conformal field theory behaves as
$$W_{\mathrm{cft}}T^{d1}u^{\frac{d(d1)}{2}},$$
(38)
where $`T`$ is the inverse of $`\beta `$, and would be interpreted as the finite temperature of the bulk theory. Thus the partition function does indeed diverge as the geometry becomes degenerate.
Given a generic $`d`$-dimensional metric which becomes degenerate in the sense considered here as some parameter $`u0`$ we can construct a $`(d+1)`$-dimensional metric satisfying the equations derived from (2) as follows. Firstly, we should find the conformal frame in which the $`d`$-dimensional metric determinant is independent of $`u`$. Then we should write the higher-dimensional metric in the form (27) and fix $`\delta `$ from the leading order terms in the Einstein equations. $`\gamma ^2`$ will follow from an expansion in powers of $`u`$.
### B Interpretation of the bulk action
If one cannot remove all the divergences in the action with covariantly defined counterterms, one has to decide how to interpret the bulk action in terms of the dual conformal field theory. One suggestion - close in spirit to interpretations of the Randall-Sundrum scenario in terms of the AdS/CFT correspondence , , \- is the following. Instead of the ultimate goal being to take the $`ϵ0`$ limit so that the cutoff boundary becomes the true boundary, we need to keep the boundary at finite $`ϵ`$. This will ensure that the background geometry for the dual conformal field theory is non-degenerate.
In the Randall-Sundrum scenario an Einstein term is induced into the effective action on the hypersurface, plus a cosmological term which we can effectively adjust to zero by adding a brane tension term . The presence of these two terms in the induced action is manifest from the counterterm action (25). However, for degenerate boundaries there is no Einstein term in the “hypersurface” action. The leading order propagator will follow from differentiating the action twice with respect to the hypersurface metric $`\stackrel{~}{\gamma }`$. Unsurprisingly one can’t get a sensible brane world scenario from a higher-dimensional metric with degenerate boundary.
The natural suggestion for the correspondence between the bulk and conformal field theory partition functions is that we should simply take
$$I_{\mathrm{bulk}}(ϵ)W_{\mathrm{cft}}(ϵ),$$
(39)
where $`W_{\mathrm{cft}}(ϵ)`$ is the partition function for the conformal field theory in a geometry regulated by $`ϵ`$. Following the discussion in the last subsection, one might question why we don’t take the correspondence to be instead
$$I_{\mathrm{nl}}(ϵ)W_{\mathrm{cft}}(ϵ),$$
(40)
where we have removed the leading order divergence of the bulk action by subtracting a counterterm of the form (31). However subtraction of such a counterterm would not be satisfactory from a holographic point of view, since one would need to know the index $`\delta `$ to carry out the subtraction but $`\delta `$ is not known by the conformal boundary geometry. Another way of saying this is that one effectively has to adjust the subtraction to the bulk geometry rather than taking a generic subtraction. Of course in the non-degenerate case one still needs to know that $`\delta =2`$ to carry out the subtraction but the regularity of the geometry of the regularisation limit implicitly tells us that $`\delta =2`$.
This proposal for the correspondence is equivalent to taking a strong version of the holographic principle : it assumes that quantum gravity on any volume contained within a manifold can be described by a theory defined on the boundary of the volume. This is the basis for the recent work of , , but our proposal extends this principle to more general negative curvature manifolds. One should be able to make more precise the correspondence between the bulk field equations and the renormalisation group flow equations in the conformal field theory along the lines of . The difference will be that, in addition to the renormalisation group flow in the conformal field theory as one flows in from infinity, one will also have a flow in the effective target space geometry for the conformal field theory.
We should also mention that, although this procedure for cutting off the interior path integral at a finite boundary seems to be the right thing to do to compare partition functions, we probably need to be more careful about how we do this. Simply cutting off the theory will throw out some physics since it will not tell us about physical processes in which particles propagate across our cutoff boundary. However, our naive approach will be adequate for the discussions here.
To derive other thermodynamic quantities in the boundary conformal field theory from the bulk, one would need to use the quasilocal tensor defined by Brown and York as
$$T^{\mu \nu }=\frac{2}{\sqrt{\stackrel{~}{\gamma }}}\frac{\delta I_{\mathrm{bulk}}}{\delta \stackrel{~}{\gamma }_{\mu \nu }},$$
(41)
and then define conserved quantities associated with Killing vectors $`\xi `$ as
$$Q_\xi (ϵ)=_{\mathrm{\Sigma }_ϵ}d^2x\sqrt{\sigma }T_{\mu \nu }u^\mu \xi ^\nu ,$$
(42)
where $`u`$ is the unit normal to a hypersurface $`\mathrm{\Sigma }_ϵ`$ in $`N_ϵ`$. The thermodynamic relation between these quantities would be defined as usual as
$$I_{\mathrm{bulk}}(ϵ)=\beta M(ϵ)+\mathrm{}S(ϵ),$$
(43)
where $`\beta `$ is the inverse temperature, and $`M(ϵ)`$ and $`S(ϵ)`$ correspond to the mass and entropy respectively of the regulated conformal field theory.
### C Example 1: The Bergman metric
To check whether the bulk/boundary partition functions do diverge in the same way, let us try to calculate both for some of the metrics discussed in §II. Suppose we introduce into the Bergman metric an IR cutoff $`\mathrm{sinh}\rho =lR1`$ so that the boundary geometry is conformal to
$$ds^2=l^2R^2(d\psi +\mathrm{cos}\theta d\varphi )^2+(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2);$$
(44)
Then the bulk Euclidean action is
$$I=\frac{5\pi l^2R^4}{4G_4}\frac{3\pi R^2}{2G_4}.$$
(45)
To calculate the surface term in (31), we need to bring the metric near the conformal boundary into the form (27). Defining $`x=2e^{\sqrt{2}\rho }`$ then the leading order terms in the metric are
$$ds^2=\frac{dx^2}{l^2x^2}+\frac{x^{\frac{4\sqrt{2}}{3}}}{2l^2}\{x^{\frac{2\sqrt{2}}{3}}(d\psi +\mathrm{cos}\theta d\varphi )^2+x^{\frac{\sqrt{2}}{3}}(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2)\},$$
(46)
from which we see that we must take $`\delta =4\sqrt{2}/3`$ in (27), and hence the first term in (31) becomes
$$I_{\mathrm{surf}}=\frac{5\pi l^2R^4}{4G_4}+\mathrm{},$$
(47)
which as expected coincides with the leading order divergence of the effective action.
As usual, strong coupling prevents us from calculating the partition function for the associated conformal theory on the squashed three sphere directly; however, in this case, we can calculate the $`R`$ dependence by an indirect method. Supergravity in negative curvature Taub-Nut and Taub-Bolt manifolds also corresponds to the $`(2+1)`$ dimensional “exotic” conformal field theory which lives on the world volume of M2-branes after placing them on a squashed three sphere. There is of course a very close relationship between the AdS Taub-Bolt manifolds and the Bergman metric. The Bergman metric is a radial extension of the second power of the Hopf bundle over $`S^2`$ whilst the nut and bolt metrics are radial extensions of the first power of the Hopf bundle over $`S^2`$ . There is no problem in calculating the regularised Euclidean action for the Taub-Nut and Taub-Bolt manifolds which have non-degenerate boundaries. The metric for the nut solution <sup>*</sup><sup>*</sup>*The bolt solution does not exist in the parameter range relevant here. is
$`ds^2`$ $`=`$ $`V(r)(d\tau +2n\mathrm{cos}\theta d\varphi )^2+V^1(r)dr^2+(r^2n^2)(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2);`$ (48)
$`V(r)`$ $`=`$ $`{\displaystyle \frac{(rn)(l^2r^2+2nl^2r+13n^2l^2)}{(r+n)}},`$ (49)
and the action was calculated using counterterm subtraction in
$$I=\frac{4\pi n^2}{G_4}(12n^2l^2),$$
(50)
with the boundary geometry behaving as
$$ds^2=4n^2l^2(d\psi +\mathrm{cos}\theta d\varphi )^2+(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$
(51)
where we identify $`\tau \psi n`$. The usual dictionary for the AdS/CFT correspondence implies that we should take
$$N^{\frac{3}{2}}\frac{1}{l^2G_4},$$
(52)
where $`N`$ is a measure of the number of unconfined degrees of freedom for the gauge theory describing the dynamics of $`N`$ parallel M2-branes wrapped on a squashed three sphere. So to compare the conformal field theory in the background geometry (51) with that in (44) we need to take the same values of $`l^2G_4`$ and set $`R=2n`$. In this limit the conformal field theory partition function behaves as
$$I=\frac{\pi R^2}{G_4}(1\frac{R^2l^2}{2}).$$
(53)
In the extreme squashing limit, the action diverges in the same way as the bulk action for the Bergman metric. Of course, we shouldn’t expect the coefficients to agree, since we can’t assume that the two spacetimes correspond to the same state in the conformal field theory Indeed, if we accept the hypothesis (39) as true, then the entropy for the Bergman metric is positive whereas that for the nut solutions is negative, so the Bergman metric corresponds to a highly excited state. Of course the use of this argument is circular. Note that the negativity of the entropy for the nut solution can be viewed as a manifestation of the pathologies in the causal structure as discussed in §2.. However, since the degeneracy of the geometry will determine the leading order divergence of the partition function, we should expect the actions to diverge in the same way as we take $`R`$ to infinity.
There is a possible flaw in the above argument. It is not obvious that we can regulate the action for the nut spacetime and then take a singular limit in $`n`$; these operations do not necessarily commute, since the spacetime becomes very singular as $`n\mathrm{}`$. Although we should be reassured that a very similar limiting process appears to work when one calculates the action for critically rotating black holes , it would nice to check the above conclusions in another way.
Since the leading order behaviour of the partition function should not depend on the details of the conformal field theory as $`R\mathrm{}`$ it should be reproduced by the partition function for free conformally coupled scalar and spinor fields in this background. A related calculation was carried out in ; it was found that if one considered eigenmodes of a scalar field on a squashed sphere satisfying
$$(+\frac{1}{4})\mathrm{\Phi }_k=\lambda _k\mathrm{\Phi }_k,$$
(54)
then the partition function obtained from the zeta function $`\zeta (s)=_k\lambda _k^s`$ did indeed behave as $`R^4`$ in the extreme oblate limit. However, this calculation is not directly relevant to conformally coupled fields as Andy Strominger has also pointed out to me.: in the large $`R`$ limit, the operator (54) is very different from the conformally invariant operator
$$(+\frac{R_g}{8})=(+\frac{1}{4}\frac{R^2}{16}),$$
(55)
where $`R_g`$ is the Ricci scalar. It is not difficult to apply the same techniques to show that the divergence as $`R^4`$ persists for the conformally coupled operator (54); the analysis mirrors that of and is summarised in the Appendix. Note that there doesn’t seem to be any natural intuitive explanation for the $`R^4`$ dependence of the partition function; it follows in a non-trivial way from the geometry.
### D Example 2: $`H^2\times H^2`$
The second example we will consider is the product of two hyperbolic spaces $`H^2\times H^2`$ whose Einstein action is
$$I=\frac{1}{48\pi G_4l^2}\frac{d\theta dx_1dx_2}{u^2\mathrm{sin}^2\theta \mathrm{cos}^2\theta }=\frac{\sqrt{2}\sigma _3}{48\pi G_4l^2u^2},$$
(56)
where we have introduced a regulated volume $`\sigma _3`$ for the volume of non-compact hypersurfaces of constant $`u`$ in the induced boundary metric
$$ds^2=\frac{u^{\frac{4}{3}}d\theta ^2}{2\mathrm{sin}^2\theta \mathrm{cos}^2\theta }+\frac{dx_1^2}{u^{\frac{2}{3}}\mathrm{cos}^2\theta }+\frac{dx_2^2}{u^{\frac{4}{3}}\mathrm{sin}^2\theta }.$$
(57)
The metric can be brought into the form (27) with the choice of coordinate
$$x=u^{\frac{\sqrt{2}}{\sqrt{3}}}\mathrm{sin}^{\frac{1}{\sqrt{6}}}\theta \mathrm{cos}^{\frac{1}{\sqrt{6}}}\theta ,$$
(58)
and hence the above analysis is applicable here. Since there is in this case no obvious supergravity background with a related conformal boundary for which we can also calculate the action, the best that we can do is to check whether we can reproduce this form of the partition function from conformally coupled scalars in the background (57). The Ricci scalar for this metric is
$$R_g=44\mathrm{cos}^2\theta \mathrm{sin}^2\theta ,$$
(59)
and modes of a conformally coupled scalar field behaving as $`\varphi (\theta ,x_1,x_2)\varphi _\alpha (\theta )e^{ik_ix_i}`$ satisfy the equation
$`(2\mathrm{cos}^2\theta \mathrm{sin}^2\theta _\theta ^2\mathrm{cos}^2\theta k_1^2u^2\mathrm{sin}^2\theta k_2^2u^2+{\displaystyle \frac{1}{2}}+{\displaystyle \frac{1}{2}}\mathrm{cos}^2\theta \mathrm{sin}^2\theta )\varphi _\alpha (\theta )`$ (60)
$`=\lambda (\alpha ,k_iu)\varphi _\alpha (\theta ).`$ (61)
In fact we don’t need to find the eigenvalues explicitly; all we need to know is that the eigenvalues $`\lambda `$ depend only on the combinations $`(k_iu)`$ and $`\alpha `$. Furthermore, since the domain over which we are solving the equation is non-compact, the index $`\alpha `$ is continuous and the zeta function summation will take the form
$`\zeta (s)`$ $`=`$ $`{\displaystyle \lambda ^s}=\sigma _3{\displaystyle 𝑑\alpha 𝑑k_i\lambda (\alpha ,k_iu)^s};`$ (62)
$`=`$ $`\sigma _3u^2\stackrel{~}{\zeta }(s),`$ (63)
where $`\sigma _3`$ is again the regulated volume and in the latter equality $`\stackrel{~}{\zeta }(s)`$ is a function only of $`s`$. Since the partition function can depend only on $`\zeta ^{}(0)`$, it manifestly exhibits the same behaviour as the bulk action (56), in agreement with our suggestion for the interpretation of the bulk action. Suppose we interpret $`x_1`$ as the Euclidean time direction; then the thermodynamic relation is given by
$$I_{\mathrm{bulk}}(u)=\beta _{x_1}M(u),$$
(64)
where $`\beta _{x_1}`$ is the inverse temperature and the cutoff mass $`M(u)`$ is negative. The entropy vanishes, which implies that $`H^2\times H^2`$ corresponds in some sense to the ground state of the conformal field theory, but the energy is negative which we should probably interpret as discussed in .
## IV Correlation functions in the boundary CFT
In the previous section we considered how the bulk action corresponds to the partition function for the conformal field theory. The next question to ask is how the bulk supergravity action acts as a generating functional for the correlation functions of the conformal field theory. The analysis for the Bergman metric was carried out in ; the $`su(2,1)`$ symmetry of the bulk corresponds to a $`su(2,1)`$ conformal symmetry group of the boundary. This conformal symmetry is enough to fix the functional form of two-point functions of scalar operators entirely, and this form is reproduced from the action for bulk scalar fields. In particular, as in other cases of the AdS/CFT correspondence, fields of a particular mass $`m`$ and spin $`s`$ are found to correspond to scalar operators of definite conformal weights $`\mathrm{\Delta }(m,s)`$ in the boundary CFT.
Manifolds of degenerate boundary which fall into the first category of §II can hence be dealt with in much the same way as in the usual AdS/CFT correspondence. However, the analysis is different for spaces falling into the second category. These spaces are characterised by the existence of more than one infinite direction, not linked by the symmetry group. As we will discuss in this section, this means that massive scalar fields will give rise to boundary data which is a sum of data of different conformal weights; the relationship between the mass and the conformal weight in the CFT is more subtle. A secondary characteristic of these spaces is that the conformal symmetry group is not large enough to fix the form of even the two point functions completely.
### A Two point functions from conformal symmetry
We will consider here the simplest non-trivial example, $`H^2\times H^2`$. Since the symmetry group of the manifold is $`sl(2,R)\times sl(2,R)`$, which has a maximal compact subgroup of $`so(2)\times so(2)`$, the boundary has only the latter group of symmetries. Expressed in terms of the $`(u,\theta )`$ coordinates, the Killing vectors in the bulk are
$`k_1`$ $`=`$ $`_{x_1};k_2=_{x_2};`$ (65)
$`l_1`$ $`=`$ $`x_1_{x_1}+u\mathrm{cos}^2\theta _u\mathrm{cos}\theta \mathrm{sin}\theta _\theta ;`$ (66)
$`l_2`$ $`=`$ $`x_2_{x_2}+u\mathrm{sin}^2\theta _u+\mathrm{cos}\theta \mathrm{sin}\theta _\theta ;`$ (67)
$`m_1`$ $`=`$ $`(x_1^2u^2\mathrm{cos}^2\theta )_{x_1}+2x_1u\mathrm{cos}^2\theta _u2x_1\mathrm{cos}\theta \mathrm{sin}\theta _\theta ;`$ (68)
$`m_2`$ $`=`$ $`(x_2^2u^2\mathrm{sin}^2\theta )_{x_2}+2x_2u\mathrm{sin}^2\theta _u+2x_2\mathrm{cos}\theta \mathrm{sin}\theta _\theta .`$ (69)
If one restricts to the boundary $`u0`$, then the $`k_i`$ remain symmetries but the $`l_i`$ are conformal symmetries only. Notice that one does not need the inverse metric to define the conformal Killing vector equations and hence the conformal symmetries are well defined even without a non-degenerate metric.
Now let us consider how the two-point function of scalar fields $`𝒪_{\mathrm{\Delta }_1}(x)𝒪_{\mathrm{\Delta }_2}(\overline{x})`$ is fixed by the requirement of invariance under conformal transformations. Under a conformal transformation generated by $`\xi `$ a field of conformal weight $`\mathrm{\Delta }`$ will transform as
$$\delta _\xi 𝒪=(_\xi +\frac{\mathrm{\Delta }}{3}D_m\xi ^m)𝒪,$$
(70)
where $``$ is the Lie derivative. Then the requirement of invariance under the isometries $`k_i`$ implies that the two-point function only depends on the translationally invariant quantities $`(x_1\overline{x}_1)`$ and $`(x_2\overline{x}_2)`$. The requirement for the two-point function to be covariant under the transformations generated by the $`l_i`$ and $`m_i`$ is
$$[l_i^{(x)}+l_i^{(y)}]𝒪_{\mathrm{\Delta }_1}(x)𝒪_{\mathrm{\Delta }_2}(\overline{x})=\frac{1}{3}[\mathrm{\Delta }_1D_ml_i^{m(x)}+\mathrm{\Delta }_2D_ml_i^{m(y)}]𝒪_{\mathrm{\Delta }_1}(x)𝒪_{\mathrm{\Delta }_2}(\overline{x}).$$
(71)
Now in this equation we need the inverse metric to be finite in order to define the right-hand side. To do this we note that if we use the conformally rescaled boundary metric (57) discussed in §III then the metric determinant is independent of $`u`$ and
$$D_ml_i^{m(x)}=\mathrm{sin}^2\theta \mathrm{cos}^2\theta _m(\mathrm{sin}^2\theta \mathrm{cos}^2\theta l_i^m).$$
(72)
Note that both the non-degenerate measure and the metric itself have conformal dimension of minus three. Using the four conformal covariance conditions we can constrain the two-point function to be of the form
$$𝒪_{\mathrm{\Delta }_1}(x)𝒪_{\mathrm{\Delta }_2}(\overline{x})=𝑑\chi f(\chi ,\mathrm{\Delta }_1,\mathrm{\Delta }_2)\frac{\mathrm{sin}^{\frac{2\mathrm{\Delta }_1}{3}\frac{\chi }{2}}\theta \mathrm{sin}^{\frac{2\mathrm{\Delta }_2}{3}\frac{\chi }{2}}\overline{\theta }\mathrm{cos}^{\frac{\chi }{2}}\theta \mathrm{cos}^{\frac{\chi }{2}}\overline{\theta }}{(x_1\overline{x}_1)^\chi (x_2\overline{x}_2)^{\frac{2\mathrm{\Delta }_1}{3}+\frac{2\mathrm{\Delta }_2}{3}\chi }}.$$
(73)
As expected the conformal symmetry group is not large enough to fix the form of the two-point function completely. The function $`f`$ is not fixed by symmetry and furthermore conformal invariance does not fix $`\mathrm{\Delta }_1=\mathrm{\Delta }_2`$; fields of unequal conformal weight are not excluded from having a non-zero correlation function.
### B Scalar fields in the bulk
Now let us consider how this form for the two-point function is reproduced by the bulk theory. One of the most interesting differences between this bulk boundary correspondence and the usual non-degenerate correspondence is that a bulk scalar field of mass $`m`$ does not correspond to a single operator of weight $`\mathrm{\Delta }(m)`$. Instead, the scalar field acts as a source for a set of operators of weights which depend not only on $`m`$ but also on the “mode” of the scalar field.
One can easily understand how this arises by looking at explicit solutions of the field equation. The field equation for a free scalar field of mass $`m`$ is
$$[z_1^2(_{z_1}^2+_{x_1}^2)+z_2^2(_{z_2}^2+_{x_2}^2)m^2]\mathrm{\Phi }^m=0,$$
(74)
and so modes of the field behave as
$$\mathrm{\Phi }^m(z_1z_2)^{1/2}K_\nu (k_1z_1)K_{\sqrt{m^2\nu ^2}}(k_2z_2)e^{ik_1x_1+ik_2x_2},$$
(75)
where we have chosen Bessel functions such that modes are bounded at the interior points $`z_1,z_2\mathrm{}`$. The allowed values of $`m`$ are determined by considering the spectrum of supergravity on $`S^7`$; we find that (in the units used here) $`m^23/8`$. We then restrict the allowed values of $`\nu `$ to $`\mathrm{Re}(\nu )>0`$ to enforce boundedness in the interior. Note that most modes will consist of decaying oscillations in one hyperbolic space and exponential decay in the other.
We should briefly mention that since the space is supersymmetric there are no unstable fluctuations of the scalar field; the point is that although modes may be normalisable on one space they cannot be simultaneously normalisable on the other. If one considers eigenmodes of $`\mathrm{\Phi }(m)`$ with eigenvalues $`\lambda _k`$ then modes of negative $`\lambda _k`$ are not normalisable.
A more elegant way of expressing the above analysis is in terms of representation theory. Solutions of the wave equation for a massive scalar field form a representation of $`sl(2,R)\times sl(2,R)`$, which can be decomposed as products of representations of $`sl(2,R)`$ and $`sl(2,R)^{}`$ with Casimirs proportional to $`\nu ^2`$ and $`(m^2\nu ^2)`$ respectively. Suppose we then consider primary fields satisfying $`k_1\mathrm{\Psi }=k_2\mathrm{\Psi }=0`$, $`l_1\mathrm{\Psi }=h_1\mathrm{\Psi }`$ and $`l_2\mathrm{\Psi }=h_2\mathrm{\Psi }`$, which behave as
$$\mathrm{\Psi }u^{h_1h_2}\mathrm{cos}^{h_1}\theta \mathrm{sin}^{h_2}\theta .$$
(76)
The quadratic Casimir is
$$m^2\mathrm{\Psi }=(l_1^2+l_2^2\{k_1,m_1\}\{k_2,m_2\})\mathrm{\Psi }=[h_1(h_1+1)+h_2(h_2+1)]\mathrm{\Psi }.$$
(77)
The conformal weights with respect to the two $`sl(2,R)`$ conformal groups are thus related by the mass of the bulk scalar field, but are not fixed; this is the origin of the $`\chi `$ integration in (73). For fields of arbitrary spin $`s`$ the mass and conformal weight relation (77) becomes
$$m^2=[h_1(h_1+1)+h_2(h_2+1)]\frac{s^2}{2}.$$
(78)
Rewriting the scalar field in terms of the $`(u,\theta )`$ variables and taking the limit $`u0`$ we get
$$\mathrm{\Phi }^mu^{1\nu \sqrt{m^2\nu ^2}}(\mathrm{cos}^{1/2\nu }\theta \mathrm{sin}^{1/2\sqrt{m^2\nu ^2}}\theta k_1^\nu k_2^{\sqrt{m^2\nu ^2}}e^{ik_1x_1+ik_2x_2}).$$
(79)
Note that the $`u`$ dependence is in general complex depending on the value of $`\nu `$. Explicitly the conformal weights $`h_i`$ are given in these variables by
$$h_1=\nu \frac{1}{2};h_2=\sqrt{m^2\nu ^2}\frac{1}{2}.$$
(80)
The total conformal weight of the boundary data will be determined by the $`u`$ dependence and is not independent of $`\nu `$; hence different modes of a massive field will give rise to boundary data of different conformal weight.
The action for a free massive scalar field reduces to the boundary term:
$$I^{(m)}=𝑑x𝑑y\frac{d\theta }{u\mathrm{sin}^2\theta \mathrm{cos}^2\theta }\mathrm{\Phi }^m_u\mathrm{\Phi }^m,$$
(81)
where in this equation and all that follow we are suppressing constant factors. Fourier transforming (75), the massive scalar field can be written in terms of propagators on each hyperbolic space as
$$\mathrm{\Phi }^{(m)}=𝑑\nu 𝑑\overline{x}_1𝑑\overline{x}_2u^{\alpha +\beta }\frac{\mathrm{cos}^\alpha \theta }{(u^2\mathrm{cos}^2\theta +(\mathrm{\Delta }x_1)^2)^\alpha }\frac{\mathrm{sin}^\beta \theta }{(u^2\mathrm{sin}^2\theta +(\mathrm{\Delta }x_2)^2)^\beta }\mathrm{\Phi }^{(m)}(\nu ,\overline{x}_1,\overline{x}_2),$$
(82)
where $`\alpha =\frac{1}{2}+\nu `$ and $`\beta =\frac{1}{2}+\sqrt{m^2\nu ^2}`$. In the limit $`u0`$,
$$\mathrm{\Phi }^{(m)}𝑑\nu u^{1\nu \sqrt{m^2\nu ^2}}\mathrm{cos}^\alpha \theta \mathrm{sin}^\beta \theta \mathrm{\Phi }^{(m)}(\nu ,x_1,x_2).$$
(83)
In these expressions we are drawing on the by now well known propagators first discussed in . Note that we have not corrected the normalisation of the propagators following to ensure the right coefficients as $`u0`$; in this expression, and all that follow, we will ignore $`\nu `$ dependent normalisation factors. We should allow the $`\nu `$ integration to run over all possible values. Furthermore,
$`_u\mathrm{\Phi }^{(m)}`$ $``$ $`{\displaystyle }d\nu d\overline{x}_1d\overline{x}_2[{\displaystyle \frac{(\alpha +\beta )u^{\alpha +\beta 1}}{(\mathrm{\Delta }x_1)^{2\alpha }(\mathrm{\Delta }x_2)^{2\beta }}}{\displaystyle \frac{2\alpha u^{1+\alpha +\beta }\mathrm{cos}^2\theta }{(\mathrm{\Delta }x_1)^{2(\alpha +1)}(\mathrm{\Delta }x_2)^{2\beta }}}`$ (85)
$`{\displaystyle \frac{2\beta u^{1+\alpha +\beta }\mathrm{sin}^2\theta }{(\mathrm{\Delta }x_1)^{2\alpha }(\mathrm{\Delta }x_2)^{2(\beta +1)}}}]\mathrm{cos}^\alpha \theta \mathrm{sin}^\beta \theta \mathrm{\Phi }^{(m)}(\nu ,x_1,x_2)`$
Since $`\mathrm{Re}(\alpha )>0`$ and $`\mathrm{Re}(\beta )>0`$, the first term is of leading order as $`u0`$. However, as we shall see below, we cannot neglect the subleading terms in this case, since these will give finite contributions to two point functions even as $`u0`$.
It is convenient at this stage to rewrite the last two integrals as integrals over conformal weight of the boundary data, since we will eventually want to compare predictions for two point functions with the boundary theory expectations. Then
$$\mathrm{\Phi }^{(m)}(u,\theta ,x_1,x_2)𝑑\lambda u^{2\lambda /3}Y^{(m)}(\lambda ,\theta )\mathrm{\Phi }^{(m)}(\lambda ,x_1,x_2),$$
(86)
where as we will see $`\lambda `$ is the conformal weight of the boundary data and we introduce the “eigenfunctions”
$$Y^{(m)}(\lambda ,\theta )=\mathrm{cos}^\alpha \theta \mathrm{sin}^\beta \theta ,$$
(87)
where $`(\alpha +\beta )=\frac{2\lambda }{3}+2`$ and in addition
$$(\alpha \frac{1}{2})^2=m^2(\beta \frac{1}{2})^2.$$
(88)
It is also helpful to introduce the notation
$$K^{(m)}(\alpha ,\beta ,\theta ,\mathrm{\Delta }x_1,\mathrm{\Delta }x_2)=\frac{\mathrm{cos}^\alpha \theta \mathrm{sin}^\beta \theta }{(\mathrm{\Delta }x_1)^{2\alpha }(\mathrm{\Delta }x_2)^{2\beta }},$$
(89)
and to simplify notation further we will suppress coordinate dependence where obvious from now on. Then,
$`_u\mathrm{\Phi }^{(m)}`$ $``$ $`{\displaystyle }d\lambda d\overline{x}_1d\overline{x}_2[(2+{\displaystyle \frac{2\lambda }{3}})u^{1+2\lambda /3}K^{(m)}(\alpha ,\beta )2\alpha u^{3+\frac{4\lambda }{3}}K^{(m)}((\alpha +1),\beta )\mathrm{cos}\theta `$ (91)
$`2\beta u^{3+\frac{4\lambda }{3}}K^{(m)}(\alpha ,(\beta +1))\mathrm{sin}\theta ]\mathrm{\Phi }^{(m)}(\lambda ,\overline{x}_1,\overline{x}_2).`$
This action is of the form
$`I^{(m)}`$ $`=`$ $`{\displaystyle }d\sigma d\lambda d\overline{x}_1d\overline{x}_2d\overline{\lambda }u^{2\overline{\lambda }/32\lambda /3}Y^{(m)}(\lambda ,\theta )\mathrm{\Phi }^{(m)}(\lambda )\{(2+{\displaystyle \frac{2\overline{\lambda }}{3}})K^{(m)}(\overline{\alpha },\overline{\beta })`$ (93)
$`2u^2(\overline{\alpha }K^{(m)}(\overline{\alpha }+1,\overline{\beta })+\overline{\beta }K^{(m)}(\overline{\alpha },(\overline{\beta }+1)))\}\mathrm{\Phi }^{(m)}(\overline{\lambda }),`$
where $`d\sigma `$ is the non-degenerate measure on the boundary and $`\alpha ,\beta `$ satisfy the constraints $`(\overline{\alpha }+\overline{\beta })=2+\frac{2\overline{\lambda }}{3}`$ as well as the constraint (88).
The bulk/boundary correspondence tells us that the bulk scalar field acts as a source for scalar operators in the boundary theory. One should hence associate the bulk action with terms in the conformal field theory action of the form
$$I=𝑑\lambda 𝑑\mathrm{\Delta }𝑑\sigma u^{2+\frac{2}{3}\mathrm{\Delta }\frac{2}{3}\lambda }\mathrm{\Phi }(\lambda ,x)𝒪_\mathrm{\Delta }(x),$$
(94)
where $`\lambda `$ is the conformal weight of the boundary scalar field data and $`\mathrm{\Delta }`$ is the conformal weight of the operator $`𝒪`$. The $`u`$ dependence of this action is determined by the requirement of conformal invariance: suppose that the scalar field data behaves as
$$\mathrm{\Phi }(u,x)u^\lambda \mathrm{\Phi }^b(x),$$
(95)
on the boundary. That is, the data scales with $`u`$ which is a positive function that has a simple zero on the boundary. Following the same arguments as in , the definition of $`\mathrm{\Phi }^b(x)`$ depends on our particular choice of function, and if we transform $`ue^wu`$ then $`\mathrm{\Phi }^b(x)e^{w\lambda }\mathrm{\Phi }^b(x)`$. Under the same transformation the measure transforms as $`d\sigma e^{2w}d\sigma `$, since the measure is of conformal weight $`3`$. The degeneracy of the boundary implies that an operator of conformal weight $`\mathrm{\Delta }`$ scales as $`e^{2w\mathrm{\Delta }/3}`$ and since the action must be conformally invariant this implies that it must be of the form (94).
Comparing the forms of (94) and (93) we see that we must identify
$`𝒪_\mathrm{\Delta }(x)`$ $`=`$ $`{\displaystyle }d\overline{x}_1d\overline{x}_2[({\displaystyle \frac{2}{3}}\mathrm{\Delta })K^{(m)}(\alpha _1,\beta _1)\mathrm{\Phi }^{(m)}(\mathrm{\Delta }3)`$ (97)
$`(2\alpha _2K^{(m)}(\alpha _2+1,\beta _2)\mathrm{cos}\theta +2\beta _2K^{(m)}(\alpha _2,\beta _2+1)\mathrm{sin}\theta )\mathrm{\Phi }^{(m)}(\mathrm{\Delta }6)],`$
where
$$\alpha _1+\beta _1=\frac{2\mathrm{\Delta }}{3};\alpha _2+\beta _2=\frac{2\mathrm{\Delta }}{3}2.$$
(98)
In addition $`\alpha _i,\beta _i`$ satisfy the constraint (88). This gives us the expectation value of the operator and functionally differentiating this expression again will give us the two-point functions. Writing the boundary value of the scalar field as the mode expansion (86), then this expression may be inverted as
$$\mathrm{\Phi }^{(m)}(\lambda ,x_1,x_2)=u^{\frac{2\lambda }{3}}\frac{d\theta }{\mathrm{sin}^2\theta \mathrm{cos}^2\theta }Y^{(m)}(\lambda ,\theta )\mathrm{\Phi }^m(u,x),$$
(99)
where we are again ignoring $`\lambda `$ dependent normalisation factors; in fact, to get the normalisation factors right, we would have to say more carefully how we are going to regularise these formally divergent integrals. The two-point function $`𝒪_\mathrm{\Delta }(x)𝒪_\mathrm{\Delta }^{}(y)`$ is given by functionally differentiating (97) with respect to
$$u^{2+\frac{2\mathrm{\Delta }^{}}{3}}\mathrm{\Phi }^{(m)}(u,x).$$
(100)
Let us consider the result of differentiating the first term in (97); then the correlation function is only non-zero for $`\mathrm{\Delta }=\mathrm{\Delta }^{}`$ and we get a two-point function
$$𝒪_\mathrm{\Delta }(x)𝒪_\mathrm{\Delta }(\overline{x})=\frac{2\mathrm{\Delta }}{3}\frac{\mathrm{cos}^{\alpha _1}\theta \mathrm{cos}^{\alpha _1}\overline{\theta }\mathrm{sin}^{\beta _1}\theta \mathrm{sin}^{\beta _1}\overline{\theta }}{(\mathrm{\Delta }x_1)^{2\alpha _1}(\mathrm{\Delta }x_2)^{2\beta _1}}.$$
(101)
(101) gives the leading order behaviour of the correlation function for operators of the same conformal weight as $`u0`$. Subleading behaviour is derived from the last two terms in (97)
$`𝒪_\mathrm{\Delta }(x)𝒪_\mathrm{\Delta }(\overline{x})`$ $`=`$ $`u^2\{2\alpha _2{\displaystyle \frac{\mathrm{cos}^{\alpha _2+2}\theta \mathrm{cos}^{\alpha _2}\overline{\theta }\mathrm{sin}^{\beta _2}\theta \mathrm{sin}^{\beta _2}\overline{\theta }}{(\mathrm{\Delta }x_1)^{2(\alpha _2+1)}(\mathrm{\Delta }x_2)^{2\beta _2}}}`$ (103)
$`+2\beta _2{\displaystyle \frac{\mathrm{cos}^{\alpha _2}\theta \mathrm{cos}^{\alpha _2}\overline{\theta }\mathrm{sin}^{\beta _2+2}\theta \mathrm{sin}^{\beta _2}\overline{\theta }}{(\mathrm{\Delta }x_1)^{2\alpha _2}(\mathrm{\Delta }x_2)^{2(\beta _2+1)}}}\}.`$
However, differentiating the last two terms in (97) also gives a non-zero leading order contribution to the correlation function
$`𝒪_\mathrm{\Delta }(x)𝒪_{\mathrm{\Delta }3}(\overline{x})`$ $`=`$ $`\{2\alpha _2{\displaystyle \frac{\mathrm{cos}^{\alpha _2+2}\theta \mathrm{cos}^{\alpha _2}\overline{\theta }\mathrm{sin}^{\beta _2}\theta \mathrm{sin}^{\beta _2}\overline{\theta }}{(\mathrm{\Delta }x_1)^{2(\alpha _2+1)}(\mathrm{\Delta }x_2)^{2\beta _2}}}`$ (105)
$`+2\beta _2{\displaystyle \frac{\mathrm{cos}^{\alpha _2}\theta \mathrm{cos}^{\alpha _2}\overline{\theta }\mathrm{sin}^{\beta _2+2}\theta \mathrm{sin}^{\beta _2}\overline{\theta }}{(\mathrm{\Delta }x_1)^{2\alpha _2}(\mathrm{\Delta }x_2)^{2(\beta _2+1)}}}\}.`$
That is, operators of unequal conformal weight have a non-vanishing two point function! Now (101) and (105) have precisely the same form as terms in the integral (73); one can check that the equivalent between indices in the two expressions. Thus we have explicitly verified that the scalar two point functions are reproduced from the action for bulk scalar fields.
For completeness, let us now sketch the principle features of the correspondence for the coset space $`SO(3,1)/SO(2)`$ given in (15). The isometries of this space are generated by the $`sl(2,R)\times sl(2,R)\times so(2)`$ algebra
$$l_0^i=i_{\varphi _i};l_{\pm 1}^i=ie^{\pm \varphi _i}(\mathrm{coth}\rho _i_{\varphi _i}i_{\rho _i});k=i_\psi .$$
(106)
One can find primary fields satisfying
$$l_0^i\mathrm{\Psi }=h^i\mathrm{\Psi };k\mathrm{\Psi }=h_k\mathrm{\Psi };l_1^i\mathrm{\Psi }=0,$$
(107)
such that the bulk mass is related to these conformal weights as
$$m^2=h_i(h_i+1)+h_k^2.$$
(108)
By analogy to the above analysis for $`H^2\times H^2`$, if one considers the correlation function of two operators of conformal weights $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$ on the hypersurface $`u0`$ defined from (18), then
$$𝒪_{\mathrm{\Delta }_1}(x)𝒪_{\mathrm{\Delta }_2}(\overline{x})=\underset{n,l}{}f_{n,l}(\mathrm{\Delta }_1,\mathrm{\Delta }_2)e^{in\mathrm{\Delta }\psi }U_l(\theta ,\overline{\theta },\mathrm{\Delta }\varphi _i,\mathrm{\Delta }_1,\mathrm{\Delta }_2),$$
(109)
where the coefficients $`f_{n,l}`$ are not fixed by the conformal symmetry but conformal covariance requires that $`U_l`$ satisfies four equations of the form
$`i(_{\varphi _1/\varphi _2}i\mathrm{sin}\theta \mathrm{cos}\theta _\theta +e^{i\mathrm{\Delta }\varphi _1/(\mathrm{\Delta }\varphi _2)}(_{\overline{\varphi }_1/\overline{\varphi }_2}i\mathrm{sin}\overline{\theta }\mathrm{cos}\overline{\theta }_{\overline{\theta }})U_l`$ (110)
$`={\displaystyle \frac{2}{3}}(\mathrm{\Delta }_1\mathrm{sin}^2\theta +\mathrm{\Delta }_2\mathrm{sin}^2\overline{\theta }e^{i\mathrm{\Delta }\varphi _1/(\mathrm{\Delta }\overline{\varphi }_2)})U_l;`$ (111)
$`i(_{\varphi _1/\varphi _2}+i\mathrm{sin}\theta \mathrm{cos}\theta _\theta +e^{(i\mathrm{\Delta }\varphi _1)/\mathrm{\Delta }\varphi _2}(_{\overline{\varphi }_1/\overline{\varphi }_2}+i\mathrm{sin}\overline{\theta }\mathrm{cos}\overline{\theta }_{\overline{\theta }})U_l`$ (112)
$`={\displaystyle \frac{2}{3}}(\mathrm{\Delta }_1\mathrm{cos}^2\theta +\mathrm{\Delta }_2\mathrm{cos}^2\overline{\theta }e^{(i\mathrm{\Delta }\varphi _1)/\mathrm{\Delta }\overline{\varphi }_2})U_l.`$ (113)
Terms in the summation (109) should then be reproduced by considering the action for bulk massive scalar fields.
It is interesting to note that although the bulk scalar field action for $`H^2\times H^2`$ contains terms which diverge as $`u0`$ the two-point functions are actually regular in this limit. The same behaviour was found for the Bergman metric; in fact in this case the action for a massive scalar is independent of the IR regularisation parameter . Thus, we don’t need to keep $`u`$ finite in the correlation functions.
If we keep $`u`$ finite in the boundary field theory, the two point functions will still be fixed by the conformal symmetry group of a constant $`u`$ hypersurface. To compare with the bulk theory, we should introduce propagators describing sources at finite $`u`$ and repeat the above analysis. It would be interesting to consider the flow of the propagators as one changes $`u`$, particularly in the context of making more precise the correspondence at finite $`u`$. One could also compare the spectrum on $`H^2\times H^2`$ with operators appearing in the boundary theory.
###### Acknowledgements.
I would like to thank Finn Larsen for discussions in the early stages of this work and to thank Harvard University, where this work was begun, for hospitality. Financial support for this work was provided by St John’s College, Cambridge.
## Effective actions on the squashed three sphere
Eigenvalues of the scalar operator (54) used in are
$$\lambda =\frac{1}{4l_3^2}(n^2+4(l_3^21)(q+\frac{1}{2})(nq\frac{1}{2}),$$
(114)
with degeneracy $`n=1,\mathrm{}`$. We have set $`l_3^2=l^2R^2`$ and $`q`$ runs from $`0`$ to $`(n1)`$. For our conformally coupled operator (55) we just need to shift the eigenvalues, and can hence write the partition function as $`W_{\mathrm{sc}}=\frac{1}{2}\zeta ^{}(0)`$ where
$$\zeta (s)=(2l_3)^{2s}\underset{n=1}{\overset{\mathrm{}}{}}\underset{q=0}{\overset{n1}{}}\frac{n}{(n^2+4(l_3^21)(q+\frac{1}{2})(nq\frac{1}{2})\frac{l_3^4}{4})^s}.$$
(115)
The approach of was to apply the Plana summation formula to the $`q`$ summation; in the extreme oblate limit $`l_3\mathrm{}`$ it is then quite easy to find the dominant term in the zeta function which behaves as $`l_3^4`$.
Following the same approach here, the key point is that, although sub-dominant terms in the Plana summation formula are affected by the shift in eigenvalues, the dominant term in the extreme prolate limit is determined by a very similar term to that in
$$\zeta (s)2i(2l_3)^{2s}_0^{\mathrm{}}\frac{dt}{exp(2\pi t)+1}\{\frac{n}{(n^2+4(l_3^21)(t^2itn)\frac{l_3^4}{4})^s}(tt)\},$$
(116)
which we can analyse by the Watson-Sommerfeld method to give a leading order contribution of
$$W_{\mathrm{sc}}\frac{3l_3^4}{2\pi ^2}\zeta _R(3),$$
(117)
as was found in . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.